Amazon’s software play - Tech2025 · NTAP, IBM, ORCL, HPE, Dell/ EMC, Veritas GOOG, MSFT, IBM, AKAM High AWS dominates storage as a service with scale, pricing and breadth Networking

Software Sector outlook

Amazon Web Services - Growth at scale

Source: CLSA, company

Find CLSA research on Bloomberg, Thomson Reuters, Factset and CapitalIQ - and profit from our evalu@tor proprietary database at clsa.com

Produced by CLSA Americas LLC. For important disclosures please refer to page 72.

0

10

20

30

40

50

60

70

80

0

5,000

10,000

15,000

20,000

25,000

2014 2015 16CL 17CL 18CL

Revenue (LHS) Revenue growth(US$m) (%)

Ed Maguire [email protected] +1 212 549 8200

Reagan Tangney +1 212 549 5028

26 September 2016

USA Technology

Akamai BUY Target US$65 Amazon¹ O-PF Target US$890 Citrix O-PF Target US$99 Microsoft O-PF Target US$60 Oracle U-PF Target US$42 PTC O-PF Target US$45 Red Hat BUY Target US$94 Salesforce BUY Target US$101 ServiceSource U-PF Target US$5 Splunk BUY Target US$86 Teradata U-PF Target US$30 Trend Micro O-PF Target ¥4,400

¹ covered by CLSA analyst James Lee

www.clsa.com

010110100101101001011010010110100101101001011010010110100101101001011010010110100101101001011010010110100101101001011010010110100101101001011010010110100101101001011010010110100101101001011010010110100101101001011010010110100101101001011010010110100101101001011010010110100101101001011010010110100101101001011010010110100101101001011010010110100101

AWS’s quiet arsenal Amazon software reshapes industry landscape Amazon Web Services (AWS) is arguably the most consequential force in enterprise IT today, with its US$12bn-plus Infrastructure-as-a-Service (IaaS) business dominating the public cloud IaaS segment and disrupting the incumbent compute and storage industry. Building on its dominant position, AWS is expanding into higher-value software services as its RedShift and Aurora offerings marginalize legacy on-premise vendors, such as Underperform-rated Oracle and Teradata.

AWS opportunity moves “up the stack” Amazon dominates IaaS, which IDC forecasts to reach US$43.6bn in 2020, with over 50% share. The market for Software as a Service (SaaS) and cloud software (eg, Software Apps as a Service, System Infrastructure Software as a Service, Platform as a Service [PaaS]) is projected to be US$152bn in 2020. With less than 2% share to date here, AWS has plenty of room to expand.

Data(base) is the key to expansion Data stored in AWS provides a vector to drive adoption of higher-value software services, database in particular. AWS’s Aurora and RedShift are its fastest-growing services, emerging as formidable competition. We expect AWS to focus aggressively on growing share in the US$34bn-plus database market directly against incumbents Oracle, Teradata, IBM and Microsoft.

Making platforms count AWS’s growing portfolio addresses adjacent markets including Internet of Things (IoT) platforms, content delivery networks (CDN) and application delivery controllers (ADCs). Its IoT suite is positioned to drive incremental demand for infrastructure services; CloudFront CDN targets mid-market AWS customers; and application load balancing indirectly targets the ADC market.

Above the PaaS layer - towards user apps AWS is making an aggressive push into business intelligence (BI) with its QuickSight offering. With a formal launch later this year, we expect this product to have significant impact on vendors in the traditional BI market. The company’s forays into virtual desktops (WorkSpaces), mail (WorkMail) and collaboration (WorkDocs) appear to be meeting less success.

http://www.clsa.com/


https://www.clsa.com/member/analysts/index.cfm?alogin=MAGUIREE&pagename=bios


Software

2 [email protected] 26 September 2016

Contents

AWS’s quiet arsenal .......................................................................... 3

AWS opportunity moves “up the stack” ............................................. 5

Data(base) is the key to expansion ..................................................16

Making platforms count ...................................................................34

Above the PaaS layer - towards user apps .........................................48

All prices quoted herein are as at close of business 21 September 2016, unless otherwise stated

Insights into disruptive trends

See also 2020 in sight for our work on innovation and key themes


https://www.clsa.com/member/report/572724066




Investment thesis Software

26 September 2016 [email protected] 3

AWS’s quiet arsenal Amazon (O-PF by CLSA analyst James Lee) has arguably emerged as the most consequential force in enterprise technology, a pioneer and the dominant leader in the market for public cloud Infrastructure as a Service. IDC expects this market to reach US$43.6bn by 2020, and Amazon Web Services holds dominant share for compute and storage as a service. The market for SaaS and cloud software is much larger, likely to reach US$152bn by 2020, and AWS has barely any presence so far.

AWS opportunity moves “up the stack” Building on its dominance in infrastructure and the strategic advantages that come from storing mission critical data, Amazon is “moving up the stack”, expanding its offerings increasingly into higher-value software categories - notably database. We’ve already seen the negative impact that AWS has (indirectly had) on server and storage vendors, including Oracle and Teradata, both of which we rate Underperform. It’s our view that database vendors are already feeling the impact of Amazon’s growing market presence.

We summarize AWS’s offerings, key competitors on premise and in the cloud and the estimated competitive impact.

Competitive matrix

Category Services On-premise vendors

Cloud vendors

Competitive impact

Comments

Compute EC2, EC2 Container Registry, Elastic Beanstalk, Lambda, Auto Scaling and Load Balancing

HP, IBM, ORCL, Dell

MSFT, GOOG, IBM, RAX, ATT, VZ, others

High AWS dominates compute as a service with scale, pricing and breadth

Storage & content delivery

S3, CloudFront, Amazon EBS, Elastic File Storage, Glacier, Import/Export Snowball and AWS Storage Gateway

NTAP, IBM, ORCL, HPE, Dell/ EMC, Veritas

GOOG, MSFT, IBM, AKAM

High AWS dominates storage as a service with scale, pricing and breadth

Networking Virtual Private Cloud, Direct Connections, Elastic Load Balancing, Application Load Balancing and Domain Name System (Route 53)

CSCO, JNPR, HPE, FFIV, CTXS, RADW

VRSN, NSR, CSCO, CTXS

Low to medium

AWS's direct impact has been minimal, but product roadmap increasingly targets the application delivery services market

Management tools

CloudWatch, CloudFormation, CloudTrail, Config, OpsWorks, Trusted Advisor

CA, IBM, VMW, MSFT, RHT, HPE, CSCO, SPLK

VMW, MSFT, ORCL, IBM, NOW, NEWR, RHT, SPLK

Low AWS management tools targeted to AWS workloads

Security & identity

Identity and Access Management, Certificate Manager, CloudHSM, Key Management Service, Directory Service, Inspector and Web Application Firewall

EMC/RSA, CA, IBM, CYBR

AKAM, IMPV, VRSN, QLYS, GDDY

Low AWS security and identity tools address AWS user needs

Database RDS (Aurora, MySQL, PostgreSQL, Oracle, SQL Server, MariaDB), Database Migration Service, DynamoDB, ElastiCache, RedShift

ORCL, IBM, MSFT, SAP, TDC, HDP

MSFT, GOOG, ORCL

Medium to high

AWS's Database-as-a-Service offerings are aggressively capturing incumbent share

Analytics QuickSight, RedShift, Machine Learning, Kinesis, Elasticsearch, Elastic MapReduce and Data Pipelines

DATA, TDC, IBM, SAP, ORCL, MSFT

MSFT, GOOG, CRM

Medium Data Warehouse as a Service is having significant market impact, but business intelligence has not entered the market

Enterprise applications

WorkSpaces, WorkMail, WorkDocs CTXS, MSFT, OTEX

MSFT, GOOG, CTXS, BOX

Low AWS footprint is limited, cloud competition well established

Internet of Things

AWS IoT, Device SDK, Registry, Device Shadows and Rules Engine

MSFT MSFT, GE, GOOG, CRM

Low to medium

Market is nascent but AWS offerings are highly competitive

Source: CLSA, Amazon

Expanding offerings into higher-value

software categories

Reshaping the industry landscape


Investment thesis Software


Data(base) is the key to expansion Data stored in AWS’s S3 service provides a vector to drive adoption of higher-value software services, database in particular. AWS’s Aurora and RedShift are its fastest-growing services, emerging as formidable competition in the relational database, data warehouse and Hadoop-as-a-Service markets.

Amazon has differentiated itself in the US$34bn-plus database market and the data warehouse market. IDC estimates Amazon’s total revenue from the database market was about US$580m in 2015, while management has disclosed >US$1bn annual run rate from database revenue including relational database services (RDS). Amazon’s database services grew 167% YoY in a market that saw only 1% growth. By 2019 we see the market expanding to US$45bn with AWS’s share growing to 8.6% with US$3.96bn in revenue. We expect AWS to focus aggressively on gaining share in the database market directly against incumbents Oracle, Teradata and IBM, in particular.

Making platforms count AWS’s growing portfolio addresses adjacent markets including Internet of Things platforms, content delivery networks and application delivery controllers. Its scale, breadth and market reach position it to be a leading IoT platform, accelerating adoption of IoT solutions and spurring incremental demand across the portfolio of cloud infrastructure services. We estimate AWS IoT could generate US$200m with an additional US$1.0-1.6bn in related cloud services by 2020, with major competition from Microsoft and Google. CloudFront CDN, by our estimates, is the second-largest CDN after Akamai, but the service predominantly targets mid-market AWS customers. Application load balancing is a new market that AWS addresses indirectly. IDC estimates about US$500m revenue addressing ADC-type use cases. We see potential impact to F5 Networks and Citrix as a medium-term risk.

Above the PaaS layer - towards user applications AWS is making an aggressive push into business intelligence with its QuickSight offering. With the formal launch later this year, we expect the as-a-service model to exert pricing pressure on traditional BI offerings from other vendors including SAP, Oracle and IBM. We’d expect a similar revenue ramp to what we saw with RedShift (US$25m the first year growing 200%-plus). However, Microsoft has beaten AWS to market with PowerBI Pro with lower pricing and a richer feature set. We expect both vendors to use respective BI offerings to drive adoption of their broader cloud portfolios.

AWS’s forays into virtual desktops (WorkSpaces), mail (WorkMail) and collaboration (WorkDocs) appear to be meeting less success. WorkDocs and WorkMail target a US$13bn market for collaborative applications in which Microsoft’s 35% share leads Google at 7.8% in a segment with hundreds of competitors. Amazon is a new entrant with just US$31m in estimated 2015 revenue (0.2% share) and 86% YoY growth. AWS WorkMail and WorkDocs are neither the cheapest, the fastest growing nor the most richly featured, but provide integrated extensions to the WorkSpaces Desktop-as-a-Service offering. AWS WorkSpaces does not appear to have made much of a dent in the market for hosted desktops. We estimate AWS could capture 5% of the 15%-Cagr Desktop-as-a-Service (DaaS) market by 2020, but given technological, organizational and structural advantages, Citrix (and Microsoft) will likely sustain dominant share.

IoT, CDN and ADC are smaller opportunities,

with less risk to incumbents for now

Aggressive push into business intelligence

Forays into virtual desktops, mail and

collaboration meeting less success

Most successful - and aggressive - targeting

the database market


Section 1: AWS opportunity moves “up the stack” Software


AWS opportunity moves “up the stack” Since the debut of AWS in 2006, Amazon has emerged as a powerful and transformative force in enterprise technology. The delivery of compute, storage and service at ever-declining price points has disrupted incumbent vendors in the traditional on-premise enterprise IT market. The AWS strategy for the first 10 years has been characterized by consistent price declines - over 50 since services were introduced - and price transparency. AWS is more than just compute and storage, however. The economics of Infrastructure as a Service exposed the weaknesses of traditional on-premise IT and have cut a swathe of disruption through the industry. AWS is accelerating the pace of innovation, adding 422 new services and features in 1H16 versus 722 in all of 2015. The vast majority of AWS revenue comes from Infrastructure as a Service (compute, storage and networking) and cloud software represents attractive opportunity for incremental growth of higher-margin services.

Figure 1

Worldwide public cloud Infrastructure-as-a-Service revenue snapshot

Source: IDC

AWS has over 1 million active customers per month, a US$11bn run rate and 58% growth YoY in 2Q16. The revenue reflects capacity that customers are actually using, not payment for unused capacity, whether they are new applications or services. There are also a large number of startups including Pinterest, AirBnB, Tinder, Twilio, Artsy, Wix and others. There is a growing cohort of enterprise companies across verticals. There are over 2,300 government customers. AWS currently has 13 regions and 35 availability zones (data centers on different flood plains, etc).

Amazon has emerged as a powerful and

transformative force in enterprise technology

AWS has >1 million active customers per month, a

US$11bn run rate and 58% growth YoY in 2Q16

The economics of Infrastructure as a

Service exposed the weaknesses of traditional

on premise IT




AWS has seen robust growth in the business overall, with revenue of US$2.5bn in 1Q16 (+64% YoY). AWS revenue growth accelerated from 49% in 2014 to 69% 2015, while operating margin increased from 14% in 2014 to 28% in 1Q16.

Figure 2

AWS revenue growing rapidly

Source: CLSA, company

“Your margin is my opportunity” - Jeff Bezos, CEO of Amazon Infrastructure as a Service is an estimated US$12.6bn market, which IDC forecasts to grow to US$43.6bn by 2020. AWS has defined the market and sprinted to early dominance. Synergy Research Group estimates Amazon’s share of the IaaS market at 31% in 2015, over three times the second-largest vendor Microsoft with 9%. Amazon achieved its dominant share through aggressive pricing tactics; however, since it has begun disclosing financial specifics including revenue growth and operating margins, the pace of price cuts has abated.

Amazon’s initial strategy was focused on share capture based on aggressive pricing, which helped the company take commanding share of the emerging IaaS market. It has cut prices 51 times since 2006 with the most recent in January 2016. However, the rate of price cuts is slowing. AWS had 42 price cuts over its first eight years with an additional nine since then. Three were announced in 2015, following an aggressive move in 2014 when AWS cut EC2 prices by up to 40% in some cases. The 2016 price cuts were less than 5% in many cases.

AWS, leading the trend towards public cloud computing services, has left a trail of casualties in its wake. Computing hardware/software vendors that have been impacted include HP, IBM, Sun (acquired by Oracle), Dell and VMware. Networking vendors including Cisco, Juniper and Lucent have seen growth slow. In storage, EMC agreed to be acquired by Dell, which went private in 2015, while NetApp has struggled. Even consulting and systems integration vendors like EDS have felt the impact.

What does this mean? We believe there are a couple of factors in play:

One, Amazon’s lead in the market is so dominant and its advantages of scale so great against all but Microsoft and Google that there are declining advantages from every marginal cost cut.

0

10

20

30

40

50

60

70

80

0

5,000

10,000

15,000

20,000

25,000

2014 2015 16CL 17CL 18CL

Revenue (LHS) Revenue growth(US$m) (%)

AWS has left a trail of casualties in its wake

The pace of price cuts is decelerating

Amazon achieved its dominant share through

aggressive pricing tactics

AWS growth has been impressive




Two, profitability is an increasing focus with the quarterly disclosure of AWS’s revenue and margins, hence the growing emphasis on higher-value offerings like platforms, database and analytics.

We’ve explored Amazon’s non-infrastructure offerings in a series of research reports to get a better sense of what existing vendors may or may not be prone to further disruption. As we explored in our 2013 CLSA U Blue Book Clouducopia and our 2013 report Soft machines, we had expected disruption to migrate “up the stack” into middleware with the emergence of Platform as a Service. We’ve already seen middleware vendors Tibco and Informatica go private. Database vendors including Oracle, IBM and particularly Teradata have struggled to accelerate their cloud offerings as revenue from on-premise licenses declines.

There’s been an important shift in adoption drivers for EC2. According to an annual IDC survey, leading drivers for adoption have consistently been reduction of development cost and reduction of infrastructure IT cost. What has changed is the growing attraction of AWS’s application ecosystem, the services in AWS.

From data storage and compute, it’s a simple step to the next market: database. Amazon introduced SimpleDB in December 2007, RDS supporting MySQL databases in 2009, DynamoDB in 2012, RedShift in 2012 and Aurora in 2014. In October 2015, Amazon disclosed that database-related revenue was roughly US$1bn/year.

In April 2015, Amazon characterized RedShift as its “fastest-growing product ever”, and then in June 2016 the company characterized the newer Aurora database as the fastest-growing service. By IDC’s estimates, Amazon’s relational database revenue grew 360% YoY to US$380m in 2015, with non-relational database revenue growing 77% YoY to US$134m in 2014 (we’d estimate the run rate over US$200m in 2015). We expect database services to remain one of AWS’s top competitive priorities.

Once data is managed in a database, the most common use case in AWS appears to be analytics. Analytics is the most frequently cited workload running on EC2, according to a 2016 survey of 287 respondents by IDC, followed by non-relational databases and relational databases.

Cloud software - larger than Infrastructure as a Service Put simply, the market for SaaS and cloud software (Software Apps as a Service, System Infrastructure Software as a Service, Platform as a Service) is over five times the size of the IaaS market that AWS already dominates. Applications represent the largest share of this market, with rapidly growing Platform as a Service set to surpass system Infrastructure Software as a Service in 2016. We expect AWS to continue to introduce higher-value services, including application development, collaboration apps, mobile application services, database, data management and analytics tools to tap larger markets.

Up the stack, that’s where the money is

Demand drivers evolve past cost savings to the ecosystem

It’s a simple step from data storage to database

From the database, the next step is analytics




Figure 3

Worldwide Software as a Service and cloud software revenue snapshot

Selected segment growth rate (%)

Applications as a Service Cagr 16.3

SIS as a Service Cagr 16.2

SaaS as a Service Cagr 16.2

PaaS as a Service Cagr 30.2

Total market Cagr 18.6 Source: IDC

Over the next five years, IDC forecasts cloud software as a proportion of the total software market to increase from 17% in 2015 to 29% in 2020.

Figure 4

Worldwide cloud software as a function of overall software market

Source: IDC

46.955.7

65.475.9

87.399.68.8

10.5

12.3

14.3

16.5

18.7

8.9

12.3

16.5

21.4

26.9

33.3

0

5

10

15

20

25

30

0

20

40

60

80

100

120

140

160

2015 2016 2017 2018 2019 2020

Application as a Service SIS as a ServicePlatform as a Service Total growth (RHS)

(US$bn) (%)

SaaS and cloud software was a US$78bn market

in 2015

Applications and platforms are migrating

to the cloud




Clearly the winds are blowing in Amazon’s favor, but the company has a lot of room to expand. Its share of cloud software is a fraction of its IaaS share, just 1.5% of the total market, according to IDC estimates, the 12th-largest vendor in the market. This offers a tempting proposition for Amazon to expand its footprint.

Figure 5

Worldwide cloud software revenue share by primary market, 2015

Source: IDC

Applications lead the move to cloud There is a new generation of application architecture native to the cloud, and development itself is moving from on premise to the cloud. Nic Poulos of venture firm Bowery Capital noted that the market for custom-built software experienced a Cagr of 33% from US$43bn in 2011 to US$136bn in 2015, vastly outpacing the 9% growth in applications, and 10% overall growth in enterprise software spending. He sees four key reasons for the robust uptake of custom development: the use of custom apps to enable hybrid on-premise public cloud deployments; the need for custom apps to serve niche and complex use cases; custom mobile application development; and traditional custom applications. The upsurge in custom development was driven by pressure to modernize and specialize software offerings, but Poulos sees SaaS applications as major beneficiaries of budget shifts over the next five years.

Looking across the software landscape, the rate of cloud penetration varies across different software categories. SaaS applications are already at around 25% cloud penetration, and this is likely to increase to 36% over the next five years.

AWS’s share of cloud software is a fraction of

its IaaS share, just 1.5% of the total market

The rate of cloud penetration varies

across different software categories

Development itself is moving from on premise

to the cloud




Figure 6

Worldwide all packaged software revenue by segment (US$m) 2015 2016 2017 2018 2019 2020 2015-2020

Cagr (%)

Applications

Total software applications (cloud + traditional) 199,216 211,398 226,725 243,014 260,598 279,501 7.0

SaaS applications penetration (%) 23.5 26.3 28.9 31.2 33.5 35.6 8.7

System infrastructure software (SIS)

Total software SIS (cloud + traditional) 96,609 103,094 107,924 113,396 119,036 124,743 5.2

SaaS SIS penetration (%) 9.1 10.2 11.4 12.6 13.8 15.0 10.4

SaaS subtotal

Total software applications and SIS (cloud + traditional) 295,825 314,492 334,648 356,410 379,635 404,244 6.4

SaaS penetration (%) 18.8 21.0 23.2 25.3 27.3 29.3 9.2

Application development and deployment (AD&D)

Total software AD&D (cloud + traditional) 90,372 95,441 102,428 110,036 118,293 127,254 7.1

PaaS penetration (%) 9.9 12.9 16.1 19.5 22.7 26.2 21.6

Total

Total software (cloud + traditional) 386,197 409,933 437,076 466,446 497,927 531,499 6.6

Cloud software penetration (%) 16.7 19.1 21.6 23.9 26.2 28.5 11.3 Source: IDC

SaaS applications are not AWS’s core competency, but independent software vendors (ISVs) are its natural customer base for Infrastructure as a Service. A growing proportion of on-premise and cloud-native ISVs are “all-in” with AWS for their cloud offerings. Salesforce, the largest, in May 2016 announced a partnership with AWS in order to accelerate its expansion into new countries. Formerly on-premise companies like Intuit, Qlik, Splunk, MicroStrategy, SoftwareAG, Autodesk, Altair and Tibco have committed to AWS for their cloud offerings. AWS also supports a broad range of cloud-native SaaS applications such as Looker, Sumo Logic, Works Applications, Okta, Netflix and others, which are all-in on AWS. By some estimates between 30-50% of all SaaS application vendors are running on AWS.

System infrastructure software includes systems management, network software, security, storage and system software. This market is about half the size of the applications market, currently with less than 10% running in the cloud, likely to increase to 15% over the next five years. This category is not large in terms of revenue opportunity, but is part of the critical portfolio of services needed to support SaaS applications. Application development and deployment encompasses structured data management software (databases), application development software, quality and life-cycle tools, application platforms, integration and orchestration middleware, and data access, analysis and delivery software. These on-premise software categories are being replaced by Platform-as-a-Service offerings. The PaaS market represents the central opportunity for AWS.

Cloud data services are AWS’s biggest opportunity, with IDC forecasting a 39% Cagr, from US$3bn in 2015 to US$15.9bn in 2020. This market includes end-user query, reporting and analysis (QuickSight), advanced and predictive analytics software (machine learning), relational database management systems (Aurora, RDS), nonrelational database management systems (SimpleDB, DynamoDB), distributed cache services managers and database development and management tools.

Independent software vendors are AWS’s

natural customer base for IaaS

Cloud data services are AWS’s biggest

opportunity

System infrastructure software is a critical part

of the portfolio of services needed to support SaaS




Figure 7

Worldwide competitive public cloud Platform-as-a-Service revenue by submarket

(US$m) 2015 2016 2017 2018 2019 2020 2015 Share

(%)

2015-2020 Cagr (%)

2020 Share

(%) Data 3,037 4,644 6,698 9,188 12,128 15,915 34.1 39.3 47.7

Application platform 2,301 3,139 4,166 5,246 6,322 7,391 25.8 26.3 22.2

Integration 2,093 2,617 3,270 4,004 4,793 5,636 23.5 21.9 16.9

Application development and life cycle

993 1,273 1,641 2,089 2,572 3,112 11.1 25.7 9.3

Other 479 593 726 885 1,071 1,291 5.4 21.9 3.9

Total 8,903 12,266 16,502 21,412 26,886 33,345 100 30.2 100 Source: IDC

Winning the developers creates a virtuous cycle The new generation of application architecture incorporates ever smaller components that become stitched together into composite applications. Containers and microservices are entering the mainstream. According to a survey by IT services firm NGINX, 2/3 of organizations are investigating containers with 20% using them in production. Of those running containers in production, over 1/3 are over 80% of their workloads. Using containers reduces dependencies between applications, and they can run in a smaller time frame.

“Data gravity” helps Amazon gain the upper hand Amazon’s software strategy is predicated on the strategic advantages that arise from storing its customers’ data. AWS attracts users with the cheap S3 storage and EC2 computing - and once the data is in AWS, it’s frictionless to sample an expanding panoply of services before buying. The flip side is that it’s easy to overspend, hence proliferation of AWS cost management startups on the expo floor.

Once the data sits in the S3 service, it tends to stay there. One of the ways that Amazon measured the adoption of its storage services is by the number of objects stored in S3. In 2006, there were 2.9 billion objects, which increased to 262 billion by 2010. By April 2013 (the most recent disclosure) Amazon disclosed that there were over 2 trillion objects, doubling from 1 trillion a year earlier. The company discontinued this disclosure in 2014, noting 102% YoY growth in usage of S3 in 1Q14. Given sustained expansion of AWS, it’s possible that the total number of objects exceeds 10 trillion or more in 2016.

“Data gravity” is a concept coined by developer Dave McCrory to describe the phenomenon in which the number, quantity and speed at which services, applications and users are attracted to data increase as the mass of that data increases. There were several important conclusions from an IDC study on data gravity in 2014:

The value of data will increasingly come from secondary uses (like analytics) not just primary use (transactions).

As databases grow in size to petabyte scale and larger, it will be easier to move applications, services and workloads to the data than move the data.

For new applications, data "born in the cloud" will increasingly remain in the cloud.

The new generation of application architecture

incorporates ever smaller components

Amazon’s strategy is predicated on advantages that arise from storing its

customers’ data

Once the data sits in the S3 service, it tends

to stay there

“Data gravity” attracts apps and services to data

wherever it resides




It will be easier to analyze disparate data sets in the cloud when compute capabilities are adjacent. More “big data” stored in datacenters will be disposed of or moved to the cloud.

More vendors will move to AWS to provide cloud-based real-time data processing as a service.

Beyond cost savings, AWS’s growing ecosystem attracts users AWS is also gaining advantages in terms of its ecosystem. According to a 2016 IDC survey of 287 users, the two top factors driving EC2 adoption was reducing development cost and gaining access to the AWS application ecosystem (the services within AWS). The ecosystem appeal saw a dramatic increase from 2015, suggesting that the appeal of the AWS portfolio is in itself attracting customers.

CLSA analyst James Lee forecasts AWS to generate about US$12.1bn in revenue in 2016. IDC estimated that AWS generated US$996m in cloud software revenue in 2015, a 132% increase from 2014. Based on this estimate, cloud software represents 12.6% of AWS 2015 revenue, an increase from 9.2% in 2014.

Figure 8

AWS historical financials and forecasts

(US$m) 2013 2014 2015 16CL 17CL 18CL Sales 3,108 4,644 7,880 12,067 15,687 19,608

Segment operating expenses 2,435 3,984 6,016 8,503 10,824 13,334

Segment operating income 673 660 1,864 3,563 4,863 6,275

Segment OPM (%) 21.7 14.2 23.7 29.5 31.0 32.0 Source: Company, CLSA

Where will AWS focus next? AWS offerings address a broad range of categories that span IaaS, PaaS and SaaS. When Amazon publicly discusses its product strategy, the company notes that many of its innovations are driven by customer requests, whether this be database, analytics or innovations in the core storage and compute service portfolio.

A brief overview of current AWS offerings Compute, storage and networking services address basic infrastructure functions and would be characterized as IaaS. AWS cloud operates 35 availability zones with 13 geographic regions around the world with nine more availability zones and four more regions coming online throughout next year. We estimate that these services currently account for over 80% of AWS revenue, and will continue to be the “bread and butter” services of the business.

Most of the applications deployed on AWS are running on virtual machines, but with the introduction of containers and Lambda (serverless computing) users have greater flexibility in how to architect their applications to incorporate more modern approaches. These services compete directly with Microsoft Azure and Google Cloud Platform along with other public cloud service providers. The impact from turning physical servers and storage into services has been felt most directly by HP, IBM, Sun/Oracle, Dell and VMware as well as EMC and NetApp.

AWS is also gaining advantages in terms

of its ecosystem

AWS should generate about US$12.1bn in

revenue in 2016

Many of its innovations are driven by

customer requests

Compute, storage and networking service

address basic infrastructure functions




Compute services include virtual servers (EC2), containers (EC2 Container Registry), 1-click web app deployment (Elastic Beanstalk), event-driven compute functions (Lambda), auto scaling and load balancing.

Storage & content delivery services include object storage (S3), content delivery networks (CloudFront), block storage (Amazon EBS), file system storage (Amazon Elastic File Storage), archive storage (Glacier), data transport (Import/Export Snowball) and integrated storage (AWS Storage Gateway).

Networking services include virtual private cloud, direct connections, load balancing and domain name system (Route 53).

Application development and enablement is strategic to AWS’s ability to extend its ecosystem of users. Developer tools enable users to manage and automate code deployment. Mobile services include Mobile Hub and API Gateway to build, test and monitor applications and APIs. These services enable developers to test apps on different devices, stream applications and analyze performance. These capabilities would be categorized as PaaS offerings.

Developer tools include source code management (CodeCommit), code deployment (CodeDeploy) and continuous delivery (CodePipeline).

Mobile services include mobile development (Mobile Hub), API management (API Gateway), identity (Cognito), app testing (Device Farm), mobile analytics, development and notifications (Amazon SNS Push Notification Service).

Application services include API management (API Gateway), app streaming (AppStream), search (CloudSearch, transcoding (Elastic Transcoder), email (SES Email Sending and Receiving Service), notifications (SNS Push Notification Service), queueing (SQS Queue Service) and workflow (SWF Workflow Service for Coordinating Application Components).

Internet of Things offerings include the IoT platform (AWS IoT), device software development kit (SDK), registry, device shadows and rules engine.

AWS has invested in a significant portfolio of management and security offerings. These are mostly designed to be used with applications and workloads running on AWS. These offerings fall under the systems infrastructure services segment of the cloud software category.

Management tools include monitoring & logs (CloudWatch), resource templates (CloudFormation), usage & resource auditing (CloudTrail, Config), dev/ops resource management (OpsWorks), service catalog and performance optimization (Trusted Advisor).

Security & identity services include access control (Identity and Access Management), SSL/TLS certificates (Certificate Manager), key storage & management (CloudHSM, Key Management Service), identity management (Directory Service), security assessment (Inspector) and web application firewall.

Big data, database and analytics offerings include a broad range of third-party, open source and proprietary offerings. Database and analytics services straddle Platform- and Application-as-a-Service categories, but offer the most significant potential, in our view, for sustained revenue growth and share gains in the market.

Application development and enablement is

strategic to AWS’s ability to extend its ecosystem

AWS has invested in a significant portfolio of

management and security offerings

Database and analytics services straddle PaaS

and SaaS capabilities




Database services include relational (managed relational database service for Aurora, MySQL, PostgreSQL, Oracle, SQL Server, MariaDB), database migration services, NoSQL (DynamoDB), caching (ElastiCache) and data warehouse (RedShift).

Analytics include business intelligence (QuickSight), data warehouse (RedShift), machine learning, streaming data (Kinesis), Elasticsearch, Hadoop (Amazon Elastic MapReduce) and data pipelines.

Amazon’s applications are the only offerings that directly target business users. WorkSpaces is a hosted virtual desktop offering, while WorkMail and WorkDocs target hosted email and document sharing services. These are large markets with established competition.

Enterprise applications include desktop virtualization (WorkSpaces), email & calendaring (WorkMail), document sharing and feedback (WorkDocs).

Focusing on priorities Given the advantages from its leading position in Infrastructure as a Service, secular migration from on premise to public cloud, strategic importance of analytics for driving business value and increasing competitive focus on differentiating features rather than pricing in the public cloud arena, we expect AWS’s emphasis to target software features in a range of categories. A summary of areas we expect to be AWS priorities:

Database and analytics. As the phenomenon of “data gravity” attracts more applications, services and users to the AWS platform, the more data managed and stored in AWS grows. This pulls demand for data-related services and applications into the AWS orbit. Amazon’s RedShift, Aurora and RDS offerings have already captured significant share in the market. The business intelligence offering and QuickSight are currently in preview. We estimate that RedShift, Aurora and other database services may account for over 12-15% of AWS revenue in 2015, which could increase to 20% by 2020.

Internet of Things. Given the tremendous potential role that AWS can play helping organizations connect devices to the internet, we would expect AWS to redouble focus on its IoT platform. While the market is nascent, the opportunity to drive pull-through for compute, storage and application services is enormous. We estimate that AWS IoT could directly and indirectly generate between US$1.2-1.8bn in revenue by 2020.

Application development. Attracting more application development creates a virtuous cycle: as more applications are developed on AWS, this will drive demand for infrastructure, application, mobile and analytics services. Application development processes are also moving into the cloud, and we would expect AWS to expand its relatively modest development service portfolio. We don’t see application development as a significant revenue contributor, but services like the recently acquired Cloud9, CodeCommit, AWS IoT and Mobile Hub should drive demand across the portfolio. Amazon Lumberyard is a free cross-platform 3D game engine for developers to create games running on the AWS cloud.

Next-gen services. AWS has positioned its newer services to support the needs of emerging types of applications, making use of approaches such as containers and microservices. Notable services include Lamba (serverless) computing for event-driven compute in ever smaller

RedShift, Aurora and RDS offerings have already

captured significant share

We estimate that AWS IoT could directly and

indirectly generate US$1.2-1.8bn by 2020

Application development creates a virtuous cycle of demand for more services

Newer services support the needs of emerging

types of applications

Applications offerings are the only ones that directly

target business users




increments and EC2 Container Services for managing clusters of containers; and real-time services like Kinesis Firehose for streaming data capture and AppStream for streaming Windows apps to mobile devices.

Application delivery and networking. AWS is involved in a range of network services that are a natural extension of IaaS capabilities. AWS’s ADC functions accelerate the delivery of application particularly for enterprises, and span several offerings most notably elastic load balancing and the recently launched application load balancing service. IDC estimates that the services that comprise AWS’s ADC capabilities (which also include CDN, the Route 53 networking services, web app firewall and API gateway) generate around US$500m annually. We expect application performance capabilities to grow in line with overall revenue and play a key role differentiating products.

Security and management. Similar to application delivery and networking, we don’t see security and performance management as significant direct contributors to revenue, but as key areas for competitive differentiation against other public cloud providers. Security consistently ranks as a top concern among businesses moving operations to the cloud. AWS has a good track record without major security breaches (and also validation from security-conscious customers like the CIO). Management capabilities are also critical to sustain the self-service nature of the experience.

Figure 9

Competitive matrix Category Services On-premise

vendors Cloud vendors

Competitive impact

Comments

Compute EC2, EC2 Container Registry, Elastic Beanstalk, Lambda, Auto Scaling and Load Balancing

HP, IBM, ORCL, Dell

MSFT, GOOG, IBM, RAX, ATT, VZ, others

High AWS dominates compute as a service with scale, pricing and breadth

Storage & content delivery

S3, CloudFront, Amazon EBS, Elastic File Storage, Glacier, Import/Export Snowball and AWS Storage Gateway

NTAP, IBM, ORCL, HPE, Dell/ EMC, Veritas

GOOG, MSFT, IBM, AKAM

High AWS dominates storage as a service with scale, pricing and breadth

Networking Virtual Private Cloud, Direct Connections, Elastic Load Balancing, Application Load Balancing and Domain Name System (Route 53)

CSCO, JNPR, HPE, FFIV, CTXS, RADW

VRSN, NSR, CSCO, CTXS

Low to medium

AWS's direct impact has been minimal, but product roadmap increasingly targets the application delivery services market

Management tools

CloudWatch, CloudFormation, CloudTrail, Config, OpsWorks, Trusted Advisor

CA, IBM, VMW, MSFT, RHT, HPE, CSCO, SPLK

VMW, MSFT, ORCL, IBM, NOW, NEWR, RHT, SPLK

Low AWS management tools targeted to AWS workloads

Security & identity

Identity and Access Management, Certificate Manager, CloudHSM, Key Management Service, Directory Service, Inspector and Web Application Firewall

EMC/RSA, CA, IBM, CYBR

AKAM, IMPV, VRSN, QLYS, GDDY

Low AWS security and identity tools address AWS user needs

Database RDS (Aurora, MySQL, PostgreSQL, Oracle, SQL Server, MariaDB), Database Migration Service, DynamoDB, ElastiCache, RedShift

ORCL, IBM, MSFT, SAP, TDC, HDP

MSFT, GOOG, ORCL

Medium to high

AWS's Database-as-a-Service offerings are aggressively capturing incumbent share

Analytics QuickSight, RedShift, Machine Learning, Kinesis, Elasticsearch, Elastic MapReduce and Data Pipelines

DATA, TDC, IBM, SAP, ORCL, MSFT

MSFT, GOOG, CRM

Medium Data Warehouse as a Service is having significant market impact, but business intelligence has not entered the market

Enterprise applications

WorkSpaces, WorkMail, WorkDocs CTXS, MSFT, OTEX

MSFT, GOOG, CTXS, BOX

Low AWS footprint is limited, cloud competition well established

Internet of Things

AWS IoT, Device SDK, Registry, Device Shadows and Rules Engine

MSFT MSFT, GE, GOOG, CRM

Low to medium

Market is nascent but AWS offerings are highly competitive

Source: CLSA, Amazon

AWS is involved in a range of network services

We see security and performance as key

areas for competitive differentiation


Section 2: Data(base) is the key to expansion Software


Data(base) is the key to expansion Database services are AWS’s most popular software option. It’s our view that Amazon is now the most potent competitor in the US$34bn-plus database market. With the lure of substantial savings from moving to its EC2 compute and S3 storage services, Amazon is aggressively courting customers to its database services. Enhanced database migration services strengthen the price/performance advantages of Aurora and RedShift, eroding technological advantages and switching costs previously enjoyed by the likes of Oracle, IBM and Teradata. Microsoft appears best poised to fend off AWS’s rising challenge.

From our conversations with AWS users and consulting partners, the overwhelming consensus is that Amazon’s database services are the most popular options beyond EC2 and S3 storage. A common trajectory is to go from Oracle (including Exadata) on premise to Oracle RDS, then to MySQL RDS, Aurora or RedShift. Migration to RedShift continues to be disproportionately sourced from Teradata.

Aurora aims a bazooka at database incumbents. Amazon’s Aurora proprietary relational database was launched in 2014 and is now AWS’s fastest-growing service to date. Designed so that users of the open-source MySQL database can move applications without any code rewrites, Aurora promises significant performance gains at a fraction of the cost of on-premise or cloud-based alternatives. Aurora appears to be AWS’s most aggressive strategic focus.

Database migration service ratchets up competitive pressure. Amazon’s database migration services provide an easy, low-cost way to migrate code, schemas and data to databases running on AWS’s RDS including Aurora. The service makes it easy to migrate to Aurora or open-source databases from proprietary systems like Oracle or Microsoft SQL Server. AWS just announced services to migrate from SAP’s Sybase ASE database, Oracle’s data warehouse and Teradata to RedShift. With over 7,000 migrations in 2016 and accelerating pace, the migration service makes AWS’s lower-cost model even more attractive and significantly increases the threat to database incumbents.

NoSQL is rapidly encroaching on the US$34bn+ database market. Amazon pioneered the NoSQL market in 2007, but the non-relational market is 13% of the total database market. With Aurora, database revenue growth accelerated in 2015 to US$580m and AWS commands 30% of the public cloud database market. We believe Amazon is positioned to capture Oracle-like share of public cloud database services over the next few years, increasing the deflationary impact on incumbents including Oracle, Teradata and IBM.

Amazon mounts a database offensive Amazon built on its success delivering compute, storage and other cloud services, and has steadily expanded into database services. The company first launched its SimpleDB database devices in 2007, and in October 2015 stunned the industry when it disclosed that its database services were tracking at a US$1bn annual run rate. CTO Werner Vogels noted that the Aurora database remains the “fastest-growing service in the history of AWS” designed to provide “commercial capabilities at an open source price”. Price is key: a deep-dive demo’ed Aurora at 56% cheaper than comparable open-source MySQL on AWS. With the expansion of database migration services, it’s our view that Amazon’s competitive threat is increasing.

It’s our view that Amazon’s competitive

threat is increasing

Amazon is now the most potent

database competitor

AWS’s database services are the most popular

options beyond EC2 and S3 storage

Database migration services provide an easy, low-cost way to migrate code, schemas and data




How Amazon classifies its database services revenue is not particularly transparent, as there are storage costs associated with its services. Additionally, Amazon’s total RDS revenue includes Oracle and Microsoft SQL Server licenses where the customer either uses his own license or purchases the license by subscription resold by Amazon. We interpret the US$1bn run rate figure as a gross revenue number, with Amazon’s revenue from the sale of database software and RDS management services representing net software revenue. The IDC forecasts we use in our analysis reflect net software revenue.

Sizing Amazon’s impact on the database market IDC estimates that Amazon generated US$380m in RDS revenue in 2015, with US$134m in non-relational database revenue in 2014 (the most recent year estimates are available). We’d conservatively estimate AWS’s 2015 revenue from database services at about US$580m, representing 167% YoY growth in a market that grew 1%. We estimate that Amazon can grow over 100% to 2016 revenue of US$1.2bn.

From modest beginnings to the first major new player in decades Amazon launched its first database service SimpleDB at the end of 2007, which is regarded as one of the starting points of the NoSQL (nonrelational) database movement. The launch of DynamoDB in 2012 helped propel Amazon to the No.5 share in the non-relational market.

Figure 10

Database software revenue (relational and non-relational, on premise and cloud) vs AWS

(US$m) 2014 2015 2016 2017 2018 2019

On-premise database software 32,963 32,484 33,198 34,286 35,303 36,349

- YoY growth (%) 2 (1) 2 3 3 3

- As a % of total market 97 94 91 88 84 81

Public cloud database software 1,118 1,932 3,154 4,693 6,539 8,687

- YoY growth (%) 90 73 63 49 39 33

- As a % of total market 3 6 9 12 16 19

Total database software market 34,081 34,416 36,352 38,979 41,843 45,036

- YoY growth (%) 4.0 1.0 5.6 7.2 7.3 7.6

AWS total database software 217 580 1,198 2,018 2,877 3,909

- YoY growth (%) 115 167 107 68 43 36

- AWS share of total market (%) 0.6 1.7 3.3 5.2 6.8 8.6

- AWS share of public cloud market (%) 19 30 38 43 44 45 Source: IDC, CLSA

With the launch of Aurora at the end of 2014 and the strong momentum from RedShift, Amazon’s database software revenue growth accelerated in 2015. According to estimates from IDC, Amazon’s share of the public cloud database services market increased from 19% to 30% as share of the total database market increased from 0.6% to 1.7%.

A broad portfolio of relational, non-relational, proprietary, 3rd party Amazon’s offerings span proprietary and 3rd party, relational and non-relational database services. RedShift, the Data-Warehouse-as-a-Service offering based on technology from ParAccel, was announced in November 2012.

We’d estimate AWS database revenue jumped

167% YoY in a market that grew 1%

Amazon’s database software revenue growth

accelerated in 2015




Figure 11

Amazon Database-as-a-Service offerings

Customer need Amazon offering Product type

A managed relational database in the cloud Amazon RDS (Amazon Aurora, MySQL, PostgreSQL, Microsoft SQL Server, Oracle, MariaDB)

Relational database

A fully managed MySQL compatible relational database with 5X performance and enterprise level features.

Amazon Aurora Relational database

A managed NoSQL database Amazon DynamoDB NoSQL database (Non-relational)

A fully managed, petabyte-scale data warehouse Amazon Redshift Data warehouse (relational)

To deploy, operate and scale in-memory cache based on Memcached or Redis in the cloud

Amazon ElastiCache In-memory cache

Help migrating databases to AWS AWS Database Migration Service Database migration Source: Amazon

Database migration service escalates competitive pressures Amazon has been increasingly focused on developing migration services. A professional services team will start with a portfolio analysis of workloads and rank them based on difficulty to move. The company has also launched migration tools including application discovery, database schema migration service and the Snowball data transfer appliance.

Figure 12

Database migration service

Source database Target database on Amazon RDS

Oracle Amazon Aurora, MySQL, PostgreSQL, MariaDB

Microsoft SQL Server Amazon Aurora, MySQL, PostgreSQL, MariaDB

Sybase ASE Amazon Aurora, MySQL, PostgreSQL, MariaDB

MySQL and MariaDB PostgreSQL

PostgreSQL Amazon Aurora, MySQL, MariaDB

Amazon Aurora PostgreSQL

Oracle Data Warehouse Amazon RedShift

Teradata Amazon RedShift Source: Amazon

In July 2016, AWS announced four new features. These include migration from Sybase/ASE and data warehousing migration from Teradata and Oracle to RedShift and application code migration, recommending appropriate translations of code in the database to the destination system and also expanding the service to support continuous replication.

Migration service lowers technology and cost barriers to adoption AWS Database Migration Service (DMS) is designed to help customers move data from on premise to cloud, from one database to another (from commercial to open source). In the past database migration required replication software costing tens of thousands of dollars and/or a lot of technical expertise. Now, the service is fast and allows users to keep applications running during the migration. There can be multiple sources and targets. The service is a logical replication tool - it uses select queries to move data selectively. The service maintains minimal application downtime using Change Data Capture to track and update changes that occur during the migration.

Amazon has been increasingly

focused on developing migration services

DMS is designed to help customers move data

from on premise to cloud




Amazon supports homogeneous database migrations in the case where customers want to move from on premise to the AWS cloud. This could be Oracle to Amazon RDS for Oracle, MySQL to Amazon Aurora, MySQL to Amazon RDS for MySQL, or Microsoft SQL Server to Amazon RDS for SQL Server. This is a simple process that transfers all of the data from a source database on premise, running on an EC2 instance or running as an Amazon RDS database.

Figure 13

AWS Database Migration Service - Heterogeneous database migrations

Source: Amazon

In the case of heterogeneous migrations (such as Oracle to Amazon Aurora, Oracle to PostgreSQL, or Microsoft SQL Server to MySQL), the schema structures data types, and database code of source and target databases can be different. This requires a two-step process of schema and code transformation before data migration starts. The AWS Schema Conversion Tool will convert the source schema and code to match the target database, and then use the AWS Database Migration Service to migrate data from source to target database.

AWS Database Migration Service can now perform continuous data replication to support use cases that can include disaster recovery instance synchronization, geographic database distribution and dev/test environment synchronization. Continuous data replication works for both homogeneous and heterogeneous data replications for all supported database engines.

Snowball is AWS’s data import/export appliance. The service for 50TB is US$200, for 80TB is US$250; customers have 10 days to load the data onto AWS S3. There is a fee of US$0.03/GB to load data out of AWS S3. Customers load the data into the appliance, which is encrypted and the shipping label is automatically updated via a Kindle display on the unit.

Amazon supports homogeneous database

migrations where customers want to move

from on premise to the AWS cloud

AWS Database Migration Service can now

perform continuous data replication

Customers load the data into the appliance, which is encrypted

Figure 14

Amazon Snowball comes in 50 and 80 terabyte capacity

Source: Amazon




Figure 15

Worldwide relational database management systems revenue snapshot

Source: IDC

Migration service accelerating pace of AWS database adoption The conversion tool is accelerating migration from on premise to Amazon. The company opened up its migration services in 2016. Over 1,000 database instances have migrated to the Amazon cloud between January and March 2016. In late June 2016 AWS’s CEO Andy Jassy announced that the service had successfully migrated over 7,000 database instances from on premise to Aurora.

Relational database market shifting to cloud model Relational database management was a US$19.9bn market in 2015, according to IDC estimates. The market for Database as a Service is a subset of the overall database segment estimated at US$1.3bn in 2015 and forecast to grow to US$7.4bn by 2020. IDC forecasts that the on-premise relational database market will see a 2.8% Cagr through 2020, while public cloud relational database as a service will grow at 41.1%. Put another way, public cloud services should increase from 4.4% of the market in 2015 to 18% in 2020. Amazon’s RDBMS offerings, RDS, Aurora and RedShift, compete purely in the cloud.

The overall relational database market is dominated by Oracle, with 44% share in 2015, followed by Microsoft, IBM, SAP and Teradata. Amazon is the insurgent with just over 1% share with an estimated US$380m in revenue in 2015.

Public cloud RDBMS services should increase from 4.4% of the market

in 2015 to 18% in 2020

Over 7,000 database instances from on

premise to Aurora have been successfully migrated

The RDBMS market is forecast to grow 6%




Figure 16

Worldwide relational database management systems software 2015 share

Source: IDC

Amazon’s RDBMS revenue growth accelerated in 2015 According to IDC’s estimates, Amazon’s relational database revenue grew by US$297m from 2014-15, up 360%. This compares to YoY growth of US$58m or 226% the prior year. This compares to US$1.2bn of revenue declines from Oracle, IBM and Teradata combined.

Figure 17

Worldwide relational database management systems revenue by vendor (US$m) 2013 2014 2015 2015

share (%) 2014-2015

growth (%) Oracle 13,389 13,756 13,115 43.9 (4.7) Microsoft 5,320 5,781 6,358 21.3 10 IBM 4,868 4,814 4,301 14.4 (10.6) SAP 1,685 1,917 2,119 7.1 10.6 Teradata 1,025 1,063 987 3.3 (7.2) Amazon.com 25 83 380 1.3 360.2 Progress Software 181 196 199 0.7 1.5 Fujitsu 232 218 191 0.6 (12.4) Hewlett Packard Enterprise 131 141 149 0.5 6 Actian 66 82 110 0.4 34.8 Other 1,658 1,757 1,980 6.6 12.7 Total 28,581 29,806 29,889 100 0.3 Source: IDC

Targeting the RDBMS market with an aggressive approach We’ve covered RedShift previously, which to date had been AWS’s fastest-growing service until Aurora. The main relational offerings are RDS and Aurora. Amazon RDS enables users to set up, operate and scale a relational database in the cloud. RDS offers a choice of six database engines: Amazon Aurora, Oracle, Microsoft SQL Server, PostgreSQL, MySQL and MariaDB (a fork of MySQL headed by Martin Mickos, former CEO of MySQL). Amazon first launched RDS in October 2009, supporting MySQL databases. This was followed by support for Oracle in June 2011, Microsoft SQL Server in May 2012, PostgreSQL in November 2013 and MariaDB in October 2015.

Amazon’s relational database revenue grew

360% in 2015

RedShift had been AWS fastest-growing service

until Aurora

The relational database market is dominated

by Oracle




RDS enables users to deploy a database, scale the compute and storage resources, launch read replicas to offload read traffic from the primary database instance and manage performance through a console. Features include easy scaling of compute and storage, automated backups and database snapshots, multi availability zone deployments, encryption of data at rest and in motion, network isolation, management metrics and alerting. Instances are offered with a range from 1-40 CPUs, 8 to 244 GBs of memory and a range of network performance. The service is offered on demand or as reserved instances with lower pricing for longer-term commitments.

RDS - a premium AWS service, still way cheaper than on premise Because of a broad array of choices, RDS pricing varies significantly depending on the needs of the customer. Not surprisingly, RDS offerings are priced at a premium to the underlying EC2 compute cloud service, meaning that users pay extra for the management, scaling and monitoring capabilities. License costs are extra, except for MySQL.

There’s a minimal premium to run PostgreSQL on RDS (just US$0.01 an hour for a basic instance). According to an analysis by Cloudyn, the incremental cost to run an Oracle license on RDS (versus MySQL, which has no licensing fee) is roughly equivalent to the annual license cost for Oracle to run on premise. Running Microsoft SQL Server on RDS is cheaper per hour than running Oracle on RDS, and also cheaper than SQL Server on-premise licensing at list price (before discounts).

Aurora expands like a nebula In November 2014, Amazon announced its proprietary relational database offering Aurora, and the product became generally available on 28 July 2015. AWS CEO Andy Jassy highlighted that based on customer requests, the company developed Aurora to be fully compatible with the open source MySQL database. The company that developed MySQL was acquired by Sun, which was in turn acquired by Oracle (Oracle continues to sell subscriptions for MySQL). Aurora has seen rapid adoption within the user base. According to CTO Werner Vogels, Aurora is the fastest-growing service in the history of AWS. In late June 2016, CEO Andy Jassy announced that the service had successfully migrated over 7,000 database instances from on premise to Aurora.

Amazon claims that Aurora delivers 5X more performance than commercial MySQL implementations and is as fault-tolerant as high-end commercial grade offerings at 1/10 the cost. At the July 2016 Summit in Santa Clara, Amazon showed how a Japanese gaming customer made the transition from running MySQL on RDS to Aurora with no code changes, improving web page response time from 15 to 5.5 milliseconds, a 3X improvement at no extra cost. Migration is a multistage process - business “lift and shift” then optimize to take advantage of AWS-native services, particularly database. Users note some limitations, but give kudos for speed, ease of use and migration.

There is no additional licensing beyond Amazon’s pricing for RDS. For Aurora pricing starts at US$0.29 per hour for an instance with 2 virtual CPUs and 15.25GB RAM. In addition, users pay US$0.10 per GB/month for storage and US$0.20 per million requests.

Pricing is significantly more aggressive than competition At the AWS Summit, the company presented a scenario analysis that showed Aurora offering anywhere from 20-56% monthly savings from running MySQL RDS based on various options. According to a pricing analysis by Nimbo

RDS offerings are priced at a premium to the

underlying EC2 service

Amazon claims that Aurora delivers

5X more performance than commercial

MySQL implementations

Based on requests, AWS developed Aurora to be

fully compatible with the MySQL database




(Equinix’s professional services), Aurora’s pricing is much cheaper than running other databases in the cloud - not necessarily by a factor of 10X in every case but significantly less. Nimbo compared running Oracle Enterprise, Microsoft SQL Server, MySQL Enterprise on a large EC2 instance (4 virtual CPUs and 30GB/storage) and on RDS with multi availability zones - versus AWS Aurora.

Based on this analysis, Oracle would cost US$52,756/year to run on EC2 with a bring-your-own-license model, versus US$14,717/year on AWS RDS. Microsoft SQL Server 2012 would cost US$19,004/year to run on EC2 with a bring-your-own-license model, versus US$13,140/year on AWS RDS. MySQL Enterprise would cost US$10,256/year to run on EC2 with a bring-your-own-license model, versus US$8,322/year on AWS RDS. Aurora would cost US$6,744 per year, representing a potential savings of US$3,000-46,000 (or 30-88%) per year versus alternatives. Performance comparisons show AWS Aurora is also highly competitive. There are also significant discounts from on-demand pricing of 44%/1 year and 63%/3 years for reserved instances.

Nonrelational database market smaller, but growing faster Relational database management was a US$4.5bn market in 2015, according to IDC estimates. The market for Database as a Service is a subset of an estimated US$614m in 2015, forecast to grow to US$3.0bn by 2019. IDC forecasts that the on-premise nonrelational database market will see a 2.7% Cagr through 2019, while public cloud relational database as a service is forecast to grow at 52%. Amazon launched its first database service SimpleDB at the end of 2007. Amazon announced DynamoDB, a fully managed proprietary NoSQL database service, in January 2012.

Figure 18

Worldwide nonrelational database management systems software revenue

Source: IDC

Nonrelational databases are growing faster than

traditional RDBMSs

Based on this analysis, Aurora could save

between 30-88% from alternative databases

Relational database management was a

US$4.5bn market in 2015




According to IDC estimates, Amazon’s was the fifth-largest non-relational database vendor in 2014, with revenue of US$134m. Estimates are not currently available, but we estimate Amazon’s DynamoDB revenue in excess of US$200m in 2015, surpassing Apple for fourth place in market share.

Figure 19

Worldwide nonrelational database management systems - 2014 share snapshot

Source: IDC

According to IDC estimates, Amazon’s DynamoDB revenue grew at 77% from 2013-14 (the most data available), representing the most rapid growth in the market. Amazon accounted for 3.1% of the market in 2014, and we expect this proportion to increase.

Figure 20

Worldwide nonrelational database management systems revenue by vendor

(US$m) 2012 2013 2014 2014 share (%)

2013-2014 growth (%)

Microsoft 1,521.1 1,726.0 1,702.3 39.6 (1.4)

IBM 885.7 881.1 866.7 20.2 (1.6)

InterSystems 300.9 310.8 342.1 8.0 10.1

CA Technologies 302.5 282.5 288.7 6.7 2.2

Apple 131.7 140.5 156.0 3.6 11.1

Amazon.com 43.5 75.8 133.9 3.1 76.8

Software AG 158.2 146.0 115.8 2.7 (20.6)

Google 44.2 49.6 68.1 1.6 37.4

MarkLogic 41.9 47.1 61.8 1.4 31.3

Rocket Software 43.5 43.5 44.9 1.0 3.3

Other 470.0 491.0 514.0 12.0 4.6

Total 3,942.9 4,194.3 4,294.1 100 2.4 Source: IDC

The US$4.3bn nonrelational market

represents 13% of the total database market

Amazon was the fifth-largest non-relational

database vendor in 2014




DynamoDB is a nonrelational database employing approaches that fall under the NoSQL category. NoSQL databases utilize a variety of data models, including document, graph, key-value and columnar. DynamoDB employs a key-value store model. Key-value store competitors include Redis, Cassandra, Azure Table Storage and similar services.

Key customers include AdRoll, BMW, Canary, Dropcam, Duolingo, MediaTek, Nordstrom, Redfin, Supercell, VidRoll, Zynga and many others. According to DB-Engines' ranking, Amazon DynamoDB is the most popular fully managed database service in the market.

DynamoDB pricing is based on two types of charges: charges for the amount of data stored and charges for provisioned read and write throughput capacity. There is a free option: DynamoDB customers get 25GB of free storage, up to 25 write capacity units and 25 read capacity units of ongoing throughput capacity (this is enough throughput to handle up to 200 million requests per month) along with 2.5 million read requests from DynamoDB Streams for free.

Amazon ElastiCache complements database, storage Amazon ElastiCache is a web service to deploy, operate, and scale an in-memory cache in the cloud. ElastiCache improves performance of web applications by enabling users to retrieve information from managed, in-memory caches, instead of on slower disk-based databases. ElastiCache detects and replaces failed nodes to provide a resilient system that reduces risks from overloaded cloud databases that can slow down websites and application loading. ElastiCache supports two open-source in-memory caching engines - Memcached and Redis - and works with Amazon CloudWatch to provide visibility into key performance metrics.

RedShift Data Warehouse as a Service emerges Amazon launched its Data-Warehouse-as-a-Service offering RedShift in 2013, which has quickly vaulted Amazon to No.6 market share behind Oracle, Microsoft, IBM, Teradata and SAP. It appears much of RedShift’s gains have come at the expense of Oracle, IBM and Teradata, whose data warehousing software revenue has declined. Hadoop too is a disruptive factor, though Microsoft and SAP have also grown revenue. As a first mover, RedShift has accelerated the shift of data warehousing towards the cloud, but we do not expect growth to remain unchallenged. Serious competition from Google and Microsoft is ramping, while IBM and Teradata are prepping their own offerings.

RedShift was the first significant managed Data-Warehouse-as-a-Service (DWaaS) alternative to a market that had been primarily on premise. Based on the ParAccel distributed database, high-performance advantages come from the massively parallel processing (MPP) architecture, columnar data storage, data compression and linear scalability. The use of industry standard (structured Query language) SQL enables users with existing data warehouse skill sets to quickly adapt to using the technology.

RedShift rolling over the data warehouse market At the recent AWS Summit in New York City, we were surprised by the superlative enthusiasm for RedShift among customers and partners. Users praised the scalability, performance and particularly the cost and opex payment model. There were a number of large users including GE, with the

DynamoDB is a nonrelational database employing approaches

that fall under the NoSQL category

ElastiCache is a web service to deploy,

operate and scale an in-memory cache

RedShift has accelerated the shift of data

warehousing towards the cloud

Superlative enthusiasm for RedShift among

customers and partners

The premier “Data Warehouse as a

Service” offering




largest implementation a multi-petabyte Chinese carrier. One partner shared an estimate of 150,000 customers, which would translate to roughly 15% penetration within the customer base. According to one large UK-based partner, roughly half of RedShift implementations are new projects and roughly half are migrations from on-premise systems. Teradata and Oracle (including Exadata) are by far the leading source systems for migrations, others include IBM/Netezza, HP/Vertica and InfoBright.

Early mover captures share in a market dominated by tech giants IDC estimates Amazon’s 2015 analytic data management software (RedShift) revenue grew 236% YoY to US$178m, the fastest growth of any vendor in a US$9.9bn market that grew 2.7% YoY. We believe much of Amazon’s US$125m YoY revenue growth in 2015 came at the expense of Oracle, IBM and Teradata, which saw 2015 analytic data management software revenue decline a combined US$356m YoY. After achieving 2% market share in just three years, we expect Amazon’s gains to continue at a moderated pace with the entry of formidable competitors with deep domain and technological depth.

Anticipating aggressive challenges from Microsoft, Google RedShift has staked early mover dominance in Data Warehouse as a Service, and while it’s cheaper than on premise, new entrants will raise the stakes for pricing and performance. Microsoft’s SQL Azure Data Warehouse as a Service (currently in beta) and Google’s Big Query are ramping their presence with aggressive pricing and differentiated features. A number of startups such as Snowflake Computing and CoolaScale are in the market, while Teradata and IBM’s Data-Warehouse-as-a-Service offerings are waiting in the wings.

Figure 21

Worldwide analytic data management software revenue by vendor 2015

Source: IDC

AWS RedShift targets the data warehouse market RedShift is based off technology from ParAccel (now owned by privately held Actian). Amazon Redshift is a relational database management system, so it is compatible with other RDBMS applications. Amazon Redshift is based on PostgreSQL 8.0.2. Although it can support online transaction processing (OLTP) functions such as inserting and deleting data, Redshift is optimized for high-performance analysis and reporting of very large datasets.

Oracle36%

Microsoft19%IBM

15%

Teradata10%

SAP9%

Amazon.com2%

Rest of Market9%

We believe much of AWS revenue growth in 2015 came at the expense of

Oracle, IBM and Teradata

A number of startups such as Snowflake Computing

and CoolaScale are in the market

Oracle dominates the analytic data

management software market




The response to the product has been overwhelmingly positive with Amazon disclosing that it’s the fastest-growing offering outside of the core AWS service. In April 2015, CTO Werner Vogels told attendees of the AWS Summit in London that Redshift is “the fastest-growing service in AWS, ever.”

In our view, RedShift has been the AWS offering with the most impact on the competitive landscape other than the core EC2 compute cloud and S3 storage services. RedShift is growing revenue faster than any other data warehouse offering in the market. We believe RedShift’s success is a significant contributing factor (along with open source Hadoop) to revenue and share declines among data warehousing revenue leaders Oracle, IBM and Teradata (according to IDC estimates).

A little bit about data warehouses The data warehouse refers to a database that is used primarily to analyze data. Data is loaded into the system where it is “warehoused” and used to support reporting and different types of analysis. (A “data mart” refers to the same concept, just a smaller scale system typically used for departmental data). This contrasts with a transactional database that is used to store “live” data for active applications (this could include anything - a point-of-sale system, financial ledger, sales management system, HR system, etc). The primary reasons why data warehouse systems are kept distinct from traditional database systems have to do with performance (analytic workloads are often computationally intensive) and data integrity (data warehousing systems operate on a “single version of the truth” - typically a static set of data, while transactional applications are constantly updating data in the system). As it is, the data warehousing (or analytic data management systems) market is a subset of the overall database market, but with varying industry analyst estimates ranging from US$9-15bn out of a US$30-45bn market.

Some data warehouse market perspective The market is relatively mature as technology goes - Teradata’s database debuted in 1979, a year before Oracle. The capital expense involved with deploying and maintaining data warehousing projects is significant, for large organizations typically in excess of US$1m and even into the tens of millions of dollars. Data warehousing as a practice dates back to the 1980s but challenges of projects early on gave rise to what was known as the “2/2/50 rule” - that data warehousing projects would take 2 years, cost US$2m and had a 50% chance of success. As technologies and best practices continued to evolve into the 2000s, time to deploy and probability of success improved while the cost of traditional systems remained high. Incumbent vendors like Teradata, Oracle and IBM have been able to charge high prices based on delivering business value with highly optimized systems. Given the concentration of established vendors with high costs associated with data warehousing projects, the opportunity has been ripe for new entrants.

Data warehouse appliances the most recent wave of innovation The early 2000s saw a generation of integrated hardware/software data warehouse “appliances” emerge. These included Vertica (acquired by HP), Netezza (acquired by IBM), Datallegro (acquired by Microsoft) and Greenplum (acquired by EMC) and helped deliver better performance for price. Appliance offerings, including Oracle’s Exadata and Teradata’s 2000 series, targeted the lower end of the data warehousing market with reduced technology integration and deployment costs.

The response to the product has been

overwhelmingly positive

Data warehouse refers to a database that is used

primarily to analyze data

The market is relatively mature as

technology goes




An opening for innovation in an established market The market for data warehousing software is highly concentrated. The top five vendors Oracle, Microsoft, IBM, Teradata and SAP account for almost 90% of the market, with the remainder distributed among another dozen leading vendors. According to IDC estimates, Amazon (driven by RedShift) has posted the highest growth of any vendor in the analytic data management software market (237% YoY growth in 2015, with Hortonworks’s Hadoop-related revenue growth of 202% the second fastest). This compares to total market growth of 2.7% YoY to US$9.9bn in 2015.

Figure 22

Worldwide analytic data management software by vendor

Vendor (Product/Database technology)

Revenue (US$m) Share (%) Growth (%)

2013 2014 2015 2013 2014 2015 2013-2014 2014-2015

Oracle (Oracle, Exadata) 3,522 3,629 3,513 38.7 37.8 35.6 3.1 (3.2)

Microsoft (SQL Server, SQL Azure) 1,573 1,725 1,917 17.3 18.0 19.4 9.7 11.1

IBM (DB2, Netezza) 1,682 1,680 1,517 18.5 17.5 15.4 (0.1) (9.7)

Teradata (Teradata) 1,025 1,063 987 11.3 11.1 10 3.7 (7.2)

SAP (HANA, SAP BW) 707 764 837 7.8 8.0 8.5 8.0 9.6

Amazon.com (RedShift, Hadoop) 25 53 178 0.3 0.5 1.8 109.2 237.4

Cloudera (Hadoop) 29 78 141 0.3 0.8 1.4 172.1 80.2

EMC (Greenplum, Pivotal) 56 68 91 0.6 0.7 0.9 22.0 33.8

Hortonworks (Hadoop) 14 26 79 0.2 0.3 0.8 83.9 202.6

SAS (SAS) 92 85 82 1.0 0.9 0.8 (7.9) (3.1)

Hewlett Packard Enterprise (Veritica) 61 65 75 0.7 0.7 0.8 7.3 14.2

Google (Big Query) 25 39 72 0.3 0.4 0.7 55.6 86.1

MarkLogic (MarkLogic) 24 32 41 0.3 0.3 0.4 31.3 29.4

Fujitsu (Symphoware Analytics Server) 37 35 33 0.4 0.4 0.3 (4.8) (7.3)

MapR (Hadoop) 8 16 31 0.1 0.2 0.3 107.2 90.8

DataStax (Cassandra) 14 21 31 0.2 0.2 0.3 50.0 46.5

Other 206 226 246 2.3 2.3 2.5 9.5 9.0

Total 9,099 9,604 9,868 100 100 100 5.6 2.7 Source: IDC

We’d note that the 2015 revenue declines for Oracle, IBM and Teradata are almost fully offset by the revenue gains seen by DWaaS and Hadoop vendors. Together, Oracle, IBM and Teradata saw analytic data management revenue decline by US$365m YoY. Amazon grew revenue US$125m and Google grew revenue US$33m for a total of US$158m. Cloudera, Hortonworks, EMC and MapR combined grew revenue by US$153m in 2015. Together, the US$313m gains from DWaaS and Hadoop vendors do not fully offset declines from Oracle, IBM and Teradata, but they come close. Notably both Microsoft’s SQL Server platform and SAP’s HANA drove growth for those vendors, but SQL Server is a traditional relational database management system and HANA’s in-memory analytics is another distinct technology.

Market forecast of high-single-digit growth IDC forecasts the market for data warehouse generation and management software to expand at an aggregate Cagr of 7.3% to US$17.6bn. Approximately 75% of the total market comes from databases and other data management systems. Data warehouse generation software includes data integration software from the likes of Informatica, SAS, IBM and other vendors.

The top five vendors account for almost 90% of the market

Revenue declines in 2015 for Oracle, IBM and

Teradata are nearly offset by the gains from DWaaS

and Hadoop vendors




Figure 23

Analytic data management software forecast

Source: IDC

Reducing hardware and storage costs - where the opportunities lie The inherent hardware, storage and interconnection costs for a data warehousing system are a big part of the overall expense. Other costs include IT staff setup networking and bandwidth costs for data transfer, software for data integration/transformation (known as extract, transform and load (ETL) and extract, load and transform (ELT)), analytical tools (business intelligence and statistical analysis).

Storage tends to be a commonly used measure of data warehousing costs, as it’s relatively straightforward for vendors to calculate and price their offerings. For Amazon, dynamics of the company’s S3 storage as a service provide an attractive alternative to on-premise systems. Traditional specialized appliance-based data warehouses can cost anywhere from US$20,000 to US$40,000 per terabyte. One alternative to reducing storage costs for analytics comes from Hadoop.

Before we address RedShift, it’s important to put in perspective Hadoop, an open source data-management platform that has gained a lot of traction. Comparing Hadoop with traditional relational database management systems is a bit like comparing apples to oranges. Hadoop systems do not rely on traditional SQL to access data. Hadoop is comprised of a distributed file system (HDFS) and the MapReduce query technology. MapReduce is a completely different query system that requires specialized skills versus the industry standard SQL.

There is a broad ecosystem of open source and proprietary tools that are associated with Hadoop. These include: Hive, an open source Java project that converts SQL to a series of Map-Reduce jobs; Pivotal HAWQ, a suite of tools that enables high-performance analytics on Hadoop; Cloudera’s Impala, an open source, interactive SQL query engine for Hadoop; Facebook’s Presto, which has similar goals to Impala; and Apache Spark, a fast and general engine for large-scale data processing.

Hadoop offers a cheaper cost of storage, which can run US$1,000-3,000 per terabyte. While the analysis tools are not nearly as mature as those for traditional SQL-based data warehouse systems, the low cost lends itself to storage of “cold” data - data that is not accessed on a regular basis. While the

0

1

2

3

4

5

6

7

8

9

0

2,000

4,000

6,000

8,000

10,000

12,000

14,000

16,000

18,000

20,000

2014 2015 2016 2017 2018 2019

Data warehouse generation Data warehouse management Growth (RHS)

2014-2019Cagr=7.3%

(%)(US$m)

Hardware, storage and interconnection costs offer savings from an

SaaS model

Hadoop offers a cheaper cost of storage,

which can run US$1,000-3,000 per terabyte

Open-source Hadoop also a powerful force

in analytics

The data warehousing and generation markets

are over US$12bn in 2014




analytic capabilities are not as robust as say a Teradata or Oracle data warehouse, Hadoop systems provide a lower-cost way to perform low-value data refining workloads (preparing and transforming data for higher-value analysis to be performed on the data warehouse).

It’s important to place open source Hadoop in perspective because it’s had such a significant impact on the data warehousing and analytics software market. There are three major Hadoop distributions (Hortonworks, Cloudera and MapR) with IBM and EMC/Pivotal also focused on the Hadoop market. Hadoop has had deflationary impact on the proprietary database market.

Amazon’s Elastic MapReduce is its Hadoop-as-a-Service offering. Pricing is simple: users pay an hourly rate for every instance hour. Hourly prices range from US$0.011/hour to US$0.27/hour (US$94/year to US$2,367/year) depending on different CPU, memory and storage options.

About RedShift’s technology Amazon’s RedShift was the most high-profile introduction of an SQL-based managed data warehouse as a service for large scale and complex analytics. The underlying technology is a distributed managed ParAccel database cluster. RedShift is particularly suited for speed, achieving high performance through extreme parallelism, columnar data storage, query optimization and smart data compression. Massively parallel processing (MPP) distributes computational workloads across many processers simultaneously. Columnar data storage is an architecture that organizes data in columns (versus rows). Data compression is critical because of the linear relationship between the size of the data set and the performance: the smaller the data set, the faster the query, hence the more compressed the data the better.

Assessing RedShift’s strengths and limitations RedShift appears to offer significant advantages in terms of speed and performance. AirBnB conducted a test comparing it to Hive on Hadoop and found that query times were up to 20 times faster. The SQL interoperability and compatibility with standard Java database connectivity (JDBC) and open database connectivity (ODBC) drivers means that users can integrate with most business intelligence and extract, transform and load (ETL) tools. RedShift can scale easily without users having to buy additional hardware or software. Security is integrated with data encrypted at rest and in transit. Data is protected in case of node failures with data replication and continuous backup.

In terms of shortcomings, users have noted that RedShift is slow in performing certain tasks (such as joins for billions of rows) and maintenance tasks like deleting data. Loading data can take a while (though Amazon’s Snowball data appliance allows users to load data onto the device and ship directly to AWS). RedShift also has limitations on the number of concurrent queries, data sources and data formats. RedShift is not a one-size-fits-all solution, and despite some limitations, user response to the technology has been overall quite enthusiastic, and this is reflected in the robust adoption.

Pricing out RedShift Amazon prices its RedShift service based on the type of instance and usage in a few different ways. Users can pay on demand as they use the service, or they can reserve instances for one- and three-year terms. The three-year term provides the lowest possible costs, under US$1,000 per terabyte per year. On an on-demand basis this can be as high as US$14,000 a year. The

Hadoop has also had a significant impact on the

data warehousing and analytics software market

RedShift offers significant advantages in terms of speed and performance

Amazon prices RedShift based on the type of

instance and usage in a few different ways

RedShift is an SQL- based managed data

warehouse as a service for large scale and complex analytics




primary attraction of RedShift is that users do not need to shell out capital costs for hardware and storage as well as new software licenses. For users that mainly use AWS for storage and other services, RedShift is generally the preferable option. The annual cost can creep up depending on how large the data warehouse and payment preference.

Figure 24

AWS RedShift pricing

Instance and payment type Effective price per TB per year (US$)

On-Demand 1yr reserved instance 3yr reserved instance

dc1.large - No Upfront 13,690 10,950 na

dc1.large - Partial Upfront 13,690 8,795 5,500

dc1.large - All Upfront 13,690 8,620 5,140

dc1.8xlarge - No Upfront 16,425 13,005 na

dc1.8xlarge - Partial Upfront 16,425 11,020 5,500

dc1.8xlarge - All Upfront 16,425 10,800 5,140

ds2.xlarge - No Upfront 3,725 2,970 na

ds2.xlarge - Partial Upfront 3,725 2,190 999

ds2.xlarge - All Upfront 3,725 2,150 935

ds2.8xlarge - No Upfront 3,725 2,970 na

ds2.8xlarge - Partial Upfront 3,725 2,190 999

ds2.8xlarge - All Upfront 3,725 2,150 935 Source: Amazon Web Services On-Demand - Pay for capacity by the hour with no commitments or upfront fees; Reserved Instance - For steady-state production workloads; No upfront - Pay nothing upfront, pay hourly at a 20% discount to on-demand; Partial upfront - Pay a portion of the Reserved Instance and the remainder over a 1- or 3-year term; All upfront - Pay the entire reserved instance upfront. Cheapest option

Comparing RedShift’s pricing to Google and Microsoft is typically a bit of a challenge to get apples-to-apples comparisons. Google’s pricing seems to be cheaper for a fully managed solution. Its pricing for Big Query is calculated based on storage and query size and the company charges a flat rate of US$0.02 per month per GB stored when doing nothing with the data (RedShift charges when servers are running even if they are not doing queries). Users pay US$5 per terabyte for queries. Microsoft’s Azure SQL Data Warehouse charges based on the instance type, with price per hour ranging from US$0.70 to US$20.60 depending on data warehousing units (RedShift starts from US$0.75 to US$9.60 and up). IBM’s new dashDB offering ranges from US$50 per month for up to 20GB of data storage to US$7,370 per month per instance.

Competitive views The database market is still sprawling and fragmented. There are nearly 300 different databases tracked by DB-Engines.com, and even within the data warehousing sub-segment there are strengths and weaknesses of different systems that apply to different types of users and use cases. RedShift is clearly picking up share from Oracle, IBM and Teradata (even as Hadoop-based projects have captured spending away as well).

Marketing for RedShift initially positioned the solution as a cheaper alternative to Teradata and Oracle data warehouses, mirroring strategies of Software-as-a-Service and Infrastructure-as-a-Service vendors as alternatives to on-premise tech. Data warehousing has been slower to migrate to a cloud model partly because of “data gravity” - that data (particularly for analytics) tends

Google’s pricing appears cheaper for a fully managed solution

RedShift initially positioned the solution as

a cheaper alternative to Teradata and Oracle

The database market is still sprawling and fragmented




to stay close to where it’s generated. Moving data from on-premise systems to the cloud was not initially practical given slow transmission speeds, but Amazon benefits as more businesses run their applications on AWS. Additionally Amazon has made it a lot easier to transfer data through innovations like Direct Connect and the Snowball appliance.

Users see RedShift as cheaper than on-premise data warehouses and its performance benchmarks are faster than Hive on Hadoop, Netezza and several other appliance-based solutions. The service is well suited for users that are already running applications on AWS (as holds true for users of Google and Microsoft’s cloud-based DBaaS offerings).

Competition in DWaaS RedShift’s key competitors in Data Warehouse as a Service appear to be Google’s Big Query and Microsoft Azure SQL Data Warehouse as a Service (which is still in beta testing). Both Google and Microsoft’s offerings have different pricing schemes, but offer similar MPP architectures and managed capabilities. Google’s Big Query’s most likely target is users of the Google Compute Cloud, while Microsoft has the potential to draw from a far larger pool of SQL Server and Azure users, so we’d see Microsoft’s potential as far greater. There are a number of startups in the market as well including Snowflake Computing and CoolaData.

Tech giants are responding Oracle offers its Database as a Service primarily to support transactional applications, and to our knowledge has not yet offered a Data Warehouse as a Service. Teradata does offer its technology on a hosted basis for some clients, and is currently developing a version of its technology to run as a service on AWS EC2 and S3. IBM is in early preview of its dashDB fully managed Data Warehouse as a Service.

Competition today and in the future We don’t expect RedShift to continue to grow at 200%-plus in a market projected to grow 7-8% over the next few years, but do believe it’s possible that RedShift could add US$100-125m in incremental revenue every year to reach a US$500m business by 2019. As a first mover offering a disruptive alternative in a mature market, RedShift enjoyed sparse direct competition. Oracle, IBM and Teradata’s on-premise businesses are likely to continue to face challenges, but the overall data analytics market is diverse and specialized enough that other players can grow as well (such as SAP’s HANA and other in-memory solutions, along with the ecosystem of open-source alternatives such as Hadoop, Cassandra and others).

Microsoft has been a formidable force in the market for databases, business intelligence and analytics for over two decades, posting the highest absolute dollar growth in 2015 (US$191m) of any vendor. We believe the challenge from SQL Azure Data Warehouse as a Service will be the most formidable competition for RedShift. Google’s Big Query is also putting pressure on per-hour pricing and bundling services that Amazon charges extra for. For the near term, we’d expect RedShift to remain the premier DWaaS offering, sustaining the competitive headwinds to Oracle, IBM and Teradata’s predominantly on-premise businesses.

The service is well suited for users that are

already running applications on AWS

Key competitors appear to be Google’s Big Query and Microsoft Azure SQL Data

Warehouse as a Service

Microsoft is a formidable force in databases,

business intelligence and analytics

RedShift could possibly add US$100-125m in every

year to reach a US$500m business by 2019




Figure 25

RedShift architectural view

Source: Amazon


Section 3: Making platforms count Software


Making platforms count The Internet of Things refers to a broad range of connected products, services and solutions, with economic impact potentially in the trillions of dollars by 2025. In October 2015, Amazon launched AWS IoT, a managed cloud platform that enables sensors, devices and applications to interact securely with one another. We believe AWS’s scale, breadth and market reach position it to be a leading IoT platform, accelerating adoption of IoT solutions and spurring incremental demand across the portfolio of cloud infrastructure services. We estimate AWS IoT could generate US$200m with an additional US$1.0-1.6bn in related cloud services by 2020.

The IoT platforms market is nascent, fragmented and broadly defined. IoT Analytics estimates the IoT application enablement platforms market at US$298m in 2015, growing to US$1.64bn in 2021, a Cagr of 33%. IDC forecasts IoT purpose-built platforms (a far broader category) growing from US$37.2bn in 2014 to US$84.1bn in 2019. Including related technologies, the addressable market is about US$288bn in 2015, expanding at a 17% Cagr to US$641bn by 2020. We’d estimate AWS IoT revenue will be less than US$50m in 2016, with IoT usage driving multiple times this in revenue for AWS cloud services.

AWS is already the leading provider of cloud-based capabilities essential to IoT, including compute, storage, database and data capture, for a broad array of use cases and third parties such as the Salesforce IoT Cloud, C3 IoT, GE Predix and others. AWS’s dominant cloud services leadership translates directly to IoT advantage as it squares off against Microsoft’s Azure IoT Suite and the Google Cloud Platform.

AWS IoT is a framework for connecting devices, data and applications securely to support IoT use cases. AWS’s ability to scale to billions of devices and trillions of message processes is an inherent attraction, and services are priced on a pay-as-you-go model. A differentiating feature is the ability to create digital device “shadows” that maintain the latest state of a device even while it is offline.

We expect IoT AWS revenue to be a small part of the portfolio near term, but over time we see the platform driving significant revenue from storage, compute, database and analytics services. AWS IoT pulls all of the pieces together and should generate a “virtuous cycle” accelerating adoption of IoT and usage of AWS cloud services.

Going after the broad IoT opportunity platform strategy The Internet of Things broadly refers to a collection of devices and services that ‘can monitor their environment, report their status, receive instructions and even take action based on the information they receive’, according to a definition by The McKinsey Global Institute. The promise of IoT results from the confluence of powerful technology innovations (eg, sensor networks, ubiquitous connectivity, cloud computing, Big Data, Internet Protocol version 6 (IPv6) and other protocols, declining chip and compute costs).

We’ve covered the IoT opportunity extensively in our September 2014 report Deep field: Discovering the Internet of Things, in which we identified several dozen IoT platform offerings. According to research by McRock Capital and IoT Analytics, there are currently over 360 IoT platform offerings that span from connectivity to telemetry data to monitoring and other functions. AWS IoT was launched in October 2015 and the company already has an impressive array of partners and customers.

We’ve covered the IoT opportunity extensively in

our September 2014 report Deep field

Targeting the Internet of Everything

A large market opportunity: fragmented

and still in nascent stages

Using cloud leadership to stake a claim to IoT

AWS IoT: over time a “revenue multiplier” across the portfolio





For Amazon there’s significant opportunity in platforms, though the scope is broadly defined. Research firm IoT Analytics estimates the market for IoT application enablement platforms to grow from US$298m in 2015 to US$1.64bn in 2021 at a Cagr of 33%.

IDC forecasts the market for IoT purpose-built platforms (encompassing device management, connectivity management, application management, dashboard and reporting tools and analytics) at a Cagr of 16.9%, from US$37.2bn in 2014 to US$84.1bn in 2019. If we include other related opportunities around IoT, AWS’s total addressable market around IoT would be close to US$300bn in 2015, expanding at a 17% Cagr to US$641bn by 2020.

Figure 26

Amazon's IoT market opportunity

(US$m) 2015 2016 2017 2018 2019 2020 2015-2020 Cagr (%)

Ongoing service or content as a service 108.5 127.4 149.9 174.9 204.3 234.8 16.7

Growth (%) 30.0 17.4 17.7 16.7 16.8 14.7

Application software 97.8 114.7 135.1 158.1 185.4 215.3 17.1

Growth (%) 11.2 17.2 17.7 17.1 17.3 16.1

IoT purpose-built platforms 41.7 48.5 56.9 66.6 78.4 91.0 16.9

Growth (%) 11.9 16.4 17.4 17.1 17.6 16.1

Analytics software 19.9 24.4 30.3 37.5 46.7 57.4 23.5

Growth (%) 22.9 22.6 24.1 23.6 24.5 23.0

Security software 10.0 12.0 14.4 17.3 20.7 24.5 19.5

Growth (%) 56.0 19.7 20.0 19.6 19.8 18.4

Servers 4.7 5.4 6.2 7.0 7.8 8.7 13.1

Growth (%) 25.3 15.7 14.3 13.3 12.1 10.5

Storage 5.0 5.7 6.5 7.4 8.2 9.0 12.6

Growth (%) 41.5 15.8 14.0 12.6 11.3 9.6

Total 287.6 338.1 399.3 468.8 551.5 640.7 17.4 Growth (%) 20.6 17.6 18.1 17.4 17.6 16.2 Source: IDC, CLSA

We estimate US$1.2-1.8bn opportunity by 2020 for AWS There’s scant data at this point to size AWS IoT’s revenue run rate. We’d assume a similar trajectory to RedShift and estimate AWS IoT revenue would be less than US$50m in 2016, with IoT use cases driving many times this in usage revenue for AWS services. We believe it is likely that IoT platform revenue could exceed US$200m by 2020, with 5-8X or US$1-1.6bn in additional revenue from IoT-related compute, storage, database and analytics services.

Developers will drive adoption The success of IoT platforms will be to a large extent dependent on the vendor’s ability to attract and cultivate developers and build an ecosystem. The winning platforms will be those that attract innovators and build connections across them. Amazon’s strength in the software developer community should prove a distinct advantage. VisionMobile’s global survey of software developers found 53% working on applications in 2015 for connected devices for the Internet of Things. It forecasts 4.5 million IoT developers worldwide by 2020.

For Amazon, there’s significant opportunity

in platforms

The winning platforms will be those that attract

innovators and build connections across them

IoT platform revenue could exceed US$200m

by 2020, with 5-8X or US$1-1.6bn in

additional revenue




Figure 27

The number of IoT developers

Note: Report: Breaking Free from Internet and Things, copyright June 2014 VisionMobile Licenses under CC by ND (Creative Commons with No Derivatives). Source: VisionMobile estimates

About AWS IoT AWS IoT platform includes software development kits (SDKs) that can be used on devices. There is a device gateway that allows devices to speak over the message queue telemetry transport (MQTT) protocol. MQTT is a 30-year old protocol, but it’s fault tolerant and bandwidth efficient.

Figure 28

AWS IoT functional overview

Source: Amazon

0.3

0.8

1.5

2.2

2.8

3.5

4.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0

2014 2015 2016 2017 2018 2019 2020

(m)

AWS IoT is a platform designed to help build Internet of

Things applications

There will be strong demand for developers

with IoT skills in coming years





Connected devices communicate via cryptographic protocols that provide communications security over computer networks: transport layer security (TLS), a successor to secure sockets layer (SSL), using x509 security certificates. Users can map the roles and policies of devices and link them to identity and access management roles. The sensors that talk to the device gateway don’t need to know where they are sending data - only those that subscribe to the data know. This helps the system scale.

The AWS IoT device gateway is fully managed and highly available. There’s also a rules engine that takes messages, orchestrates and routes them. SQL-like statements can be added to the rules engine to detect sensors that exceed data thresholds. The rules engine integrates and routes information to provide alerts that can trigger business processes.

Lambda “serverless computing” allows services to be event-based AWS’s Lambda function is known as “serverless computing”. This is a capability that enables computing services to be used only when there is a need to perform a specific function. For instance, there may be a calculation that needs to be performed with new data once an hour. Rather than maintain server instances running the whole time, a user can set a Lambda function to be triggered by the new data, paying only for the compute cycles used.

One of the differentiating capabilities of AWS IoT is Device Shadows. Any connected device can have its state tracked within the system so that applications can work with the latest state even if the device is offline. AWS IoT Device Shadows can read the last state of a device, then set a desired future state, controlling via representational state transfer application programming interface (RESTful API). There are SDKs for C and Java, and an Arduino (open source hardware platform) library.

Pay as you go, pay extra for compute, storage and other services Pricing is a pay-as you-go model based on the number of messages published to AWS services and the number of messages delivered to devices or applications. Customers are currently charged US$5 per million messages published. AWS IoT does not charge to deliver data to Amazon services like Amazon S3, DynamoDB, Lambda or Kinesis. However, AWS does charge for different services, so users must assemble a portfolio of functions and services needed for their applications.

Already a strong partner ecosystem AWS already has partnerships with leading industrial IoT vendors such as ThingWorx, Schneider Electric and Ayla Networks that help AWS IoT to compete as a cloud-based industrial IoT platform. Other IoT offerings that run on AWS include Salesforce’s IoT Cloud, C3 IoT and GE’s Predix platform (though not exclusively). AWS IoT is complementary to other platforms like Cisco’s Jasper, PTC’s ThingWorx and GE Predix - even though there is some overlap. The range of required functions is so broad that AWS’s strengths in device authentication and management are perfectly complementary to ThingWorx’s mashup building capabilities, GE Predix’s operational analytics and other capabilities. Some of the prominent customers that endorse AWS IoT include Philips, Sealed Air Corp, NASA and Jet Propulsion Laboratory (JPL).

AWS’s Lambda function is known as

“serverless computing”

Pricing is a pay-as- you-go model

AWS already has partnerships with leading

industrial IoT vendors

The AWS IoT device gateway is fully managed

and highly available

Device shadows is a “digital twin”

for offline tracking




Figure 29

The AWS IoT pragma architecture

Source: Amazon

Competition: AWS IoT a strong offering, but Microsoft has slight edge AWS IoT is still early in the market, but the breadth of capabilities and integration with the rest of the AWS portfolio are formidable. At this point only Microsoft’s Azure IoT and Google Cloud Platform have comparable scale, breadth and integration. According to a direct comparison by Raheel Retiwalla, principal architect for IoT and Big Data at BlueMetal Inc, Microsoft’s Azure IoT Suite and AWS IoT are comparable platforms. However, he gives a slight edge to Microsoft for the following reasons. Microsoft has a more mature stream analytics platform that supports SQL-based analysis. Amazon's Kinesis offering is not too far behind, but Kinesis Analytics is still in beta.

Microsoft also offers a machine learning algorithm marketplace and a data marketplace. The PowerBI data visualization service is more mature while Amazon's QuickSight is in preview. Microsoft’s Azure Data Factory data integration capabilities are more robust. Microsoft’s Azure Data Lake Store can be used natively by a wide variety of cloud data services without having to move data around, while with AWS one needs to move data from S3 to use Redshift. AWS IoT’s unbundled pricing model does mean that costs can add up over time for users - and there’s also the need to assemble the appropriate collection of services.

The CloudFront content delivery network Amazon Web Services is arguably the most consequential force in enterprise IT today, but it is far more than just compute and storage as a service. We profile key aspects of AWS’s portfolio and take measure of its potential impact on the competition in new areas. Here, we look at Amazon CloudFront, the company’s content delivery network offering with respect to its potential impact on Akamai and other players in the CDN market.

Microsoft has more mature visualization and stream analytics

platforms

Microsoft also offers a machine learning

algorithm marketplace and a data marketplace

A look at Amazon’s content delivery network offering




What is CloudFront and how is it different? CloudFront is Amazon’s CDN offering, enabling AWS’s >1 million users the ability to have content served from both S3 storage services and CloudFront. CloudFront is appropriate for frequently accessed static content like popular website images, videos, media files or software downloads. Pricing is competitive, comparable to Microsoft’s Azure CDN services (powered by Akamai and Verizon) and Google - but not the cheapest. CloudFront is well positioned for the small- to mid-sized enterprises.

Revenue estimates vary from about US$200-300m a year to US$1.8bn a year (which includes the S3 storage services). Overall CDN market growth forecasts range from 16-25% per year. We estimate CloudFront’s run rate at around US$250m per year, which would make it the second-largest CDN provider. While AWS overall grew 79% in 2015, we believe CloudFront is growing slower than AWS in total, but faster than the CDN market. Over the next five years, we expect AWS CloudFront to gain share of the overall market. At 25% growth through 2020, CloudFront could be a US$763m business, nearly half the size of Akamai’s CDN business.

Pricing in the CDN market is declining at a steady 20% level. AWS’s strategy for its compute and storage services is based around aggressive pricing, but the company is showing increasing operating margins with more value-added services. Akamai uses software engineering to sustain its margin structure, while its performance and security offerings drive growth and sustain profitability. The financials from smaller incumbents like Limelight and ChinaCache reveal difficult market dynamics.

As servers and storage move from hardware to the cloud, so too does security, which is increasingly interlinked with CDN for performance. We see CloudFront naturally gaining share from cloud adoption, with Akamai and integrated CDN/security firms seeing profitable growth driven by security.

Figure 30

CDN market share 2015 (estimated)

Source: CLSA

Akamai22%

AWS7%

Verizon6%

Limelight5%

ChinaCache6%

Others54%

Exact data is scarce, but CloudFront is likely the

No.2 largest CDN

Pricing dynamics favor the big players

and startups

Akamai enjoys leading share in the

market, with AWS No.2




About CloudFront To use the service, customers first store their files on an “origin server” (this can be an Amazon S3 storage bucket, EC2 computing instance or the customer’s own server). They then register the server with the Amazon CloudFront network and use a unique domain name in their web page or application that automatically routes the request to the nearest edge location. Amazon currently operates 54 edge locations globally for content delivery.

Sizing the CDN market opportunity and adjacencies Quantifying the addressable market for CDN services is inherently difficult to delineate. There are over 40 CDN providers actively in the market. With the exception of Akamai, ChinaCache and Limelight, the others are either part of a larger business (as in the case of Verizon, Amazon and Comcast) or privately held. Most estimates size the core CDN market somewhere between US$3.1-4bn in 2015.

We’ve noted that the core CDN market is estimated at around US$4bn in 2014, with growth projected to be in the 25%-plus range through 2020. Industry expert Dan Rayburn of Streaming Media estimates a US$3.6bn market in 2016, growing 14% YoY. Research firm MarketsandMarkets forecasts US$5bn in 2015, at a 26% Cagr to US$12bn in 2019 and US$15.7bn in 2020.

How big is AWS CloudFront? It’s difficult to arrive at an exact estimate for the size of CloudFront. Estimates range from about US$200m a year (Ernie Regalado of Bizety) to US$250-300m a year (analyst firm Stratusly) to US$1.8bn a year (Dan Rayburn of Streaming Media - although his estimates also incorporate the S3 cloud storage services). For our purposes, we’d estimate CloudFront’s run rate somewhere around US$250m per year, which would make it the second-largest CDN provider. We’ve aggregated revenue for leading CDN service providers for comparison. Assuming this is in the ballpark, this would make CloudFront the second-largest CDN vendor in terms of sales.

Figure 31

Estimated revenue for CDN service providers

Provider 2016 annual revenue (US$m)

Akamai (AKAM) 857

Amazon (AMZN) 250

Verizon (VZ) 220

ChinaCache (CCHI) 220

Limelight (LLNW) 182

Level 3 (LVLT) 170

Cloudflare >200

Highwinds 150

Fastly 100

IBM 50

Comcast 25 Source: CLSA for Akamai, Bizety/Stratusly for Amazon, Stratusly for CloudFront, FactSet for ChinaCache and Limelight, Dan Rayburn/Streaming Media for all others

CDN market estimated between US$3.1-4bn

for 2015 . . .

. . . with growth forecasts in the 14-26% range

We’d estimate CloudFront’s run rate

somewhere around US$250m per year




How fast is the CDN business growing? If we assume the overall CDN market is growing somewhere between 16-26%, and Akamai’s media delivery growth will be in the low teens this year (excluding the impact from the top 2 customers’ insourcing efforts), it follows that the market is supporting a number of smaller vendors with higher growth than Akamai.

Akamai has targeted an 18% sustained Cagr in its business (inclusive of M&A) towards a US$5bn annual revenue goal by 2020. This includes performance and security (the highest growth segment). It’s a highly variable business however; looking at YoY revenue growth for Akamai’s media and entertainment vertical from 2008-2013 and the media delivery solutions business segment from 2013 illustrates the wide variability of the company’s CDN services growth. We expect Akamai’s CDN business to return to positive growth in 2017.

Figure 32

Akamai’s content delivery growth is variable from year to year

Source: CLSA, Akamai (reported)

Dan Rayburn has estimated that revenue from S3 storage plus CloudFront increased from US$1.5bn to US$1.8bn from 2014 to 2015, representing a 20% growth rate. Disaggregating storage from our estimated CDN revenue would size S3 storage services at around US$1.5-1.6bn in 2015.

We’d estimate conservatively that CloudFront is growing at a 25% pace, which would size the business around US$200m in 2014 and US$250m in 2015. We assume that AWS CloudFront should be able to continue to grow at this pace through 2020, which would put revenue at US$763m, roughly half of Akamai’s size.

Where the market is headed Based on the growth trajectory of the overall CDN market, we estimate what the overall market will look like in 2020. Assuming the overall market sees a conservative 16% Cagr through 2020, we forecast Akamai’s media delivery business rebounding to 14% growth in 2017 and declining 1% each year. This would reduce Akamai’s share of the market from 23% to 17%. If AWS sees a 25% Cagr, it would increase its share from 7% to 10% of the market. Verizon, Limelight and ChinaCache would all see their market shares decline with less than market growth rates, while the share of others (which include high-growth startups like Fastly and CloudFlare) would increase from 55% to 61% of the overall market.

(10)

(5)

0

5

10

15

20

25

30

2008 2009 2010 2011 2012 2013 2014 2015 16CL 17CL 18CL

Media and Delivery SolutionMedia & Entertainment

(% YoY growth)

We assume the overall market sees a

conservative 16% Cagr through 2020

We expect Akamai’s CDN business to return to

positive growth in 2017

The market is supporting a number of smaller vendors with higher growth than Akamai




Figure 33 Figure 34

CDN market share - 2015 CDN market share - 2020

Source: CLSA, FactSet, company, Dan Rayburn/Streaming Media

The pricing dynamic and the impact on margins Pricing declines are a constant in the CDN market - and the pace appears to have stabilized in the industry for the time being. According to Streaming Media’s Dan Rayburn, 2015 saw pricing declines of 20% on average, with declines for all of 2016 expected to be 20-25%, representing no effective change to the overall environment.

AWS has been the preeminent price leader in Infrastructure as a Service and is well known for its aggressiveness in compute and storage costs. CloudFront is one of the less expensive CDNs, but it is far from the cheapest. Akamai and Verizon do not disclose pricing, so direct comparisons are near impossible.

CloudFront is comparable with Microsoft Azure’s CDN services (they resell both Verizon and Akamai CDN). MaxCDN is the low-cost leader in the market. Of course, pricing is highly dependent on the type of content, location of the edge servers (for instance, prices are higher for Latin America, India and other emerging markets).

Figure 35

CDN pricing across different vendors (US$) CloudFront (AWS) Rackspace Fastly Microsoft-standard CacheFly MaxCDN MaxCDN First 10TB/GB 0.085 0.100 0.12 0.087 0.100 Under 25TB 0.06 Next 40TB/GB 0.080 0.090 0.08 0.080 0.060 51-150TB 0.04 Next 150TB/GB 0.060 0.070 0.060 0.054 351-1023TB 0.03 Next 300TB/GB 0.040 0.045 0.040 0.030 1PB-2PB 0.02 Next 524TB/GB 0.030 0.035 0.030 0.027 3PB-5PB 0.02-0.08 Next 4PB/GB 0.025 0.025 0.025 0.024 5PB+/GB 0.020 0.020 Contact Azure 0.022 Note: Microsoft Azure resells both Akamai and Verizon CDN - Pricing is the same. Source: Companies

Ernie Regalado of Bizety compared CDN pricing and found that KeyCDN’s pricing is lower than CloudFront at the highest tier. Google is the newcomer, matching Amazon CloudFront prices:

CDN77.com: US$0.049/GB - US$0.035/GB

KeyCDN: US$0.04/GB - US$0.01/GB

Google CDN: US$0.08/GB - US$0.02/GB

AWS CloudFront: US$0.085/GB - US$0.02/GB

Akamai23%

AWS Cloudfront

7%

Verizon6%

Limelight4%

ChinaCache5%

Others55%

Akamai17%

AWS Cloudfront

10%

Verizon CDN5%

Limelight4%

ChinaCache4%

Others60%

CloudFront pricing is comparable with

Microsoft Azure’s CDN services - they resell both

Verizon and Akamai

2015 saw pricing declines of 20% on average




One of the tactics that larger customers are employing is the use of multiple CDNs. Formerly, companies of the size of Netflix, Apple and Amazon had the scope of business to justify multi-CDN contracts, but this is becoming easier for enterprises with the availability of multi-CDN platforms from the likes of Cedexis, Conviva and Nice People At Work that enable switching traffic.

In a difficult market, Akamai holds its margins The pricing dynamic is one that Akamai has had to deal with throughout the life of the company, but due to its unique scale and expertise in software engineering, the company has been able to sustain steady margins for its media delivery business. Akamai does not provide granular segment detail for its media delivery solutions on a quarterly basis, but it does disclose this annually at its investor day. Gross margin tends to vary but has averaged around 69% over the past four years. Ebitda margin also improved from 2012 to 2014 but declined in 2015, likely resulting from the deceleration of growth from 2014.

Figure 36

Akamai media delivery solutions financial model (from analyst days) 2012 2013 2014 2015 Cash gross margin (%) 65 69 72 69 Adjusted Ebitda (%) 41 46 50 44 Capex (% of sales) 24 20 19 24 Operational FCF (% of sales) 17 26 31 20 Revenue (US$m)¹ 653 757 912 977 YoY growth (%) 15 17 21 10 ¹ segment revenue disclosures include re-classified offerings. Source: Akamai

Pure-play CDNs struggle in a game of scale As a pure-play CDN, Limelight operates at far lower margins than does Akamai and has struggled to deliver growth in the business.

Figure 37

Limelight financial model 2012 2013 2014 2015 Gross margin (%) 38 37 40 41 Ebitda margin (%) 5 1 3 4 Capex (% of sales) 10 11 11 14 Operational cashflow (% of sales) 9 3 0 4 Revenue (US$m) 180 173 162 171 YoY growth (%) 5 (4) (6) 5 Source: Limelight, FactSet

ChinaCache is the largest public pure-play CDN in Asia, but the company too has also seen margins and revenue growth decline as it retrenches the business following operational issues with its platform. The company’s margins are below Limelight and Akamai.

Figure 38

ChinaCache financial model 2012 2013 2014 2015 Gross margin (%) 31 31 30 23 Ebitda margin (%) 10 na na na Capex (% of sales) na 11 11 9 Operational cashflow (% of sales) (5) (6) (14) Revenue (US$m) 131 182 223 209 YoY growth (%) 40 22 (6) Source: ChinaCache, FactSet

The growth of multi- CDN contracts

Akamai’s media delivery cash gross margins are

high 60s, Ebitda in low to mid 40s

Limelight’s margins are high 30s- 40s, Ebitda in

low-single digits

ChinaCache’s gross margins around 30%

except 2015 with negative Ebitda




What’s defensible - and what is a commodity? It’s increasingly apparent that CDN services are a commodity and the key for leading providers is higher-value services. It is a game of scale and pricing, and there are other aspects at work. According to Dan Rayburn, referring specifically to CDN for media delivery, ‘CDN is a check box. It’s a platform play now. Amazon acquired Elemental. Level 3 acquired Servecast. Limelight acquired Delve. Verizon acquired Uplynk and Volicon. Comcast acquired the Platform. IBM acquired Ustream and Clearleap. Live linear platforms, holistic video ecosystem are the future.’ For Akamai, there may be some risk in 2016 as Amazon increasingly brings Amazon instant video delivery onto its own platform (Dan Rayburn estimates this could account for US$20-25m in revenue in 2016), but this is a small portion of Akamai’s overall business.

Figure 39

CDN market

FY15 FY16CL FY17CL FY18CL FY19CL FY20CL

Akamai CDN revenue (US$m) 869 858 974 1,101 1,233 1,368

YoY growth (%) 7 (1) 14 13 12 11

Market share (%) 23 19 19 19 18 17

AWS CloudFront revenue (US$m) 250 313 391 488 610 763

YoY growth (%) 25 25 25 25 25 25

Market share (%) 7 7 8 8 9 10

Verizon CDN revenue (US$m) 220 246 276 309 346 388

YoY growth (%) 12 12 12 12 12

Market share (%) 6 6 5 5 5 5

ChinaCache revenue (US$m) 209 220 240 262 286 312

YoY growth (%) 5 9 9 9 9

Market share (%) 6 5 5 4 4 4

Limelight revenue (US$m) 171 182 206 230 258 289

YoY growth (%) 6 13 12 12 12

Market share (%) 5 4 4 4 4 4

Others revenue (US$m) 2,081 2,589 3,027 3,542 4,148 4,862

YoY growth (%) 24 17 17 17 17

Market share (%) 55 59 59 60 60 61

Total market revenue (US$m) 3,800 4,408 5,113 5,931 6,880 7,981

YoY growth (%) 16 16 16 16 16 16 Source: CLSA. FactSet, companies, Dan Rayburn/Streaming Media

Security - the big differentiator and market growth driver We believe security will increasingly be the preferred avenue for Akamai and other CDNs to drive profitable growth. Security gross margins are in excess of 90% even for SaaS offerings. Security is a market with powerful secular fundamentals, a rich and dynamic opportunity for innovation and an attractive market for Amazon, Akamai and other vendors. With the broader shift from on-premise IT to cloud computing, next-generation security is moving into the cloud as well. The CDN market increasingly moved into security, and the new cloud security could likely become a multibillion dollar market over the next five years.

CDN is a check box. It’s a platform play now

Security is an attractive market for CDN vendors




Akamai has already established significant security market share, with its Kona and Prolexic offerings driving 45%-plus growth to a US$350m annual run rate for 2016. There’s a whole generation of startups and new entrants to the market that bundle CDN and security. These include CloudFlare, ZScaler, Cato Networks and Incapsula (part of Imperva). The total IT security market was sized at roughly US$32.8bn in 2014 and is projected by IDC to expand at a 7% Cagr through 2019. Akamai’s security business exceeded the 10% revenue threshold in 2015 and has grown at a 45%-plus organic clip the past three quarters, as its DDoS mitigation and other security services complement the core offerings.

DDoS mitigation and web application firewalls The key security offerings that CDN providers offer are distributed denial of service (DDoS) protection and web application firewalls (WAF). These are now differentiating features. According to Bizety, there are a number of CDNs with DDoS Mitigation and WAF, including Akamai, Verizon, Level 3, CDNetworks, Limelight, CloudFlare, Distil Networks, Fireblade, Fastly, Yottaa, Internap, Instart Logic, Incapsula/Imperva and Zenedge. Providers without DDoS mitigation include Amazon CloudFront (AWS offers a WAF), Highwinds, Mirror Image, Cachefly, MaxCDN, HiberniaCDN, KeyCDN, CDN77, CDNify, Streamzilla, Twin Prime and Kwicr. Amazon’s security offerings at present are limited, but we believe this represents a significant opportunity for revenue expansion.

Amazon CloudFront pros Amazon Web Services has over 1m customers on its platform, which

provides an unmatched base to sell CDN services. There are likely hundreds of thousands of CDN clients (CloudFlare, the market leader has over 4m clients). As long as clients sign up for AWS, it should be easy to add CloudFlare services.

CloudFront CDN pricing is competitive for small-medium-sized businesses/enterprises (SMB/SME) customers and the company will continue to be aggressive to expand the market.

The dynamic content delivery option allows customers to choose between S3 storage for large files and CloudFront caching for smaller objects.

Amazon CloudFront cons CloudFront’s overall performance ranks in the lower third of 40-plus CDNs

in the market, according to Bizety, making it less appropriate for large enterprises. Because the company’s networks are not dedicated CDN networks, it’s more difficult to guarantee service levels on a consistent basis for CDN.

CloudFront is better suited to customers already using S3 storage, while those storing content on their own origin servers may be better served by using a 3rd-party CDN.

Compared to Akamai, Amazon has fewer international points of presence.

Amazon lacks DDoS capabilities.

Amazon CloudFront opportunities Amazon is making increasing investments in video delivery, acquiring

Elemental’s software defined video platform.

AWS has WAF but lacks significant DDoS protection

Amazon Web Services has over 1m customers

on its platform

CloudFront’s overall performance ranks in the

lower third of 40-plus CDNs in the market




AWS’s impact on the application delivery controller market Application delivery controllers optimize network performance and web-based applications. An ADC appliance or software performs functions like compression, caching, security, offloading of encryption and integrating with load balancing function in the network. Basically ADCs help the network and apps run faster. The market is at US$1.8bn, expanding at a 4% Cagr.

Figure 40

Worldwide application delivery controller revenue snapshot

Source: IDC

Leading vendors in the market with 2015 revenue, according to IDC:

F5 Networks - US$867m (47.6% share)

Citrix - US$468.2m (25.7% share)

Cisco - US$15.1m (0.8% share)

Others represent US$109m (6.0% share), including Radware, A10 Networks and Brocade

What’s happening in the ADC market? When companies have been migrating applications from on premise to the cloud (and to AWS), they have been using existing software products (buying them through the Amazon Marketplace). Some vendors like Citrix and F5 note increasing sales from their virtualized software through AWS.

Application delivery controllers optimize

network performance and web-based applications

A US$1.8bn market, expanding at a 4% Cagr

Some vendors note increasing sales from

their virtualized software through AWS




What’s AWS doing? AWS is involved in a range of network services that are a natural extension of IaaS capabilities. AWS’s application delivery controller functions accelerate the delivery of application particularly for enterprises, and span several offerings, most notably elastic load balancing and the recently launched application load balancing service. IDC estimates that the services that comprise AWS’s ADC capabilities (which also include CDN, the Route 53 networking services, web app firewall and API gateway) generate around US$500m annually. We expect application performance capabilities to grow in line with overall revenue and play a key role differentiating products.

What could happen? As customers explore Amazon’s offerings over time, it’s possible that they will choose the AWS native services rather than continue to purchase software from the 3rd parties. The fundamental premise for ADCs is to handle a lot of the tasks necessary to optimize apps running in a data center. When you move to the cloud, a lot of these management functions are handled by AWS. Over time, customers could choose the AWS services over 3rd-party controllers. This has not shown up in a meaningful way yet, but based on conversations we believe this could be a medium-term phenomenon.

We don’t expect AWS to impact NetScaler in a meaningful way NetScaler represents about 23% of Citrix’s business pre spinoff of GetGo and will be about 1/3 post spin. Roughly 35-50% of Citrix’s ADC revenue comes from service providers (not at risk) with roughly 1/3 from enterprise. We don’t see any risk from AWS to hyperscale internet customers like Apple, Yahoo, Google, Microsoft and even Amazon - even though the business tends to be lumpy from quarter to quarter. Additionally, NetScaler’s expanded management and security capabilities offer a far richer and deeper set of capabilities, and they are fully integrated as well.

For Citrix, we estimate that between 5-10% of NetScaler-optimized enterprise workloads (2-3% of total or US$14-22m) could be impacted in 2017-18 from customers switching from NetScaler software to a portfolio of AWS optimization services.

AWS is involved in a range of network services

Customers over time are likely to choose AWS

native services

We don’t see any risk from AWS to hyperscale

internet customers


Section 4: Above the PaaS layer - towards user apps Software


Above the PaaS layer - towards user apps As more and more users generate and store data on Amazon Web Services, the company is providing more tools for its customers to analyze and derive value from that data. We expect the QuickSight business intelligence offering to disrupt on-premise BI vendors’ pricing models, though Microsoft PowerBI has taken the lead on cost and availability. For Splunk, Kinesis has become quite complementary, while potential threat is muted from Elasticsearch Service’s limited capabilities.

AWS’s QuickSight business intelligence offering enables users of Amazon’s data services to easily create queries, reports and visualizations starting at US$9-18 per user per month. QuickSight is currently in preview, but we’ve heard from partners that the subscription model is luring buyers away from perpetual license vendors like Tableau. The product is not generally available yet, but we expect the as-a-service model to exert pricing pressure on traditional BI offerings from other vendors including SAP, Oracle and IBM. We’d expect a similar revenue ramp to what we saw with RedShift (US$25m the first year growing 200%-plus). However, Microsoft has beaten AWS to market with PowerBI Pro with lower pricing and a richer feature set. We expect both vendors to use respective BI offerings to drive adoption of their broader cloud portfolios.

AWS’s Elasticsearch Service allows users to analyze data using the open-source project along with Logstash and Kibana (the ELK stack). The open-source license cost is free, but capabilities are limited, setup time is far slower and AWS does not support the latest version. The Kinesis streaming data capture service started with some analytic capabilities seen initially as competitive to Splunk; however, as the product has matured, it’s increasingly deployed as a joint solution feeding data to Splunk’s rich analytics.

AWS’s machine learning (ML) service enables users to incorporate ML capabilities as components in applications. Amazon currently has >1% of a broadly defined US$4.5bn cognitive software market. We’d expect rapid growth off a small base in a greenfield market opportunity. The Elastic MapReduce service provides access to a growing array of open-source Hadoop capabilities, but this market remains nascent and the self-service nature of the solution appeals to a segment of highly savvy technical users.

Looking at AWS’s intelligence offerings The market for business intelligence and analytics tools is large (over US$15bn in 2015) and likely to grow at nearly 9% through 2020. The market is relatively mature, but demand continues to be driven by the surge of data generated by new and existing applications.

Going up the stack to create value from

all that data

QuickSight’s disruption will be the subscription

business model

Elasticsearch Service lags, Kinesis

complements Splunk

Targeting a large and growing market

opportunity




Figure 41

Worldwide business intelligence and analytics tools forecast

(US$m) 2014 2015 2016 2017 2018 2019 2014-2019 Cagr (%)

Advanced and predictive analytics 2,423 2,562 2,801 3,058 3,345 3,664 8.6

Query, reporting and analysis 10,079 10,603 11,488 12,468 13,516 14,653 7.8

Spatial information analytics 873 904 967 1,040 1,119 1,208 6.7

Content analytics and cognitive software platform

1,130 1,331 1,572 1,845 2,143 2,453 16.8

Total 14,505 15,400 16,827 18,411 20,124 21,979 8.7 Source: IDC

The newer market for big data technology infrastructure offers a particularly attractive opportunity for Amazon, with IDC estimating a market at US$11bn in 2015, forecast to grow nearly 22% through 2019. This is an obvious target market for Amazon’s infrastructure services, and AWS has been quick to capitalize on opportunities with its compute, storage and data services.

Figure 42

Worldwide big data technology infrastructure revenue by segment

(US$bn) 2014 2015 2016 2017 2018 2019 2014-2019 Cagr (%)

Compute 2.97 3.70 4.60 5.70 7.01 8.61 23.7

Storage 4.45 5.62 7.00 8.54 10.20 12.02 22.0

Networking 0.88 1.02 1.19 1.36 1.55 1.76 14.9

Other infrastructure 0.55 0.65 0.77 0.90 1.06 1.23 17.3

Total 8.85 11.00 13.56 16.50 19.82 23.62 21.7 Source: IDC

Meanwhile, big data software offers even more robust growth potential.

Figure 43

Worldwide big data technology software revenue by segment

(US$bn) 2014 2015 2016 2017 2018 2019 2014-2019 Cagr (%)

Information management software 1.66 2.04 2.53 3.15 3.89 4.73 23.3

Discovery and analytics software 1.93 2.49 3.22 4.13 5.20 6.63 28.0

Applications software 0.41 0.52 0.66 0.86 1.10 1.46 28.9

Total 4.00 5.05 6.41 8.14 10.19 12.82 26.2 Source: IDC

Lastly, there’s a large market for business analytics services, accounting for US$58.7bn in 2015 global spending across a range of professional services. This is important for two reasons: First, because it’s large and growing, it attracts professional services firms investing in practices to serve demand. Second, the potential benefits in cost savings and agility from the automation of manual processes through the cloud model makes AWS an attractive option to traditional on-premise software. For Amazon, the opportunity lies in tapping into the budgets for systems integration and application management, which together accounted for US$18.6bn of spend in 2015.

Large market for business analytics services

Big data technology infrastructure offers a particularly attractive

opportunity for Amazon




Figure 44

Worldwide business analytics services revenue snapshot for select markets

Source: IDC

It’s been a logical evolution for Amazon Web Services to expand into different flavors of analytic offerings, given the massive amounts of data that users generate and store on the various data services. These include Amazon Redshift Data Warehouse as a Service, Amazon RDS (relational database service), Amazon Aurora (database), Amazon EMR (Elastic Map Reduce), Amazon DynamoDB (database), Amazon S3 (storage) and Amazon Kinesis (streaming data). We look at Amazon QuickSight (business intelligence), Amazon Elasticsearch Service, Amazon EMR (Elastic Map Reduce), AWS Data Pipeline, Amazon Kinesis and Amazon Machine Learning.

QuickSight: Business intelligence as a service QuickSight enables users to build visualizations, perform ad-hoc analysis and quickly get insights from data. QuickSight is one of AWS’s newer offerings. It was introduced on 7 October 2015 at re:Invent and is currently in preview. The QuickSight business intelligence product is still in beta, but based on user interest we believe it’s poised to be disruptive to traditional offerings from the likes of Tableau, SAP, IBM and Oracle (Microsoft PowerBI’s deeper features are more aggressively priced). The product will offer self-service capabilities that are designed to accelerate the ability to realize analysis from data stored in AWS. There are some unique offerings including super fast, parallel, in memory calculation engine (SPICE) that will differentiate the offering against BI incumbents. We attended a preview session that outlined a rich array of features coming with the formal release likely at the end of 2016.

Massive spend on service provides ripe opportunity

for AWS and other cloud alternatives

QuickSight business intelligence product

is still in beta

Logical evolution for Amazon Web

Services to expand into different flavors of

analytic offerings




A fragmented market The market for business intelligence tools is relatively fragmented, with market leader SAP accounting for less than 16% share. The top 6 vendors accounted for roughly 59% share in 2014 (by comparison the top 6 data warehousing vendors accounted for nearly 90% share). As the market is already fragmented, there’s quite a bit of competition (and much opportunity for differentiation). However, the predominant business model is traditional, on-premise licenses.

Figure 45

Business intelligence and analytics tools market share 2014

Source: IDC

Taking a look at QuickSight Traditional business intelligence can cost between US$150-250 per month, with lengthy implementation cycles. Amazon is not trying to achieve feature parity - the goal is a fast experience for 1/10 the cost of traditional BI. The experience will enable business users to analyze data easily and quickly. The service is cloud-powered so there are no back-end requirements for the user. The service starts at US$9/month or US$108/ year for a one-year commitment.

For now, AWS is not exposing its back-end API’s. There are a number of services to connect data to QuickSight, including flat file uploads, Amazon RDS, RedShift and S3. There will also be support for Kinesis streaming data, DynamoDB, EMR and 3rd-party web applications like Salesforce. Data prep is an optional step that will enable users to preview reports and change the presentation. The metadata is stored on the back end (not in the data source) and replicated. Auto-Graph feature detects data types and optimally generates queries.

SPICE is an intermediate, optional in-memory columnar data store and calculation engine. Compression speeds up performance by 2-4X. In some respects, SPICE is a data mart that’s fully managed. QuickSight will offer suggestions to users based on machine learning user behavior and automatically generate visuals that users might want to see. AWS will open up to 3rd-party tools like Tibco Spotfire, Qlik and Domo so that users can mix and match different capabilities. At launch, there will be native support for IOS with support for Android to follow.

$14,505M▲7.5%

SAP $2,305, +6.6% y/y

IBM $1,532, -4.6% y/y

MSFT $1,396, +8.0% y/y

SAS $1,300, +4.6% y/y

Rest $5,512 +11% y/y

Tableau $399, +77.3% y/y

Oracle $1,110, +1.3% y/yQLIK $503, +16.6% y/yMicrostrategy $447, +2.1% y/y

QuickSight targets a large market with

an aggressively priced offering

The market for business intelligence tools is

relatively fragmented

Traditional business intelligence can cost

US$150-250 per month




Potential for significant disruption In our view, QuickSight’s business intelligence offering has the potential to disrupt incumbent offerings (from the likes of SAP, IBM, Oracle, Tableau and Qlik), not so much from the standpoint of functionality but from the aggressive on-demand subscription pricing model. Microsoft’s PowerBI as a service has a head start, both on cost and functionality.

For users of existing AWS services, QuickSight offers a number of attractive capabilities. There’s a data discovery function that identifies data stored in the various AWS services. QuickSight formats the data and moves it to SPICE, which can perform advanced calculations and render visualizations rapidly. SPICE uses columnar storage, in-memory technologies, machine code generation and data compression to enable users to run interactive queries on large datasets.

QuickSight offers users the ability to connect easily to AWS data services, including Amazon Redshift, Amazon RDS, Amazon Aurora, Amazon EMR, Amazon DynamoDB, Amazon S3 and Amazon Kinesis. Users can also upload comma separated value (CSV), tab separated value (TSV) and spreadsheet files or connect to third-party data sources such as Salesforce.

Figure 46

QuickSight screenshot

Source: Amazon

QuickSight offers basic visualization and reporting capabilities. A suggestion engine provides users with recommended visualizations based on the characteristics of the underlying data. StoryBoards enable users to annotate specific stages of analysis to tell a story or make a point. We expect that users demanding more sophisticated visualizations or analysis may be inclined to use Tableau or Qlik in conjunction with QuickSight.

Reporting and visualization capabilities

appear basic for now

QuickSight’s business intelligence offering has the potential to disrupt incumbent BI offerings

QuickSight offers users the ability to connect

easily to AWS data services

QuickSight offers basic visualization and

reporting capabilities




Aggressive pricing, though Microsoft PowerBI is cheaper Pricing for QuickSight starts at US$9/user/month for basic services to US$18/user/month for enterprise users (which adds features like active directory integration) with 10GB of SPICE storage with a one-year commitment. Qlik offers Qlik Sense Cloud Basic for free and Cloud Plus for US$20/user/month with maximum storage of 10GB. Tableau Online charges US$500/user/year with Tableau Desktop required to author. Salesforce Wave is an add-on BI offering for Sales Cloud and Service Cloud users, with pricing that ranges from US$75 to US$150/user/month. For primarily on-premise BI vendors (which include SAP, Oracle, IBM, Tableau and Qlik), we expect QuickSight’s aggressive pricing and subscription model to exert deflationary pressures on license growth. Other SaaS BI vendors such as Good Data, Birst, Domo and others offer opaque pricing models.

The most directly competitive offering is Microsoft’s PowerBI, which is free for up to 1GB/user. For the PowerBI Pro version, Microsoft charges US$9.99/user/month, which includes the ability to refresh from live data sources, collaboration and 10GB/user storage. Microsoft’s PowerBI Pro offers more functionality, integration with Excel and other Microsoft applications at a lower price. We believe Microsoft will use PowerBI to attract users to Azure versus AWS.

How big can it be? We’d note that in its first year offering RedShift, Amazon generated an estimated US$25m in revenue, according to IDC. We’d estimate QuickSight could see a similar trajectory as an inexpensive add-on for existing AWS customers. We’d expect both AWS and Microsoft to take share from Tableau, Qlik and legacy BI offerings from SAP, Oracle and IBM.

Elasticsearch Service - alternative analytics In November 2015, AWS launched a managed service that enables users to run the open-source Elasticsearch in the cloud. Elasticsearch is an open-source search and analytics engine, typically deployed with Logstash for data collection and Kibana for data visualization. The combination of Elasticsearch, Logstash and Kibana is commonly known as the ELK stack. The AWS service provisions the resources for the ELK cluster and launches it, and also automatically detects and replaces failed nodes. Amazon Elasticsearch is used for IT-related uses such as log analytics, real-time application monitoring, and click stream analytics. AWS also offers Cloudsearch, a search engine that EC2 users can use to add search to their websites.

The service is priced per instance, based on usage. Customers are charged for Elasticsearch instance hours, storage (if chosen) and data transfer fees. There’s a free tier for up 750 hours per month of a single instance and up to 10GB of optional storage. Pricing ranges from US$0.018 per hour for a single CPU with EBS storage to US$3.92 per hour for a memory optimized instance with 244GB of memory and 32 CPUs. EBS storage costs from US$0.067/GB per month for magnetic storage to US$0.169/GB per month for AWS provisioned solid state storage. Elastic (the company) also offers a managed service on AWS with monitoring, alerting and notifications and different tiers of support starting at US$45/month. Pricing is not apples to apples - while AWS Elasticsearch charges based on CPU usage and storage at rest, Splunk Cloud offers monthly subscriptions for US$1,380 per month for 10GB of data analyzed per day.

AWS provides a managed service to run

Elasticsearch in the cloud

Pricing for QuickSight starts at US$9/user/mth

for basic services

The most directly competitive offering is

Microsoft’s Power BI

The service is priced per instance, based on usage




There are shortcomings to Elasticsearch Service, according to users. There were limitations to the number and types of nodes, which limit the ability to scale. In May 2016, AWS doubled capacity to 20 data nodes from 10 previously. The AWS-supported version of Elasticsearch is 1.5.2, while the latest version of Elasticsearch is version 2.3. The service also does not support Elasticsearch’s commercial products including Shield for security, Watcher for alerts and notifications, and Marvel for cluster monitoring. According to users, Splunk sets up more quickly as compared to the ELK stack, while both take some time to learn and apply. Both also provide a powerful interface model for data visualization.

At this point, Elasticsearch Service does not appear to be a meaningful threat to Splunk’s core business. Although Elasticsearch and Splunk can be used to analyze and visualize log data, the Splunk product is more feature-rich with a broader ecosystem of over 1,000 applications and templates available on the Splunk store. We’d note that Splunk Cloud is a featured solution on the Partner Directory page of the Amazon Web Services site as of 6 July 2016.

Amazon Kinesis - high-volume data streaming complements Splunk In December 2013, Amazon Web Services launched Kinesis, a managed service for processing streaming data in real-time. The technology supports a variety of custom applications that enable real-time processing of streaming data from a variety of sources. Kinesis acts as a conduit to stream messages between producers and receivers of data. Data can come from nearly any source: telemetry connected devices, system data, web logs, mobile apps, social networks, financial trading information, geospatial data, etc. Data is streamed to processing and storage services such as Hadoop, S3 storage, Splunk and Elasticsearch. Amazon offers Kinesis Firehouse, Kinesis Streams (the original offering) and is currently developing Kinesis Analytics, a way to run standard SQL queries against streaming data.

Kinesis Firehose is a managed service to capture and automatically load streaming data into Amazon S3, Amazon Redshift and Amazon Elasticsearch Service. The service enables near real-time analytics using existing business intelligence tools and dashboards. Pricing is based on volume of data ingested into Firehose, which is calculated as the number of data records. The price is US$0.035 per GB of data ingested, with additional costs for S3, RedShift and Elasticsearch Service usage.

Kinesis Streams is the name of the original Kinesis service. Kinesis Streams is a service that captures and stores data from sources like website clickstreams, financial transactions, social media feeds, IT logs and location-tracking events. The Kinesis Client Library provides tools to build apps that use streaming data for real-time dashboards, alert and other use cases. Pricing is pay as you go.

Amazon Kinesis does perform some of the functions Splunk offers. These include streaming data, processing data and generating high-level statistics that can be used for real-time dashboards and analytics. For more advanced analytics or to run queries, users will need to use other tools like RedShift or Elastic MapReduce. Since the introduction of Kinesis, Splunk users have found that the Kinesis streaming and capture capabilities complement the

Elasticsearch is not currently a threat to

Splunk’s core business

There are shortcomings to Elasticsearch Service,

according to users

Kinesis acts as a conduit to stream messages

between producers and receivers of data

Splunk users find Kinesis capabilities enhance

Splunk real-time operational intelligence




Splunk real-time operational intelligence capabilities, and the two technologies are increasingly deployed together. For Splunk, Kinesis appears more complementary, while Elasticsearch Service provides comparable functions for log analysis and other use cases for users with extensive programming skills.

Figure 47

Amazon Kinesis - Firehose, Analytics and Streams

Source: Amazon

Amazon Machine Learning - AI for the masses Amazon Machine Learning is a service that makes it easier for developers to use machine-learning technology. The service offers visualization tools that guide users through the process of creating machine-learning models without having to learn complex algorithms and technology. The service is highly scalable and can generate billions of predictions daily.

Amazon Machine Learning is a service that makes it

easier for developers to use machine-

learning technology




Figure 48

AWS Machine Learning screenshot

Source: Amazon

Pricing is pay per use with no minimum fees or commitments. AWS charges an hourly rate for the compute time used to build predictive models and then users pay for the number of predictions generated for their application. There is a data analysis and model building fee of US$0.42 per hour. Batch predictions cost US$0.10 per 1,000 predictions, rounded up to the next 1,000. Real-time predictions cost US$0.0001 per prediction, rounded up to the nearest penny.

Figure 49

Worldwide cognitive systems, content analytics and discovery software 2015

Source: IDC

Pay per use with no minimum fees or

commitments

Creating ML models without having to learn

complex algorithms and technology

Amazon generated around US$16m in

revenue from cognitive software in 2015




IDC estimates that Amazon generated around US$16m in revenue (+69% YoY) from cognitive software in 2015. This represents just 0.3% share of a broadly defined US$4.8bn market that grew 11% YoY. The market for applied predictive analytics/AI/machine learning tends to be dominated by application companies like Nuance (speech recognition) and Palantir (deep pattern recognition), with Microsoft, Google and IBM monetizing the application of the technologies. Amazon’s Machine Learning Service, IBM’s Watson and Microsoft’s Azure Machine Learning platform are early entrants offering developers machine-learning components. We expect Amazon to continue to grow at a similar trajectory, but given the wide open nature of the market in nascent stages, we see little to no competitive impact on other firms for the time being.

Amazon Elastic MapReduce for Hadoop as a Service Amazon offers an Elastic MapReduce service, launched in 2009, that enables users to take advantage of big data capabilities. This is a managed Hadoop framework that runs on the EC2 computing cloud or S3 storage service. MapReduce was developed to sort and navigate large batch processing jobs in Hadoop. Pricing is straightforward - users pay an hourly rate for every instance hour used. Hourly prices range from US$0.011/hour to US$0.27/hour (US$94/year to US$2,367/year) depending on whether customers are accessing on-demand or reserved instances.

The service supports Apache Spark (an open-source distributed analytic processing system for big data workloads), Presto (an open-source distributed SQL Query engine for ad hoc data analysis), Apache Hadoop (the open-source data processing system) and Apache HBase (a massively scalable, distributed big data store in the Hadoop ecosystem). AWS EMR can handle tasks such as web indexing, scientific simulation, log analysis, bioinformatics, machine learning, financial analysis and data warehousing.

Allied Market Research forecasts the Hadoop-as-a-Service market at a 71% Cagr to reach US$16.1bn by 2020. Competition includes Microsoft’s Azure HDInsight, which is built off a partnership with Hortonworks, as well as managed offerings from the likes of CSC Big Data Platform as a Service, CenturyLink, Google Compute Platform, IBM BigInsights on Cloud and Rackspace Big Data. There are also a number of startups including Altiscale, Qubole, Mirantis and Skytap.

Implementing Hadoop technology at scale is not a trivial undertaking, and to date the hype around Hadoop has outpaced the ability for smaller firms to implement Hadoop technologies. The managed service approach makes a lot of sense by taking the complexity out of implementation. AWS’s MapReduce service does require a significant level of expertise and management capabilities by the user, and costs can grow to be quite significant. Additionally, Amazon has been slower to support all of the projects in the Hadoop ecosystem to the same extent as others, although the service continues to add capabilities. Managed-Hadoop-as-a-Service companies like Altiscale handle more of the tasks that AWS leaves to users to handle.

A drop in the bucket so far for a large

market opportunity

Hadoop-as-a-Service market should increase at

a 71% Cagr to reach US$16.1bn by 2020

Elastic MapReduce enables users to take advantage of big data

capabilities




Figure 50

Amazon EMR Release Velocity includes more of the Hadoop ecosystem

Source: Amazon

Data Pipeline - moving data from place to place Amazon’s Data Pipeline is a data orchestration web service that helps users manage workflows around the movement of data. Data Pipeline allows users to move data between the different AWS services such as Amazon S3 storage), Amazon RDS (relational database service), Dynamo DB (NoSQL

Amazon’s Data Pipeline is a data orchestration

web service




database service) and Amazon Elastic MapReduce as well as on-premise data sources. Customers define a pipeline based on the data sources and destinations for their data. For example, users could define an hourly job that runs on Elastic MapReduce based on AWS S3 log data, and then load the results into a relational database that sends a daily summary. Data Pipeline incorporates some capabilities of ETL technologies from the likes of Informatica, IBM, Dell’s Boomi and SnapLogic, although it does not offer the breadth of transformation capabilities. A comparable offering is Microsoft’s Azure Data Factory. Pricing is based on how often jobs run. Pricing ranges from US$0.60-2.50 per month per pipeline depending on how activities are scheduled and where they run (AWS or on premise).

Targeting business users with email, file sharing AWS’s portfolio of offerings is predominantly concentrated in IT infrastructure with a few exceptions. The WorkDocs secure file-sharing service and WorkMail hosted email services target business users as alternatives to Microsoft’s Office365 and Google Apps for Work. Amazon’s offerings do not particularly stand out in terms of features or pricing, and revenue so far is small. Given Microsoft and Google’s aggressive efforts to expand established market share, we believe it is likely Amazon will focus the bulk of its AWS efforts in other markets.

WorkDocs and WorkMail target a US$13bn market for collaborative applications in which Microsoft’s 35% share leads Google’s 7.8% in a market with hundreds of competitors. Amazon is a new entrant with just US$31m in estimated 2015 revenue (0.2% share) and 86% YoY growth. AWS WorkMail and WorkDocs are neither the cheapest, the fastest-growing services nor the most richly featured in the market, but provide integrated extensions to the WorkSpaces Desktop-as-a-Service offering.

IDC estimates the worldwide email applications market at US$4.3bn in 2015, expanding at a 4.1% Cagr to US$5.1bn in 2019. A Gartner study early this year found 13% of business using Google and Microsoft’s cloud email services, with 8.5% using Office365 and 4.7% using Google Apps for Work. AWS’s WorkMail costs US$4-6 per user per month, comparable in pricing to the lower tiers of Microsoft and Google’s offering, but without features like conferencing and document editing.

The US$2.4bn file synchronization and sharing software segment accounts for roughly 18% of the collaborative apps market as of 2015, likely to grow to 27% by 2020. Microsoft, Google, Box, Dropbox and Citrix all have established presence and business momentum. Amazon has just entered the market with its WorkDocs offering, and as of yet does not have measurable market share. Given the highly competitive nature of the market at this point, it seems unlikely that Amazon will dedicate significant resources to try to disrupt this market. In our view, there are other - significantly more attractive - fish to fry.

Exploring collaborative applications Amazon’s hosted mail and file-sharing services WorkMail and WorkDocs play in the broadly defined collaborative applications market, which accounted for US$13bn in revenue in 2015, according to IDC. The market encompasses categories that include conferencing apps, email, file synchronization, team collaborative apps and enterprise social networks. This is a large market opportunity, but even more highly fragmented with aggressive competition

Email, file-sharing services appear more

modest ambitions

A drop in a US$13bn bucket so far

WorkMail pricing is competitive but features

trail the competition

WorkDocs file sharing makes sense bundled

with other services




than in the database and data warehouse markets. Microsoft is the leader and Google has made significant inroads, but both companies bundle core functions of WorkMail and WorkDocs into broader suites of productivity and collaborative offerings.

Figure 51

Global collaborative applications market share

Source: IDC

The market is highly fragmented. Microsoft is the dominant competitor in the market, with 35% share in 2015, over four and a half times the next largest competitors Cisco and Google, each with less than 8%. IDC lists over 100 vendors with offerings in the market, and Amazon ranks No.43 with US$31m in 2015 revenue (+86% YoY). The shift towards public cloud offerings mirrors the dynamics of other software markets. IDC forecasts the collaborative applications market to expand at an 8.5% Cagr through 2020.

Hosted email applications - an attractive US$4.3bn market IDC estimates worldwide email applications was a US$4.3bn market in 2015, forecast to expand at a 4.1% Cagr to US$5.1bn in 2019. The story is similar to many other sectors of software: on-premise software is expected to decline at a 2.4% Cagr while public cloud offerings grow nearly 17%, increasing from 27% of the market in 2014 to 47% in 2015.

According to a Gartner study that analyzed the server addresses of 40,000 public companies globally, 13% are using one of the two main cloud email vendors. 8.5% of companies in the sample are using Microsoft's Office365 service, while 4.7% are using Google Apps for Work. The remaining 87% use on-premise, hybrid, hosted or private cloud email managed by smaller vendors. It’s this opportunity that Amazon targets with its WorkMail offering.

Collaborative applications was a US$13bn market

in 2015

Worldwide email applications was a

US$4.3bn market in 2015

Thirteen percent are using one of the two main

cloud email vendors -Microsoft Office365 and

Google Apps for Work




Figure 52

Worldwide email applications revenue snapshot

Source: IDC

The Radicati Group estimates that the cloud business email market was over US$8.7bn in 2015, growing to over US$34.8bn by 2019. This market encompasses offerings like Microsoft Office365, hosted Exchange providers, Google Apps for Work and other basic email services for businesses.

The market is dominated by Microsoft with 55% share in 2014, trailed by IBM’s Lotus platform at 18% and Google a strong insurgent with 16% share following 52% YoY growth. Amazon’s share is small, but the service has been posting the highest growth rates of any vendor in the market.

Figure 53

Worldwide email applications revenue by vendor

(US$m) 2012 2013 2014 2014 share (%)

2013-2014 growth (%)

Microsoft 2,096.7 2,334.5 2,291.7 54.6 (1.8)

IBM 894.2 795.9 737.5 17.6 (7.3)

Google 250 428.8 651.1 15.5 51.9

Micro Focus 88.2 91.7 95.6 2.3 4.2

NEC 87.5 70.5 64.3 1.5 (8.9)

Hitachi 27.3 20.9 18.4 0.4 (11.7)

Amazon.com 6.7 10.0 16.5 0.4 65.3

Other 346.2 326.1 322.3 7.7 (1.2)

Total 3,796.8 4,078.4 4,197.4 100 2.9 Source: IDC

The worldwide email applications market is

expanding at a 4.1% Cagr to US$5.1bn in 2019

The Radicati Group estimates that the cloud

business email market was over US$8.7bn

Microsoft dominates with 55% share in 2014




Amazon’s WorkMail is a cloud-based managed calendaring and email solution that works with existing desktop and mobile clients. The service offers data encryption, regional data controls, support for Microsoft Outlook 2007, 2010, 2013 and 2016 on Microsoft Windows, support for clients that run on OSX including Apple Mail and Outlook. WorkMail supports clients that use the Microsoft Exchange ActiveSync protocol including iPhone, iPad, Kindle Fire, Fire Phone, Android, Windows Phone and BlackBerry 10. Amazon WorkMail makes use of several AWS services including Amazon WorkDocs, the Directory Service, AWS Identity and Access Management (IAM), AWS Key Management Service (KMS) and Amazon Simple Email Service (SES). Organizations can use their existing domain names and integrate with an existing Microsoft Active Directory.

Pricing - inexpensive but not lots of features Amazon WorkMail costs US$4 per user per month and includes 50GB of storage for each user with a 30-day free trial for up to 25 users. Amazon WorkMail and Amazon WorkDocs can also be purchased together for US$6 per user per month, when used in the same region. This option includes 50GB of Amazon WorkMail mailbox storage and 200GB of Amazon WorkDocs storage for each user.

Microsoft’s basic plans include Office365 Business Essentials for US$5/user per month with 50GB mailbox storage, Microsoft Office 365 Business for US$8.25/month with storage extra and Microsoft Office 365 Business Premium for US$12.50 per month with 50GB storage. The basic US$5/month Business Essentials plan includes email, 1TB of file storage, HD conferencing and Office Online. Office 365 Business Premium for US$12.50/month includes email with 50GB mailbox, 1TB file storage, HD video conferencing, fully installed Office on PC/Mac and Office apps on tablets and phones.

Google apps is priced simply at US$5/month per user with 30GB of storage. For US$10/month, users get unlimited storage and Vault.

Rackspace email works with Outlook, smartphones and the Rackspace Webmail app. Hosted Exchange-Rackspace manages Microsoft Exchange and provides enterprise-level syncing of Outlook Web Apps across mobile and tablet devices.

WorkDocs targets the file synchronization and sharing subsegment The US$2.4bn file synchronization and sharing software segment accounts for roughly 18% of the collaborative apps market as of 2015, which IDC expects to grow to 27% by 2020. Amazon has just entered the market with its WorkDocs offering, and as of yet does not have measurable market share.

WorkMail is a cloud-based managed calendaring and email solution that works with existing desktop and

mobile clients

WorkMail costs US$4 per user per month and

includes 50GB of storage for each user

Google apps is priced simply at US$5/month

per user with 30GB of storage




Figure 54

Worldwide file synchronization and sharing revenue snapshot

Source: IDC

Market leadership in the file synchronization market is split between Dropbox, Box and Microsoft, each in the low 20% range. Citrix’s Sharefile is also a strong competitor in the market.

Figure 55

Worldwide business file synchronization and sharing revenue share by vendor, 2014

Source: IDC

Amazon WorkDocs Previously called Zocalo, Amazon WorkDocs is a managed, secure enterprise storage and sharing service with administrative controls and feedback capabilities that improve user productivity. WorkDocs costs US$5 per user per month including 200GB of storage for each user up to 50 users. WorkSpaces (virtual desktop) users receive access to Amazon WorkDocs for no additional charge that can be upgraded from an additional US$2/month to 200GB of storage.

The US$2.4bn file synchronization and

sharing software segment accounts for 18% of the

collaborative apps market

Market leadership in the file synchronization

market is split between Dropbox, Box and Microsoft

WorkDocs costs US$5 per user per month including

200GB of storage for each user up to 50 users




WorkDocs provides users with a central location for both the documents and files they are reviewing as well as those they own and are soliciting feedback. WorkDocs supports common file formats including .doc, .docx, .xls and .xlsx. Users can access store documents from laptops, iOS and Android smartphones and tablets. The WorkDocs Management Console enables administrators to view audit logs to track file and user activity by time, IP address and device, and choose whether to allow users to share files with others outside their organization.

Cheaper than Dropbox and Google, more expensive than Microsoft Major competition for storage and file sync services includes Microsoft’s OneDrive for Business, Google Drive, Box and Dropbox. OneDrive is bundled with Microsoft’s Office 365 with different versions ranging from US$5 to US$12.50 per user per month with 50GB of data. All three include 1TB of storage and sharing per user. In terms of pricing, Microsoft offers better pricing and more extras bundled in its Office365 versions. Microsoft’s OneDrive for Business does have a 20,000-file limit per user. Amazon WorkDocs is priced so that it would cost US$30 per user per month to offer comparable storage to OneDrive, without the extras that Microsoft offers. Google Drive for Work starts at US$5 per user per month with 30GB of storage and 99.9% guaranteed uptime. Unlimited storage is US$10 per user per month. Dropbox for business is more expensive, around US$18 per user per month for unlimited storage.

AWS’s toe in the Desktop-as-a-Service pool Virtual client computing is moving to the cloud, and Desktop as a Service is how on-premise desktop virtualization is being delivered to users over the internet. Amazon Web Services entered this market in 2014 with the introduction of WorkSpaces, its tiered DaaS offering. At the time, there were fears AWS would disrupt market leader Citrix (as well as VMware) through its aggressive pricing. Two years later, concerns have not been borne out as AWS WorkSpaces does not appear to have made much of a dent in the market. Meanwhile, Citrix’s over 2,000 DaaS partners generated over 10% of license sales in 1Q16, climbing at a 38% Cagr. We estimate AWS could capture 5% of the 15%-Cagr DaaS market by 2020, but given technological, organizational and structural advantages, believe Citrix (and Microsoft) will sustain their current dominant market share.

Research firm MarketsandMarkets forecasts the Workspace-as-a-Service market to expand at a 14.6% Cagr from US$4.8bn in 2014 to US$9.4bn in 2019. IDC estimates the virtual client computing DaaS-enabling software market to expand at a 29% Cagr from US$376m in 2014 to US$1.35bn in 2019. Citrix maintains 43% share of the overall virtual client computing market in 2015, ahead of Microsoft at 20% and VMware at 15%. We estimate AWS WorkSpaces at a 46% Cagr from US$75m in 2015 to US$491m in 2020, representing 5% of the total market in 2020.

AWS WorkSpaces are Windows 7 desktops running on a Windows Server 2008 R2 environment. This is limited due to Microsoft’s stringent licensing requirements. Pricing per user is US$25-60/month, depending on capacity (US$4/month less if users have their own Windows license), US$15/month extra for Office and security services. VMware starts at US$40/month, while Citrix Service Providers are at similar levels. Microsoft’s Azure Remote App offers limited functionality starting at US$10-15/month, with a Desktop-as-a-Service offering possible later in 2016.

Major competition for storage and file sync

services includes Microsoft’s OneDrive for Business, Google Drive,

Box and Dropbox

Two years on, AWS not yet a force in DaaS

DaaS an attractive market opportunity

AWS Workspaces a bare-bones offering




Share data are hard to come by: 2ndWatch’s survey of its 100,000 managed AWS instances indicates that it’s the 23rd-most popular AWS service, with roughly 19% of customers using it in one form or another. Johnson & Johnson is the biggest customer with a target of 25,000 users in its deployment. Currently, AWS offers only Windows 7 on Windows Server 2008 R2, while Citrix and VMware support Windows 8.1, Windows 10 and Windows Server 2012 R2.

Microsoft holds most of the cards when it comes to desktop computing for business. AWS can only offer what Microsoft’s licensing model allows. While not blocking AWS’s progress overall, Microsoft can use its leverage to keep dedicated partners like Citrix and its own offerings one step ahead. Citrix indicates that its own CSP business is running around US$65m in annual recurring revenue. Characteristically, AWS has not said much about Workspaces in 2016 since Matt Wood, head of product strategy at AWS, hinted in August 2015 that desktop would be a new area for aggressive pricing. Regardless of the success of its own WorkSpaces offering, AWS remains the go-to Infrastructure-as-a-Service provider for Citrix Service Providers.

WorkSpaces is a basic offering that allows end users to access a virtual Windows 7 desktop powered by Windows Server 2008 R2. There was some concern initially that AWS would be a significant threat to Citrix’s business given AWS’s aggressive pricing strategy around its compute and storage services.

Over the past two years, Citrix has aggressively expanded its third-party ecosystem of service providers where there are now over 2,000 partners selling DaaS offerings powered on its platform. Citrix has sold over 680k monthly DaaS subscriptions as of March 2016 in addition to a large proportion of perpetual licenses that are being delivered as a service.

Some key observations Pricing for AWS WorkSpaces is not appreciably cheaper than comparable

offerings from Citrix Service Providers, VMware Horizon Air DaaS offerings, while Citrix and VMware have deeper technology footprints, enterprise expertise and support capabilities. While there is certainly the possibility that AWS could get more aggressive, the company has dropped the lower tier of pricing only once from US$35 to US$25 a month while leaving higher-tier pricing intact.

AWS has not provided specifics in terms of sizing or growth, and given the lack of buzz in the market, we believe it remains small. The largest reference customer is Johnson & Johnson with a projected deployment of 25,000 clients. We’d estimate the total number of Workspaces users is likely around 100-150,000 worldwide. By comparison there are 50 million active users of Citrix’s Xen family of workspace products (XenDesktop and XenApp).

While WorkSpaces have some initial reference accounts, adoption appears to be tracking more modestly than other services. In terms of popularity among AWS offerings, WorkSpaces ranks No.23 out of 30 in terms of penetration based on 2ndWatch’s survey of customers that run over 100,000 AWS instances, with 19% reporting some usage.

AWS’s software for virtual computing is not yet on industry analysts’ radars. We estimate AWS’s revenue from its own WorkSpaces Desktop at around US$75m per year out of the approximate US$4.8bn DaaS market. Of course, AWS also benefits indirectly from other DaaS providers; the majority of Citrix’s CSP partners deliver DaaS using AWS infrastructure services.

AWS having some successes so far, but not

rocking the boat

WorkSpaces is a basic offering that allows end users to access a virtual

Windows 7 desktop

Pricing for AWS WorkSpaces is not

appreciably cheaper than comparable offerings




The overall desktop computing market continues to be closely tied to Microsoft’s Windows OS, and Microsoft’s licensing schemes limit the flexibility of what service providers can offer as a service. We believe Amazon can be successful in growing a business from a small base, but Microsoft, Citrix and VMware retain advantages in terms of channel reach, technological head start and enterprise sales and support capabilities.

What could the DaaS market look like in 2020? We took a crack at estimating what the market for Desktop as a Service could like in five years. We started with the MarketsandMarkets estimate of a US$4.8bn DaaS market in 2014 expanding at a 14.6% Cagr through 2020. We assumed that the 2015 market share for hosted virtual client computing solutions corresponds roughly to on-premise share, according to IDC.

We assume that Citrix DaaS grows faster than the market, AWS WorkSpaces grows consistent with overall AWS growth, VMware grows slightly slower than the market, Microsoft only grows high-single digits (though its total economic benefit from DaaS will track the entire market because of the Windows client licenses), Huawei grows rapidly off a small base and the share of other vendors declines as AWS, Huawei, Citrix and Microsoft grow faster than the market overall. By our estimation, AWS WorkSpaces could reach nearly US$500m in revenue by 2020, representing 5% of the total. We believe that share gains would come mostly at the expense of smaller vendors in the market.

These estimates refer to total revenue from third parties selling DaaS and also include the underlying software, storage, compute, networking and IT administration costs associated with a bundled offering.

Figure 56

Estimated Desktop-as-a-Service provider revenue by platform

(US$m) 2015 2016 2017 2018 2019 2020 2014-2020 Cagr (%)

Citrix DaaS 2,322 2,832 3,371 3,910 4,457 4,992 17

YoY growth (%) 22 19 16 14 12

Market share (%) 43 45 47 48 47 46

MSFT DaaS 1,112 1,267 1,432 1,618 1,829 2,048 13

YoY growth (%) 14 13 13 13 12

Market share (%) 18 20 20 20 19 19

AWS WorkSpaces 75 120 180 261 365 493 46

YoY growth (%) 60 50 45 40 35

Market share (%) 1 2 3 3 4 5

VMW DaaS revenue 801 913 1,041 1,177 1,329 1,489 13

YoY growth (%) 14 14 13 13 12

Market share (%) 15 15 15 14 14 14

Huawei DaaS 54.5 74 96 124 168 227 33

YoY growth (%) 35 30 30 35 35

Market share (%) 1 1 1 2 2 2

Other 1,086 1,039 1,038 1,112 1,251 1,523 7

YoY growth (%) (4) 0 7 12 22

Market share (%) 20 17 15 14 13 14

Total DaaS market 5,450 6,246 7,157 8,202 9,400 10,772

YoY growth (%) 15 15 15 15 15 Source: CLSA, IDC market share estimates for 2015

AWS WorkSpaces could reach nearly US$500m in

revenue by 2020

We estimate what Desktop as a Service could look like

in five years




The total market size for virtual client computing software is an estimated US$3.3bn, with Citrix leading with 43% share, Microsoft at 20%, VMware at 15%, Huawei at 1% and the rest of the market at 21%.

Figure 57

Worldwide virtual client computing software 2015 market share

Source: IDC

IDC forecasts revenue for DaaS-enabling software at a 29% Cagr through 2019.

Figure 58

Worldwide hosted Desktop-as-a-Service-enabling software revenue

Source: IDC

Citrix43%

Microsoft20%

VmWare15%

Huawei AP1%

Rest of Market21%

0

10

20

30

40

50

60

0

200

400

600

800

1,000

1,200

1,400

1,600

2014 2015 2016 2017 2018 2019

Centralized virtual desktop Virtual user session

Other client virtualization Total growth (RHS)(%)

Total market Cagr 2014-2019=29.1%

(US$m)

Citrix leads the market for virtual client

computing software (including on premise)

The market for software powering DaaS is forecast

at a 29% Cagr




About WorkSpaces AWS Workspaces is a basic DaaS offering likely to be attractive to existing users of AWS. Advantages include ease of management and deployment, security and monthly subscription model. The Workspace Application Manager (with applications that can be purchased and deployed to AWS WorkSpaces) costs another US$5/month.

AWS Workspaces is a Desktop-as-a-Service offering that enables end users to access a full Windows 7 desktop experience hosted on AWS and delivered as a service. WorkSpaces is based on a heavily modified version of the open-source Xen hypervisor, and uses the PCoIP protocol licensed from Teredici (owned by VMware). Pricing tiers range from US$25-60 per month based on hardware requirements, including a Windows client license; US$21-56 if the customer owns their own corporate Windows license. The basic package includes web browsers and file compression. For an additional US$15/month, users can get Office Professional and Trend Micro security services.

Figure 59

AWS Workspaces pricing options

WorkSpaces Bundle Hardware resources Monthly price (US$)

Bring your own license monthly price (US$)

Value 1 vCPU, 2 GB Memory, 10 GB User Storage 25 21

Standard 2 vCPU, 4 GB Memory, 50 GB User Storage 35 31

Performance 2 vCPU, 7.5 GB Memory, 100 GB User Storage 60 56

WorkSpaces Bundle Applications Additional monthly price

Value, Standard, Performance Utilities (Internet Explorer 11, Firefox, 7-Zip) No additional charge

Value Plus, Standard Plus, Performance Plus

Microsoft Office Professional, Trend Micro Worry-Free Business Security Services, Utilities (Internet Explorer 11, Firefox, WinZip)

Additional US$15

Source: Amazon

In terms of usability, Amazon WorkSpaces is still an early stage product. We reference a March 2016 hands-on by author Tim Anderson of The Register, testing it with a variety of clients, including Windows, Mac, iOS, Android and Chromebook. Operation is generally smooth, but it did not work on a Windows 10 desktop at the time. There are some features that make Windows 7 easier to use on a tablet. Printing can be an issue, and there is no assistance with patch management. The solution works well for Chromebooks with an Office365 subscription, and appears to be improving over time.

Citrix XenDesktop offers Windows Server 2008 R2, Windows Server 2012 R2, and Windows XP, 7, 8 client desktops. There are roughly 400,000 organizations with 50 million active users using the Xen Family of products. The product is again delivering new innovations following a four-year effort to re-engineer XenApp and XenDesktop on a common backplane that impacted growth prospects. It runs on the HDX protocol, and service can be obtained through authorized service providers. Prices vary depending on the provider.

WorkSpaces is improving since release,

but still buggy

Citrix XenDesktop still king of the hill

AWS Workspaces is a Desktop-as-a-

Service offering




Four years ago Citrix undertook an extensive effort to re-engineer XenDesktop (desktop virtualization) and XenApp (app virtualization) on a common backplane. This sidelined a lot of innovation, as the company took a lot of features out of the product until work was done. This created significant challenges from 2013-14 but the hard work has been done. The product has the same control plane that runs in the cloud. This effort hurt growth and competitiveness against VMware in particular.

Figure 60

Citrix Service Providers increasing as a proportion of the business

Source: Citrix

For Citrix, CSPs as a standalone piece of the business currently represent about a US$65m annual recurring revenue run rate, rising at a 38% Cagr over the last two years. The CSP model has about a two-year break-even pricing model with a 50% higher total customer value over a four-year life of customers. In terms of cloud business, CSPs are generally net new customers. The cloud should unlock opportunities in the midmarket that have been elusive to date.

Citrix’s Workspace Services segment includes on-premise and cloud, license, subscriptions, support and services. The business had slowed in FY14 and FY15 and in FY16 reflects the impact of some divestitures; however, with the new CEO, Windows 10 enterprise upgrade cycle, refreshed platform and growing interest in high-performance virtual desktops, we are confident that revenue growth will accelerate. AWS remains a partner of Citrix and a key provider of the infrastructure service for CSPs, so growth in DaaS should benefit both vendors.

Figure 61

Citrix Workspace Services (desktop) segment revenue

(US$m) FY14 FY15 FY16CL FY17CL FY18CL

Workspace services (US$) 1,603.4 1,641.6 1,616.1 1,671.4 1,771.7

As a % of revenue 51.0 50.1 48.2 47.0 47.0

YoY growth (%) 3.5 2.4 (1.6) 3.4 6.0

Source: CLSA, Citrix

100

120140

160180

200

220240

260280300

1Q15 2Q15 3Q15 4Q15 1Q160

2

4

6

8

10

12

License revenue (LHS) Term and CSP revenue contribution

(%)(US$m)

CSP subscriptions are over 10% of license sales

The CSP model has 2-year break-even pricing with

50% higher total value of the customer

Tanned, rested and ready for 2016

We are confident that revenue growth

will accelerate




VMware's Horizon Air For comparison, VMware’s Horizon Air offers DaaS at prices starting from US$35 per user per month. Platform support is broader: Horizon Air can provide users with a Windows 7, Windows 8.1 and Windows 10 Windows Server 2008 R2 or Windows Server 2012 R2 virtual desktop.

Figure 62

VMware Horizon Air Options

Desktop Model Standard Advanced Enterprise Enterprise Plus

Price US$35/ desktop/month MSRP

(minimum quantity of 50)

US$50/ desktop/month MSRP






Processor 1 vCPU 2 vCPU 4 vCPU 8 vCPU

Memory 2 GB vRAM 4 GB vRAM 8 GB vRAM 16 GB vRAM

Hard disk 30 GB 60 GB 120 GB 240 GB

Source: CLSA

Desktops can be persistent or non-persistent and utilize the PCoIP protocol Service, which can be provided by VMware or authorized service providers, and pricing may vary depending on which option users choose and the services offered.

Microsoft’s Azure RemoteApp Microsoft’s Azure RemoteApp offering is not a true DaaS offering, rather it’s a service for offering server-hosted applications on the Azure platform. It’s a way to deliver applications to end users regardless of client devices, with all of the processing handled on the server itself. In August 2016, Microsoft announced that it would discontinue Azure RemoteApp in favor of Citrix XenApp on Azure. The product will be available in 1Q17 and supercede Microsoft’s offerings.

Figure 63

Microsoft Azure RemoteApp pricing

Basic Standard Premium Premium plus

Target user Task worker Information worker Information worker Designer/Engineer

Example applications Lightweight LOB applications (eg, data entry,

expense reporting)

Productivity apps (eg, Office, SAP,

Dynamics)

Productivity apps with large data files

(eg, Excel with macros)

Design or heavy-compute apps

(eg, Matlab)

Storage (user) 50 GB 50 GB 50 GB 50 GB

Starting price (user/month) (US$) 10 15 20 25

Hours included in starting price (month) 40 40 40 40

Hourly overage rate (US$) 0.18 0.20 0.30 0.45

Price for unlimited usage, per user/month (US$)

17 23 32 43

Source: Microsoft

Microsoft’s Azure RemoteApp offering is not

a true DaaS offering

VMware’s Horizon Air offers DaaS at prices

starting from US$35 per user per month


Important disclosures Software


Companies mentioned A10 Networks (N-R) Actian (N-R) Adobe Systems (N-R) ADP (N-R) Aetna (N-R) AirBnB (N-R) Akamai (AKAM US - US$52.21 - BUY) Alcatel-Lucent (N-R) Alphabet (GOOGL US - US$805.03 - BUY) Altair (N-R) Altiscale (N-R) Amazon (AMZN US - US$789.74 - OUTPERFORM) Apple (AAPL US - US$113.55 - BUY) Artsy (N-R) AT&T (N-R) Birst (N-R) BlueMetal Inc (N-R) Box-Pak (N-R) Brocade Comms (N-R) Bureau Veritas (N-R) CA Technologies (N-R) Cachefly (N-R) Cato Networks (N-R) CDN77 (N-R) CDNetworks (N-R) CDNify (N-R) Cedexis (N-R) ChinaCache (N-R) Cisco (N-R) Citrix (CTXS US - US$84.51 - OUTPERFORM) Cloudera (N-R) Comcast (N-R) Conviva (N-R) CoolaScale (N-R) CyberArk (N-R) DataStax (N-R) Dell (N-R) Distil Networks (N-R) Domo (N-R) DropBox (N-R) EDS (N-R) Elastic (N-R) EMC (EMC US - US$29.05 - NO REC) F5 Networks (N-R) Fastly (N-R) Fireblade (N-R) Fujitsu (6702 JP - ¥547 - OUTPERFORM) General Electric (N-R) GoDaddy (N-R) Good Data (N-R) Highwinds (N-R) Hortonworks (N-R) HP Enterprise (N-R) Huawei (N-R) IBM (IBM US - US$155.53 - OUTPERFORM) Imperva (N-R) Incapsula (N-R) Informatica (N-R) Intersystems (N-R) Intuit (N-R) Jet Propulsion Laboratory (N-R)

Johnson & Johnson (N-R) Juniper Networks (N-R) Kwicr (N-R) Level 3 (N-R) Limelight (N-R) Madhucon (N-R) MarkLogic (N-R) MathWorks (N-R) MaxCDN (N-R) Microsoft (MSFT US - US$57.76 - OUTPERFORM) Microstrategy (N-R) Mirantis (N-R) NetApp (NTAP US - US$35.70 - UNDERPERFORM) Netflix (N-R) Netsuite (N-R) NeuStar (N-R) New Relic, Inc (N-R) Nice People At Work (N-R) Nuance (N-R) Okta (N-R) OpenText (N-R) Oracle (ORCL US - US$39.51 - UNDERPERFORM) Palantir (N-R) Philips (N-R) Pinterest (N-R) Progress Software (N-R) PTC (PTC US - US$44.11 - OUTPERFORM) QlikTech (N-R) Qualys (N-R) Qubole (N-R) Rackspace (RAX US - US$31.55 - SELL) Radware (N-R) Red Hat (RHT US - US$77.04 - BUY) Rocket Software (N-R) Salesforce.com (CRM US - US$74.18 - BUY) SAP (N-R) Sealed Air Corp (N-R) ServiceNow (N-R) ServiceSource (SREV US - US$4.85 - UNDERPERFORM) Skytap (N-R) Snaplogic (N-R) Snowflake Computing (N-R) Software AG (N-R) Splunk (SPLK US - US$59.21 - BUY) Streamzilla (N-R) Tableau (N-R) Teradata (TDC US - US$30.34 - UNDERPERFORM) Thingworx (N-R) Tibco (N-R) Tinder (N-R) Trend Micro (4704 JP - ¥3,520 - OUTPERFORM) Twilio (N-R) TwinPrime (N-R) Ultimate Software (N-R) VeriSign (N-R) Verizon (N-R) VMware (VMW US - US$73.42 - UNDERPERFORM) Wix (N-R) Workday (N-R) Zscaler (N-R)




Analyst certification The analyst(s) of this report hereby certify that the views expressed in this research report accurately reflect my/our own personal views about the securities and/or the issuers and that no part of my/our compensation was, is, or will be directly or indirectly related to the specific recommendation or views contained in this research report.

Important disclosures The policy of CLSA (which for the purpose of this

disclosure includes subsidiaries of CLSA B.V. and CLSA Americas, LLC ("CLSA Americas")), and CL Securities Taiwan Co., Ltd. (“CLST”) is to only publish research that is impartial, independent, clear, fair, and not misleading. Analysts may not receive compensation from the companies they cover. Regulations or market practice of some jurisdictions/markets prescribe certain disclosures to be made for certain actual, potential or perceived conflicts of interests relating to a research report as below. This research disclosure should be read in conjunction with the research disclaimer as set out at www.clsa.com/disclaimer.html and the applicable regulation of the concerned market where the analyst is stationed and hence subject to. This research disclosure is for your information only and does not constitute any recommendation, representation or warranty. Absence of a discloseable position should not be taken as endorsement on the validity or quality of the research report or recommendation.

To maintain the independence and integrity of

CLSA’s research, our Corporate Finance, Sales Trading and Research business lines are distinct from one another. This means that CLSA’s Research department is not part of and does not report to CLSA Corporate Finance (or “investment banking”) department or CLSA’s Sales and Trading business. Accordingly, neither the Corporate Finance nor the Sales and Trading department supervises or controls the activities of CLSA’s research analysts. CLSA’s research analysts report to the management of the Research department, who in turn report to CLSA’s senior management.

CLSA has put in place a number of internal

controls designed to manage conflicts of interest that may arise as a result of CLSA engaging in Corporate Finance, Sales and Trading and Research activities. Some examples of these controls include: the use of information barriers and other information controls designed to ensure that confidential information is only shared on a “need to know” basis and in compliance with CLSA’s Chinese Wall policies and procedures; measures designed to ensure that interactions that may occur among CLSA’s Research

personnel, Corporate Finance and Sales and Trading personnel, CLSA’s financial product issuers and CLSA’s research analysts do not compromise the integrity and independence of CLSA’s research.

Neither analysts nor their household

members/associates/may have a financial interest in, or be an officer, director or advisory board member of companies covered by the analyst unless disclosed herein. In circumstances where an analyst has a pre-existing holding in any securities under coverage, those holdings are grandfathered and the analyst is prohibited from trading such securities.

Unless specified otherwise, CLSA/CLSA

Americas/CLST did not receive investment banking/non-investment banking income from, and did not manage/co-manage a public offering for, the listed company during the past 12 months, and it does not expect to receive investment banking compensation from the listed company within the coming three months. Unless mentioned otherwise, CLSA/CLSA Americas/CLST does not own a material discloseable position, and does not make a market, in the securities.

As analyst(s) of this report, I/we hereby certify

that the views expressed in this research report accurately reflect my/our own personal views about the securities and/or the issuers and that no part of my/our compensation was, is, or will be directly or indirectly related to the specific recommendation or views contained in this report or to any investment banking relationship with the subject company covered in this report (for the past one year) or otherwise any other relationship with such company which leads to receipt of fees from the company except in ordinary course of business of the company. The analyst/s also state/s and confirm/s that he/she/they has/have not been placed under any undue influence, intervention or pressure by any person/s in compiling this research report. In addition, the analysts included herein attest that they were not in possession of any material, nonpublic information regarding the subject company at the time of publication of the report. Save from the disclosure below (if any), the analyst(s) is/are not aware of any material conflict of interest.




Key to CLSA/CLSA Americas/CLST investment rankings: BUY: Total stock return (including dividends) expected to exceed 20%; O-PF: Total expected return below 20% but exceeding market return; U-PF: Total expected return positive but below market return; SELL: Total return expected to be negative. For relative performance, we benchmark the 12-month total forecast return (including dividends) for the stock against the 12-month forecast return (including dividends) for the market on which the stock trades.

In the case of US stocks, the recommendation is

relative to the expected return for the S&P500 of 10%. Exceptions may be made depending upon prevailing market conditions. We define as “Double Baggers” stocks we expect to yield 100% or more (including dividends) within three years at the time the stocks are introduced to our “Double Bagger” list. "High Conviction" Ideas are not necessarily stocks with the most upside/downside, but those where the Research Head/Strategist believes there is the highest likelihood of positive/negative returns. The list for each market is monitored weekly.

Overall rating distribution for CLSA/CLSA

Americas only /CLST only Universe: Overall rating distribution: Buy / Outperform -

CLSA: 59.66%; CLSA Americas only: 57.77%; CLST only: 76.81%, Underperform / Sell - CLSA: 40.34%; CLSA Americas only: 42.23%; CLST only: 23.19%, Restricted - CLSA: 0.00%; CLSA Americas only: 0.00%; CLST only: 0.00%. Data as of 30 June 2016.

Investment banking clients as a % of rating

category: Buy / Outperform - CLSA: 2.13%; CLSA Americas only: 0.00%; CLST only: 0.00%, Underperform / Sell - CLSA: 1.67%; CLSA Americas only: 0.00%; CLST only: 0.00%, Restricted - CLSA: 0.00%; CLSA Americas only: 0.00%; CLST only: 0.00% . Data for 12-month period ending 30 June 2016.

There are no numbers for Hold/Neutral as

CLSA/CLSA Americas/CLST do not have such investment rankings.

For a history of the recommendations and price

targets for companies mentioned in this report, as well as company specific disclosures, please write to: (a) CLSA Americas, Compliance Department, 1301 Avenue of the Americas, 15th Floor, New York, New York 10019-6022; (b) CLSA, Group Compliance, 18/F, One Pacific Place, 88 Queensway, Hong Kong

and/or; (c) CLST Compliance (27/F, 95, Section 2 Dun Hua South Road, Taipei 10682, Taiwan, telephone (886) 2 2326 8188). © 2016 CLSA Limited, CLSA Americas, and/or CLST.

© 2016 CLSA Limited, CLSA Americas, LLC

(“CLSA Americas”) and/or CL Securities Taiwan Co., Ltd. (“CLST”)

This publication/communication is subject to and

incorporates the terms and conditions of use set out on the www.clsa.com website (www.clsa.com/disclaimer.html.). Neither the publication/communication nor any portion hereof may be reprinted, sold, resold, copied, reproduced, distributed, redistributed, published, republished, displayed, posted or transmitted in any form or media or by any means without the written consent of CLSA group of companies (excluding CLSA Americas, LLC) (“CLSA”), CLSA Americas (a broker-dealer registered with the US Securities and Exchange Commission and an affiliate of CLSA) and/or CLST.

CLSA, CLSA Americas and CLST has/have

produced this publication/communication for private circulation to professional, institutional and/or wholesale clients only. This publication/communication may not be distributed or redistributed to retail investors. The information, opinions and estimates herein are not directed at, or intended for distribution to or use by, any person or entity in any jurisdiction where doing so would be contrary to law or regulation or which would subject CLSA, CLSA Americas and/or CLST to any additional registration or licensing requirement within such jurisdiction. The information and statistical data herein have been obtained from sources we believe to be reliable. Such information has not been independently verified and we make no representation or warranty as to its accuracy, completeness or correctness. Any opinions or estimates herein reflect the judgment of CLSA, CLSA Americas and/or CLST at the date of this publication/communication and are subject to change at any time without notice. Where any part of the information, opinions or estimates contained herein reflects the views and opinions of a sales person or a non-analyst, such views and opinions may not correspond to the published view of CLSA, CLSA Americas and/or CLST. This is not a solicitation or any offer to buy or sell. This publication/communication is for information purposes only and does not constitute any recommendation, representation, warranty or




guarantee of performance. Any price target given in the report may be projected from one or more valuation models and hence any price target may be subject to the inherent risk of the selected model as well as other external risk factors. This is not intended to provide professional, investment or any other type of advice or recommendation and does not take into account the particular investment objectives, financial situation or needs of individual recipients. Before acting on any information in this publication/communication, you should consider whether it is suitable for your particular circumstances and, if appropriate, seek professional advice, including tax advice. CLSA, CLSA Americas and/or CLST do/does not accept any responsibility and cannot be held liable for any person’s use of or reliance on the information and opinions contained herein. To the extent permitted by applicable securities laws and regulations, CLSA, CLSA Americas and/or CLST accept(s) no liability whatsoever for any direct or consequential loss arising from the use of this publication/communication or its contents. Where the publication does not contain ratings, the material should not be construed as research but is offered as factual commentary. It is not intended to, nor should it be used to form an investment opinion about the non-rated companies.

Subject to any applicable laws and regulations at

any given time, CLSA, CLSA Americas, CLST, their respective affiliates or companies or individuals connected with CLSA/CLSA Americas/CLST may have used the information contained herein before publication and may have positions in, may from time to time purchase or sell or have a material interest in any of the securities mentioned or related securities, or may currently or in future have or have had a business or financial relationship with, or may provide or have provided investment banking, capital markets and/or other services to, the entities referred to herein, their advisors and/or any other connected parties. As a result, investors should be aware that CLSA, CLSA Americas, CLST and/or their respective affiliates or companies or such individuals may have one or more conflicts of interest. Regulations or market practice of some jurisdictions/markets prescribe certain disclosures to be made for certain actual, potential or perceived conflicts of interests relating to research reports. Details of the disclosable interest can be found in certain reports as required by the relevant rules and regulation and the full details are available at http://www.clsa.com/member/research_disclosures/. Disclosures therein include the position of CLSA,

CLSA Americas and CLST only. Unless specified otherwise, CLSA did not receive any compensation or other benefits from the subject company covered in this publication/communication.. If investors have any difficulty accessing this website, please contact [email protected] on +852 2600 8111. If you require disclosure information on previous dates, please contact [email protected].

This publication/communication is distributed for

and on behalf of CLSA Limited (for research compiled by non-US and non-Taiwan analyst(s)), CLSA Americas (for research compiled by US analyst(s)) and/or CLST (for research compiled by Taiwan analyst(s)) in Australia by CLSA Australia Pty Ltd; in Hong Kong by CLSA Limited; in India by CLSA India Private Limited (formerly CLSA India Limited), (Address: 8/F, Dalamal House, Nariman Point, Mumbai 400021. Tel No: +91-22-66505050. Fax No: +91-22-22840271; CIN: U67120MH1994PLC083118; SEBI Registration No: INZ000001735; in Indonesia by PT CLSA Sekuritas Indonesia; in Japan by CLSA Securities Japan Co., Ltd; in Korea by CLSA Securities Korea Ltd; in Malaysia by CLSA Securities Malaysia Sdn Bhd; in the Philippines by CLSA Philippines Inc (a member of Philippine Stock Exchange and Securities Investors Protection Fund); in Thailand by CLSA Securities (Thailand) Limited; in Taiwan by CLST and in United Kingdom by CLSA (UK).

India: CLSA India Private Limited, incorporated in

November 1994 provides equity brokerage services (SEBI Registration No: INZ000001735), research services (SEBI Registration No: INH000001113) and merchant banking services (SEBI Registration No.INM000010619) to global institutional investors, pension funds and corporates. CLSA and its associates may have debt holdings in the subject company. Further, CLSA and its associates, in the past 12 months, may have received compensation for non-investment banking securities and/or non-securities related services from the subject company. For further details of “associates” of CLSA India please contact [email protected].

United States of America: Where any section of

the research is compiled by US analyst(s), it is distributed by CLSA Americas. Where any section is compiled by non-US analyst(s), it is distributed into the United States by CLSA solely to persons who qualify as "Major US Institutional Investors" as defined in Rule 15a-6 under the Securities and Exchange Act of 1934 and who deal with CLSA Americas. However, the delivery of this research




report to any person in the United States shall not be deemed a recommendation to effect any transactions in the securities discussed herein or an endorsement of any opinion expressed herein. Any recipient of this research in the United States wishing to effect a transaction in any security mentioned herein should do so by contacting CLSA Americas.

Canada: The delivery of this research report to

any person in Canada shall not be deemed a recommendation to effect any transactions in the securities discussed herein or an endorsement of any opinion expressed herein. Any recipient of this research in Canada wishing to effect a transaction in any security mentioned herein should do so by contacting CLSA Americas.

United Kingdom: In the United Kingdom, this

research is a marketing communication. It has not been prepared in accordance with the legal requirements designed to promote the independence of investment research, and is not subject to any prohibition on dealing ahead of the dissemination of investment research. The research is disseminated in the EU by CLSA (UK), which is authorised and regulated by the Financial Conduct Authority. This document is directed at persons having professional experience in matters relating to investments as defined in Article 19 of the FSMA 2000 (Financial Promotion) Order 2005. Any investment activity to which it relates is only available to such persons. If you do not have professional experience in matters relating to investments you should not rely on this document. Where the research material is compiled by the UK analyst(s), it is produced and disseminated by CLSA (UK). For the purposes of the Financial Conduct Rules this research is prepared and intended as substantive research material.

Singapore: In Singapore, research is issues

and/or distributed by CLSA Singapore Pte Ltd (Company Registration No.: 198703750W), a Capital Markets Services license holder to deal in securities and an exempt financial adviser, a solely to persons who qualify as an institutional investor, accredited investor or expert investor, as defined in s.4A(1) of the Securities and Futures Act. Pursuant to

Paragraphs 33, 34, 35 and 36 of the Financial Advisers (Amendment) Regulations 2005 of the Financial Advisers Act (Cap 110) with regards to an institutional investor, accredited investor expert investor or Overseas Investor, sections 25, 27 and 36 of the Financial Adviser Act (Cap 110) shall not apply to CLSA Singapore Pte Ltd. Please contact CLSA Singapore Pte Ltd (telephone No.: +65 6416 7888) in connection with queries on the report. MCI (P) 013 11 2015

The analysts/contributors to this

publication/communication may be employed by any relevant CLSA entity (including CLSA Americas), CLST or a subsidiary of CITIC Securities Company Limited which is different from the entity that distributes the publication/communication in the respective jurisdictions.

MSCI-sourced information is the exclusive

property of Morgan Stanley Capital International Inc (MSCI). Without prior written permission of MSCI, this information and any other MSCI intellectual property may not be reproduced, redisseminated or used to create any financial products, including any indices. This information is provided on an "as is" basis. The user assumes the entire risk of any use made of this information. MSCI, its affiliates and any third party involved in, or related to, computing or compiling the information hereby expressly disclaim all warranties of originality, accuracy, completeness, merchantability or fitness for a particular purpose with respect to any of this information. Without limiting any of the foregoing, in no event shall MSCI, any of its affiliates or any third party involved in, or related to, computing or compiling the information have any liability for any damages of any kind. MSCI, Morgan Stanley Capital International and the MSCI indexes are service marks of MSCI and its affiliates. The Global Industry Classification Standard (GICS) was developed by and is the exclusive property of MSCI and Standard & Poor's. GICS is a service mark of MSCI and S&P and has been licensed for use by CLSA.

EVA® is a registered trademark of Stern, Stewart

& Co. "CL" in charts and tables stands for CLSA/CLSA Americas and “CT” stands for CLST estimates unless otherwise noted in the source.


Research & sales offices www.clsa.com

Australia - Melbourne CLSA Australia Pty Ltd Level 30 35 Collins Street Melbourne VIC 3000 Tel: +61 2 8571 4200 Fax: +61 2 9221 1188

India CLSA India Private Limited (formerly CLSA India Limited) 8/F, Dalamal House Nariman Point Mumbai 400021 Tel: +91 22 6650 5050 Fax: +91 22 2284 0271

Philippines CLSA Philippines, Inc 19/F, Tower 2 The Enterprise Center 6766 Ayala corner Paseo de Roxas Makati City Tel: +63 2 860 4000 Fax: +63 2 860 4051

USA - Boston CLSA Americas, LLC 99 Summer Street, Suite 220 Boston, MA 02110 Tel: +1 617 295 0100 Fax: +1 617 295 0140

Australia - Sydney CLSA Australia Pty Ltd CLSA House Level 15 20 Hunter Street Sydney NSW 2000 Tel: +61 2 8571 4200 Fax: +61 2 9221 1188

Indonesia PT CLSA Sekuritas Indonesia WISMA GKBI Suite 901 Jl Jendral Sudirman No.28 Jakarta 10210 Tel: +62 21 2554 8888 Fax: +62 21 574 6920

Singapore CLSA Singapore Pte Ltd 80 Raffles Place, No.18-01 UOB Plaza 1 Singapore 048624 Tel: +65 6416 7888 Fax: +65 6533 8922

USA - Chicago CLSA Americas, LLC One South Dearborn Street Suite 2100 Chicago IL 60603 Tel: +1 312 220 7346 Fax: +1 312 212 4401

China - Beijing CLSA Limited - Beijing Rep Office Unit 10-12, Level 25 China World Trade Center Tower 2 1 Jian Guo Men Wai Ave Beijing 100004 Tel: +86 10 5965 2188 Fax: +86 10 6505 2209

Japan CLSA Securities Japan Co., Ltd. 16/F, Shiodome Sumitomo Building 1-9-2, Higashi-Shimbashi Minato-ku, Tokyo 105-0021 Tel: +81 3 4578 8000 Fax: +81 3 4578 8080

Taiwan CL Securities Taiwan Company Limited* 27/F, 95, Section 2 Dun Hua South Road Taipei 10682 Tel: +886 2 2326 8188 Fax: +886 2 2326 8166

USA - Dallas CLSA Americas, LLC 100 Crescent Court Suite 700 Dallas TX 75201 Tel: +1 214 530 9289

China - Shanghai CLSA Limited - Shanghai Rep Office Room 910, 9/F 100 Century Avenue Pudong New Area Shanghai 200120 Tel: +86 21 2020 5888 Fax: +86 21 2020 5777

Korea CLSA Securities Korea Ltd 30/F, One IFC 10 Gukjegeumyung-ro Yeongdeungpo-gu Seoul, 150-712 Tel: +82 2 397 8400 Fax: +82 2 771 8583

Thailand CLSA Securities (Thailand) Ltd 16/F, M. Thai Tower All Seasons Place 87 Wireless Road, Pathumwan, Bangkok 10330 Tel: +66 2 257 4600 Fax: +66 2 253 0532

USA - New York CLSA Americas, LLC 1301 Avenue of The Americas 15th Floor, New York 10019 Tel: +1 212 408 5888 Fax: +1 212 261 2502

China - Shenzhen CLSA Limited - Shenzhen Rep Office Room 3111, Shun Hing Square Di Wang Commercial Centre 5002 Shennan Road East Shenzhen 518008 Tel: +86 755 8246 1755 Fax: +86 755 8246 1754

Malaysia CLSA Securities Malaysia Sdn Bhd Suite 20-01, Level 20 Menara Dion 27 Jalan Sultan Ismail 50250 Kuala Lumpur Tel: +60 3 2056 7888 Fax: +60 3 2056 7988

United Kingdom CLSA (UK) 12/F, Moor House 120 London Wall London EC2Y 5ET Tel: +44 20 7614 7000 Fax: +44 20 7614 7070

USA - San Francisco CLSA Americas, LLC 50 California Street, Suite 850 San Francisco, CA 94111 Tel: +1 415 544 6100 Fax: +1 415 434 6140

Hong Kong CLSA Limited 18/F, One Pacific Place 88 Queensway Hong Kong Tel: +852 2600 8888 Fax: +852 2868 0189

* CLST is an exclusive Taiwan research provider to CLSA

CLSA Sales Trading Team

Australia +61 2 8571 4201 China (Shanghai) +86 21 2020 5810 Hong Kong +852 2600 7003 India +91 22 6622 5000 Indonesia +62 21 573 9460 Japan +81 3 4580 5169 Korea +82 2 397 8512

Malaysia +60 3 2056 7852 Philippines +63 2 860 4030 Singapore +65 6416 7878 Taiwan* +886 2 2326 8124 Thailand +66 2 257 4611 UK +44 207 614 7260 US +1 212 408 5800

Research subscriptions To change your report distribution requirements, please contact your CLSA sales representative or email us at [email protected]. You can also fine-tune your Research Alert email preferences at https://www.clsa.com/member/tools/email_alert/.

© 2016 CLSA Limited (“CLSA”), CLSA Americas, LLC (“CLSA Americas”) and/or CL Securities Taiwan Co., Ltd. (“CLST”) Key to CLSA/CLSA Americas/CLST investment rankings: BUY: Total stock return (including dividends) expected to exceed 20%; O-PF: Total expected return below 20% but exceeding market return; U-PF: Total expected return positive but below market return; SELL: Total expected return to be negative. For relative performance, we benchmark the 12-month total forecast return (including dividends) for the stock against the 12-month forecast return (including dividends) for the market on which the stock trades. For example, in the case of US stock, the recommendation is relative to the expected return for S&P of 10%. Exceptions may be made depending upon prevailing market conditions. • We define as “Double Baggers” stocks we expect to yield 100% or more (including dividends) within three years at the time the stocks are introduced to our “Double Bagger” list. "High Conviction" Ideas are not necessarily stocks with the most upside/downside but those where the Research Head/Strategist believes there is the highest likelihood of positive/negative returns. The list for each market is monitored weekly. 12/09/2016

CLSA is certified ISO14001:2004

CLEANGREEN& TM

At CLSA we supportsustainable development.

We print on paper sourced fromenvironmentally conservativefactories that only use fibres

from plantation forests.Please recycle.

mailto:[email protected]

https://www.clsa.com/member/tools/email_alert/


Documents

Amazon’s software play - Tech2025 · NTAP, IBM, ORCL, HPE, Dell/ EMC, Veritas GOOG, MSFT, IBM, AKAM High AWS dominates storage as a service with scale, pricing and breadth Networking