Clouducopia - Into The Era of Abundance

Blue BooksExperts’ views for expert investors

The group of companies that comprise CLSA are affiliates of Credit Agricole Securities (USA) Inc. For important disclosure information please refer to page 40.

Global

Technology

January 2013

Ed Maguire Managing Director Software Research [email protected] (1) 212 549 8200

Guest authors

John Zysman Professor, University of California, Berkeley

Jonathan Murray Chief Technology Officer, Warner Music Group

Kenji Kushida Takahashi Research Associate, Stanford University

www.clsau.com

ClouducopiaInto the era of abundance zero-one

Blue Books

2 www.clsau.com January 2013

Produced byProduced byProduced by

Straight to the source with CLSA When industry innovations change as quickly as they are created, yourability to respond could mean the difference between success and failure. Inthis volatile environment, why rely entirely on broker research when you cantap into unfiltered, unbiased primary research?

CLSA U® is a value-added executive education programme created toallow you to gain firsthand information and draw your ownconclusions and make better informed investment decisions.

CLSA U® offers tailored courses on a broad range of macro themes with aspecial focus on technology and telecoms. The format ensures you learn aswe do and obtain firsthand information about prospects and trends inindustries and sectors that underline the companies in your portfolio.

You will interact and learn from the trailblazers at the centre oftoday’s fastest moving industries - experts, engineers and scientistswho design, implement and shape the new technologies today,which impact the market tomorrow.

CLSA U® is not a one-off event. It is an ongoing education programmerestricted to CLSA’s top clients. The syllabus will constantly evolve to meetyour needs and help you debunk the latest technologies, investment stylesand industry trends that affect the markets and sectors you invest in.

For more details, please email [email protected] or log on to www.clsau.com

John Zysman John has been a faculty member of the University of California, Berkeley,teaching and researching high-technology issues, European politics and politicaleconomy, since 1974. He is also co-director of the Berkeley Roundtable on theInternational Economy (BRIE) since its establishment in 1982. He has writtenextensively on European and Japanese policy and corporate strategy. His path-breaking book, Manufacturing Matters: The Myth of the Post-IndustrialEconomy, remains an enduring driver of economic development policy andstrategy for governments and companies. He has hosted international eventsand been a featured panelist at numerous industry and policy conferences.

Jonathan Murray Jonathan is the CTO and EVP for Warner Music Group, where he is responsiblefor global technology strategy, end-to-end IT service delivery and the designand delivery of the new cloud-based Digital Services Platform. CloudAverecently recognized him as one of three cloud-computing visionaries for 2012. Jonathan joined WMG from Innovia Ventures, where he was a founding partner.He spent 16 years in various executive roles with Microsoft, most recently asvice president of the Public Sector Technology Office. Jonathan is a publishedresearch fellow with BRIE and blogs occasionally at http://adamalthus.com.

Kenji Kushida Kenji is the Takahashi Research Associate in Japanese Studies at the Walter HShorenstein Asia-Pacific Research Center at Stanford University. He is anaffiliated researcher at BRIE, with a PhD in political science from the Universityof California, Berkeley. He specializes in the political economic issuessurrounding information technology and has published on topics such as cloudcomputing and the transformation of services with IT tools, along with JonathanMurray and John Zysman. Kenji is involved with startups and investors, workingto bridge the gap between IT startups in Silicon Valley and Japan.

CLSA U® logo, CLSA U® (word mark) and CLSA University are registered trademarks of CLSA in the USA and elsewhere.

Blue Books

January 2013 www.clsau.com 3

Contents

Executive summary ............................................................................ 5

Abundant ramifications ...................................................................... 6

Bracing for disruption ...................................................................... 11

Driving disruptive impact ................................................................. 17

Battle lines square off ...................................................................... 24

In-depth insights

Blue Books



Foreword In trying to understand the implications that the information-technology revolution has on the broader economy, I happened upon a brilliant report: Services with Everything: The ICT-Enabled Digital Transformation of Services by John Zysman, Stuart Feldman, Jonathan Murray, Niels Christian Nielsen and Kenji Kushida. The report identified how technology is changing the nature of services from an economic sinkhole to a productivity driver.

I was deeply impressed by the team’s cogent insights into the profound economic implications of the Information and Communications Technology (ICT) revolution. I reached out to Professor John Zysman to gauge his interest in a CLSA U Blue Book on cloud computing, and he pulled in his colleagues Kenji Kushida and Jonathan Murray. Clouducopia is the result of our collaborations.

We explore why cloud computing is distinct from prior technology waves around the mainframe and PC eras. Each transition has successively decoupled and disintegrated the technology “stack”. Cloud computing creates the experience of unlimited computing power on demand for users, accelerating capacity for innovation and scaling new ideas. For providers, the dynamics around offering cloud services are far more nuanced than utilities. We look at the looming threats that “arms dealers” Intel, Cisco and Oracle face because of the cloud.

Adoption of cloud computing is “wiring the world”. This transition offers parallels with the rise of electrification in the late 18th Century. The introduction of electric power transformed industries and economies, paving the way for massive growth and improvement in living standards in the developed world. We are at the cusp of a globally connected world, and the changes afoot are profound for users and participants - and especially for the ICT “49-ers” staking claims to virtual landscapes. I hope you find this report as illuminating as I have.

Ed Maguire Managing Director, Software Research

Executive summary Blue Books


Clouducopia Enterprise computing is undergoing a fundamental generational change as significant as the mainframe to PC in the 1980s and 1990s, and it is poised to be just as disruptive. Cloud computing is ushering in an era of abundance that will force major shifts in the strategies of firms that have built their success on enterprise-computing models of scarcity. While many of the underlying concepts and technologies are not new, cloud computing is a radical new way of delivering computing resources. We discuss here the context, characteristics and profound ramifications for users and providers.

Cloud computing is an economic story. Early days of computing were characterized by scarcity, which in turn constrained software capabilities. Scarcity enabled tightly coupled systems to flourish. The rise of the PC loosened the binding between software and hardware, changed the economic model and unleashed new industries fuelled by Moore’s Law and consumer demand. Cloud computing delivers computing resources in abundance, further loosening software constraints with profound implications for incumbents.

It is uniquely new, simultaneously serving as an innovation ecosystem, production environment and a marketplace. Ready accessibility to low-cost, nearly unlimited computing resources provides a baseline for agile innovation. Cloud services have crossed the threshold of early adoption to support mainstream production environments. Global reach enables businesses to operate and scale globally, giving rise to collaborative models.

As the PC unbundled software and hardware, leading vendors pursued different paths to escape commoditization, embedding business logic and offering high-end, customized “solutions”. Cloud computing disintegrates the vertical stacks, further commoditizing the physical layers. A more radical form of disruption results from decoupling business logic from the database and infrastructure software, further undermining vendor “lock-in”.

There are four types of providers, each with different competitive dynamics: cloud-service providers, connectors (network providers), device providers and “arms suppliers”. For service providers, global scale is an advantage, but regulations offer entry opportunities for localized providers. The starting points of Amazon, Microsoft, Google, Salesforce.com and Apple each define their strategies. Network providers must seek higher value to avoid commoditization. “Arms suppliers” Intel, Cisco and Oracle must adjust strategies to address the cloud’s existential threats to their core businesses.

Cloud computing can be an innovation ecosystem, production environment and a marketplace

Innovation Production Marketplace

Lowers the bar for new entrants Scale enables lower operating costs Provides global reach for applications

Facilitates experimentation Highly efficient management and upgrades Ability to scale user base rapidly

Lower capital outlays for startups Provides the building blocks for new systems Enables multisourcing (Mechanical Turk)

Enterprises use cloud to handle demand bursts

Supports new businesses (Dropbox, Netflix streaming, Foursquare)

Facilitates crowdsourcing (Kickstarter)

Improves allocation efficiency of internal IT resources

Source: John Zysman, Jonathan Murray, Kenji Kushida

Abundant ramifications

Bracing for disruption

Driving disruptive impact

Battle lines square off

Ushering in an era of abundance

Section 1: Abundant ramifications Blue Books


Abundant ramifications Cloud computing is an economic story. Early days of computing were characterized by scarcity, which in turn constrained software capabilities. Scarcity enabled tightly coupled systems to flourish. The rise of the PC in the 1980s and 1990s loosened the binding between software and hardware, changed the economic model and unleashed new industries fuelled by Moore’s Law 1 and consumer demand. In that disruption, existing leaders such as Digital Equipment Corporation (DEC) and Wang Laboratories were driven out of the business. IBM survived, but only after a major turnaround. The transformation now under way will force major shifts in the strategies of firms that have built their success on the current enterprise-computing model. Cloud computing is a radical new way of delivering computing resources, which we will explain.

What do we mean by cloud computing? The label “cloud computing” is undoubtedly overused, particularly by the marketing departments of those firms most at risk from the disruption. Cloud computing is not simply anything on the web, nor does it automatically imply social networks, such as Twitter or Facebook. Most importantly, it is not a new word for old models of IT outsourcing or application hosting. Such definitions attempt to render the term too broadly or to use the term “cloud computing” to re-label the old as the new, and are therefore not useful. What we need is a specific definition that captures the heart of what cloud computing really delivers and differentiates it from previous generations of computing.

In simple terms: Cloud computing delivers computing services - data storage, computation and networking - to users at the time, to the location and in the quantity they wish to consume, with costs based only on the amount of resource used.

In laymen’s terms: Cloud computing delivers the computing services to support business or personal needs without the user having to know how the underlying physical devices and software are configured or managed.

Cloud computing and the economics of abundance2 In the early days of the computing industry, hardware resources were extremely scarce. Processors had limited computational capacity. The limited size of computer program memory and disk storage put severe constraints on the size and complexity of computer applications. The earliest network connections could only transmit data slowly and at very high cost.

The cost and limited capacity of these foundational components of computing infrastructure - computation, memory, storage and network bandwidth - placed severe limits on the complexity of the software operating systems and applications which ran on top. Optimizing for these limitations meant that software was always written - or targeted - for the specific underlying hardware on which it would run. IBM produced highly optimized operating systems dedicated to specific mainframe hardware memory and storage architectures. DEC did the same for its mini-computers, and even later-generation companies such as Sun Microsystems followed the same model to extract every last ounce of performance from their workstation and service products.

1 Moore’s Law is the observation by Intel co-founder Gordon Moore that the number of transistors (in effect, processing power) in integrated circuits doubles every two years. 2 We owe thanks to Jonathan Murray for this conception.

The term “cloud computing” is overused,

particularly by firms most at risk from disruption

Cost and capacity constraints have limited capabilities of software

Forcing major shifts in the strategies of firms



Figure 1

The evolution of computing waves


The early success of Sun Microsystems in markets for engineering workstations and high-performance servers for the financial-services industry show just how successful this “tight coupling” of software and hardware was as an economic model in the computer industry. While the demands of software developers and their algorithms outstripped the capacity of hardware resources, this tightly coupled model would hold sway.

The economics of scarcity laid the foundation for the success of most pioneering companies of the enterprise-computing industry. Companies were willing to pay a premium for software and hardware solutions that could optimize scarce resources and keep operating costs to a minimum.

Figure 2 Figure 3

Digital Equipment Corporation minicomputer (1965) Sun Microsystems workstation computer (Sun-1) (1982)

Source: Smithsonian Institution Source: Sun Microsystems

Computer industry arises around a tightly coupled

model

Economics of scarcity helped computing pioneers succeed

As the client gets thinner, the number of users

increases



Cisco, Oracle and EMC are but three companies whose fundamental value proposition was based on the economics of scarcity. Each of these companies created value by being best in class in helping customers optimize and manage network, database scale and disk storage, respectively. They could each charge a premium price for their products and services because the costs of the underlying infrastructure resources were so high that marginal improvements in optimization led to significantly positive outcomes in cost control.

The introduction of the IBM PC in the early 80s laid the foundation for a transformation in this economic model. IBM broke with its traditional model when it outsourced the supply of the computer processor and operating system for this personal computer. Little did IBM know at the time that this single decision would transform an industry.

The decision to source processors from Intel and an operating system from Microsoft and, critically, to not bind both these suppliers to exclusive supply agreements created the opportunity to build a PC industry around IBM’s generic blueprint. The market for PCs grew and other players such as Compaq created competitive alternatives to IBM.

Three factors were to transform the previous era of computing resource scarcity into one of overwhelming abundance. Decomposition of previous systems into components unleashed innovation and competition at every layer. Exponential growth in consumer demand led to vast economies of scale in the production of microprocessors, disk storage and memory chips. At the same time, improvements in hardware fabrication technology (following Gordon Moore’s Law) have seen a doubling of performance in computational capacity, program memory and disk storage capacity every 12-18 months continuously for over 40 years.

Figure 4

Moore’s Law

Source: Wikipedia

Rise of the PC loosens the bindings of the industry

Decomposition, consumer demand and Moore’s Law

convert scarcity to abundance

Improvements in processor capacity

continue at an exponential rate

Cisco, Oracle and EMC are companies whose value

proposition was based on the economics of scarcity



It is hard to understate the impact of this transition from scarcity to abundance. There is a close analogy to the introduction of the moveable-type printing press. IBM’s invention of the open PC architecture and selection of Intel processors democratized computing in a similar way that Guttenberg’s printing press democratized printing, with arguably an even greater impact on our societies and economies. We can speculate whether this democratization of information access has as great an impact on our societies and economies.

An environment where underlying hardware resources are, to all intents, a limitless commodity changes the calculus of software development. The emergence of the Linux operating systems is a good exemplar of these changes.

In the era of scarcity, the most direct analog to Linux was the Unix operating system (OS) invented at Bell Labs. Unix was designed to be a general-purpose operating system which could run on many different types of underlying hardware. However, when hardware resources are scarce, the operating system needs to be tuned and optimized to be as efficient as possible on each type of computer. Tuning Unix for each different system turned it from being a single, general-purpose operating system to being a collection of related but quite different operating systems. This in turn meant that an application written to run on one variant of Unix was not guaranteed to run on any other variant. There was no economy of scale.

Linux, on the other hand, is an operating system for the era of abundance. So much computational capacity is now available on personal computer class machines that some of this capacity can be “wasted” by layers of software designed to insulate application developers and users from underlying differences in hardware architecture from machine to machine.

This ability to “waste” computing cycles means that Linux truly has become a general-purpose operating system capable of running on nearly any hardware architecture available. However, we are still talking about an environment where there is a one-to-one relationship between a computer - hardware - and the operating system. Applications written for Linux will only run on Linux and not on Microsoft Windows as an example. But massive abundance is now enabling us to solve even this problem.

The Mac, on which this report is being written, can run not only its native OSX operating system developed by Apple, but also Microsoft Windows and any variation of Linux. It is also capable of running Sun’s (now Oracle’s) Open Solaris operating system. How is this possible?

There is now such an abundance of computational capacity that other, multiple different hardware architectures can be emulated in software. “Virtualization” software allows us to run a copy of Microsoft Windows or Linux on a Mac computer in parallel with the native OSX operating system. The Windows operating system or Linux is running in a “virtual” container, which emulates the hardware resources of a generic IBM PC. The virtualization software is smart enough to allow the sharing of the Mac’s underlying physical hardware by multiple different operating systems and their applications. It is an advancement in the ability to “virtualize” underlying hardware resources that has laid the foundation for cloud computing.

Abundance changes the calculus of software

development

Unix operating system -ecosystems of scarcity

Linux is the OS of abundance

Virtualization breaks couplings even further

It is hard to understate the impact of this

transition from scarcity to abundance



Cloud as a resource-management model Cloud computing, in its simplest sense, is a computing-resource-management model. It is a method for pooling and sharing hardware-infrastructure resources on a massive scale. Finite hardware resources are shared between competing demands, giving each user the illusion of exclusive control over the underlying environment and without the user needing to know anything about how the physical resources are configured. Cloud computing is an operating and economic model for the era of abundance.

The implications for companies that built their business models on the economics of scarcity are profound. If the cloud-computing-management model makes computation, disk storage and program memory a limitless commodity resource, then users are unlikely to pay a premium for solutions designed to optimize scarce resources.

Oracle’s success in database technology came through its ability to provide high levels of performance and scaling in resource-constrained environments. When those environments ran out of capacity, Oracle would be happy to sell you very expensive dedicated hardware solutions to overcome the problem. In a cloud-computing environment, the same problem can be overcome by scaling open source (free) database technologies over vast commodity hardware infrastructures.

Profound implications for companies built

on scarcity

Cloud paves the way for open-source disruption

Cloud computing is a computing-resource-management model

Section 2: Bracing for disruption Blue Books


Bracing for disruption Cloud computing is uniquely new by simultaneously being an innovation ecosystem, a production environment and a marketplace, despite many of the underlying technologies and concepts, such as virtualization and applications residing on remote servers, not being new. We focus on the impacts and disruptions generated by the advent of cloud computing, with emphasis on who are the winners and losers. The significance of this decoupling of applications, platforms and infrastructure that cloud computing enables cannot be exaggerated. The decoupling affects how computing is used and what kinds of services and products will emerge.

Two groups of actors: Providers and users We divide our analysis into two groups of actors: those involved in the provision of cloud services (including those that provide the equipment, connection and tools), and users, which include large and small businesses and consumers.

The impact of cloud computing on specific regions and sectors may be powerfully influenced by government policy. Issues such as privacy, security and reliability will matter a great deal. Many of these issues are not new, such as privacy or security. However, cloud computing changes their manifestation, opening old debates and adding twists to traditional policy positions.

What is cloud computing? First, we need to go further in depth to elaborate on the definition and main characteristics of cloud computing, including some examples.

Cloud computing delivers computing services - data storage, computation and networking - to users at the time, to the location and in the quantity they wish to consume, with costs based only on the resources used.

Let us unpack this definition by adding characteristics and examples.

Users procure the “amount of computing” they want without investing in their own infrastructure. Only an internet connection is required. Cloud services provide the illusion of infinite resources on demand available to users, regardless of their size and number.

Physical infrastructure is decoupled from applications and platforms, which allocate computing, memory and storage resources without reference to underlying physical infrastructures. This is known as virtualization. Note also that the physical location is decoupled between the physical location of users and the cloud datacenters.

Cloud services transform computing from a capital expense to an operating expense. This changes the role of IT expenditures within the firm.

Providers can dynamically add, remove or modify hardware resources without reconfiguring the services that depend on them. This is a major difference from traditional datacenter outsourcing.

Cloud computing changes the location of data processing. Processing moves from the “edge” of the network, in PCs and private datacenters, towards the center of the network, in shared cloud datacenters.

Only a few firms are able to offer truly global-scale cloud infrastructure (eg, Amazon, Google, Microsoft), with each firm requiring numerous datacenters costing more than US$500 million each, worldwide.

Decoupling applications, platforms and

infrastructure has profound impact

Cloud services provide the illusion of infinite

resources on demand

Cloud services transform computing from a capital expense to an operating

expense



There are a few things that cloud computing is not.

Cloud computing is not simply all datacenter outsourcing, and a large enterprise with a single datacenter is not a cloud-service provider. The real power of cloud computing is in the dynamic allocation of resources and the “illusion” of infinite scale. Social networking services, such as Facebook, Twitter or LinkedIn, are not necessarily cloud services in of themselves. They can deploy cloud computing to deliver their services, and they can provide cloud-based platforms for third-party applications. However, not all do.

Cloud computing does not automatically imply dumb terminals or “thin” clients with little power at the user’s end. Many cloud services depend on powerful client-processing capabilities, and networks are not always reliably fast enough. Processing power and storage capacity on user devices (smartphones, PCs) continues to improve in line with Moore’s Law.

We highlight a few key examples:

Firms that use Amazon’s virtual servers, such as Dropbox or Netflix, don’t have to worry that they will use up Amazon’s storage or processing capability. It is Amazon’s responsibility to give the illusion of infinite resources, not the users’.

You really don’t know, and in most cases don’t care about, where your data are stored when using Google Docs or Microsoft Office 365. (We will get into the exceptions later.) You also don’t care what kind of servers they are running - you just want the functionality.

Google, Microsoft and Amazon build new, massive datacenters with seemingly continuous frequency, but this doesn’t affect the applications and services that run on their infrastructure. You don’t have to fundamentally reconfigure your Google database when it adds new datacenters or migrates to new versions of its servers’ operating systems.

Dropbox, the popular file-storage and synchronization service, uses Amazon’s virtual storage to deliver its functionality. It did not have to build massive datacenters on its own to start and rapidly expand its services.

Figure 5

What cloud is and what it is not

Is Not

Users procure IT capacity without investing in their own infrastructure

Not simple IT outsourcing

Physical infrastructure not tied to specific applications and platforms

Platforms and applications not closely coupledto underlying hardware

Opex for IT consumers Not capex for IT consumers

Computing moves from the edge to centralized datacenters

Endpoints are not necessarily dumb terminals, some processing remains


Dynamic utility For users, cloud computing is a dynamic utility. As with a traditional utility, cloud-computing resources are always available, paid for according to the amount consumed and can be consumed in any quantity. (More precisely, there are contractual levels of availability and reliability.) Services are delivered through internet connections, and the provider does not care about the device used to consume the service. Users do not care about how

Users do not care about the backend as long as

service quality and price are acceptable

Cloud computing is not simply datacenter

outsourcing

Cloud computing does not automatically imply

dumb terminals

You really don’t know, and in most cases don’t care about where your

data are stored

Cloud computing provides user flexibility as it

decouples software from the physical layer



providers technically configure or operate the service on the backend as long as quality and price are acceptable, and users are free to use the resources as they see fit.

Figure 6

Representative cloud Infrastructure-as-a-Service offerings and pricing Provider Categories Description Pricing Google On-demand front-end instances 28 free hours per app US$0.08/hour (US$60/month) App Engine High-replication datastore (HRD) 1GB free US$0.24/GB/month (US$0.00032 /GB/hr) Each API has free quota and price per usage Blobstore API 5GB free; US$0.13/GB/month Datastore API 50k free; US$0.10/100k write operations Amazon Standard on-demand instances (EC2) Linux/UNIX US$0.065/hour (US$50 /month) AWS Windows US$0.115/hour (US$80/month) Simple storage service (S3) First 1TB US$0.095/GB + US$0.10 per mn requests Relational database service (Oracle) Small DB instance US$0.14/hour (US$105/month) Elastic block store (EBS) DB storage US$0.10/GB + US$0.10 per mn requests Microsoft Computing nodes Shared CPU, 768MB RAM US$0.02/hour (US$15/month) Azure 1 Core, 1.75GB RAM US$0.12/hour (US$90/month) Storage US$0.105/GB + US$0.01 per 10k operations SQL Azure database 100MB - US$0.0067/hour (US$5/month) 1GB - US$0.0133/hour (US$10/month) Source: Amazon, Google, Microsoft

Cloud providers, like utility providers, are large companies operating at significant scale, serving small users as well as giant corporations. Aggregate demand can then be amortized over this highly scalable infrastructure and sold back to the user at a much lower - per unit resource - cost than users could provide themselves.

Cloud computing is commonly delineated into three layers: Infrastructure as a Service (IaaS), Platform as a Service (PaaS) and Software as a Service (SaaS). We include brief definitions:

IaaS can be thought of as a “management” model (ie, how computing, storage and memory resources are allocated to applications based on demand from users over the internet).

PaaS is a “development” model, which defines how software developers design and build applications that access and make use of the computing resources managed by the underlying IaaS layer. PaaS typically offers software developers common services such as user authentication and database access designed to enable written applications to take advantage of the scalability and resilience of the underlying infrastructure.

SaaS is a “delivery” model that defines how software written by developers using the PaaS layer is made available to users over the internet. Users typically only require an internet connection and web browser to access SaaS-based applications and SaaS is often associated with a subscription- based economic model.

Cloud computing to become critical social infrastructure Cloud computing is poised to become part of societies’ critical infrastructure, as an increasingly dominant means through which the world’s computational demands are met. It will approach the level of economic critical dependency as electricity, gas, water and telephony.

These utility-like characteristics create incentives for national governments to search for regulatory frameworks that approach cloud services as critical national infrastructure. This is particularly important for non-US governments, since the major global services providers are American and hence subject also to American regulations and rules.

Cloud providers, like utility providers, are large

companies operating at significant scale

Cloud computing is poised to become part of societies’ critical

infrastructure

Cloud computing is commonly delineated

into three layers



For providers, cloud services are competitive propositions The competitiveness of the cloud-computing service provision critically depends on the providers’ ability to build out capacity at a scale far greater than any individual user or firm could afford. For providers, cloud services are not utilities; they are competitive propositions that differ from utilities in several important ways, and providers would certainly resist being regulated as utilities.

First, cloud providers do not want to offer commodities - goods offering little value-add - that are interchangeable with others, and which compete primarily on the basis of price. Cloud providers are competing on value-based differentiation on attributes such as service level and functionality.

Second, cloud providers do not enjoy inherent, geographic lock-in of users. Public-utility providers are granted local monopolies, but cloud services are not geographically bound. Therefore, cloud providers face pressure to create their own service level lock-in mechanisms. These may include proprietary software components such as a Platform-as-a-Service layer, or specific characteristics tailored for vertical industry or regulatory requirements.

Third, the actual data bits delivered in cloud services are not interchangeable in the manner of electrons, or molecules in traditional utilities. The bits combine into quite different uses, some of which are mission critical and some of which are more casual. Users care a great deal about the whereabouts of the bits carrying sensitive personal or mission-critical corporate data, but far less about the location of the constituent bits of a photo or video.

Datacenters are not inherently tied to specific locations close to the markets they serve. Other factors, such as real estate, electricity and access to network infrastructure, matter greatly in siting decisions.

Arguably, the biggest difference between cloud services and traditional utilities lies in the degree to which cloud services are uniquely and dynamically configured to the needs of each application and class of user.

Cloud services are built from a common set of building blocks, but unlike the electricity provider, cloud providers configure them in unique ways for each specific application. For example, the building block configuration for a global public email system differs from an airline reservation system.

Why cloud computing is uniquely new We contend that cloud computing is uniquely new by simultaneously being an innovation ecosystem, a production environment and a marketplace, despite many technologies and concepts underlying cloud computing, such as virtualization, and applications residing on remote servers, not being new.

Figure 7

Characteristics of cloud computing Innovation Production Marketplace

Lowers the bar for new entrants Scale enables lower operating costs Provides global reach for applications

Facilitates experimentation Highly efficient management and upgrades Ability to scale user base rapidly

Lower capital outlays for startups Provides the building blocks for new systems Enables multisourcing (Mechanical Turk)

Enterprises use cloud to handle demand bursts

Supports new businesses (Dropbox, Netflix streaming, Foursquare)

Facilitates crowdsourcing (Kickstarter)

Improves allocation efficiency of internal IT resources


Simultaneously aninnovation ecosystem, aproduction environment

and a marketplace

Unlike traditional utilities, cloud services are

uniquely and dynamically configured

For providers, cloud services are competitive

propositions that differ from utilities



Cloud computing feeds the innovation ecosystem by lowering the bar for new entrants and facilitating experimentation. Most startup firms no longer require substantial capital outlays to build ICT capabilities. They can rapidly scale up or scale down operations as needed, and they can experiment with highly computing-intensive tasks. Cloud-based tools further lower startup costs. Larger enterprises with pre-existing datacenters can also utilize cloud resources for bursts of computing capacity for experimentation. Within their own datacenters, cloud computing-style architectures increase allocation efficiency of internal IT resources - to the extent possible with their corporate organization.

Cloud computing is becoming the baseline for efficiency and functionality for firms’ IT infrastructure. Global cloud providers’ scale enables far lower total operating costs than consumers’ own infrastructure. Since cloud providers can upgrade services in real time, users do not incur costly IT infrastructure upgrades. This accelerates the implementation of new technologies, to the detriment of those that do not adopt cloud services.

Cloud computing is also quickly becoming a production environment. We are now in an era when IT services are best considered part of production; systems are built, which then deliver services via IT networks. Cloud services, including raw storage and processing power, and platform-level tools, provide the building blocks for creating systems. For example, Dropbox’s popular file-synchronization and storage service and Netflix’s video-streaming service both use Amazon’s cloud infrastructure. Google and Microsoft’s powerful developer tools enable the ability to automatically generate cloud-based services and applications.

Cloud services extend the innovation platforms worldwide, becoming marketplaces with global reach. This is accentuated by the spread of apps for smartphones, tablets and browsers, putting within reach powerful building blocks, tools and entire ecosystems of third-party tools to anywhere with an internet connection. Phil Libin, CEO of Evernote, a popular note-taking and archiving software service founded in 2008, with more than 11 million users in three years and having raised over US$100 million in venture capital, made the following point: In his first two software startups, distribution was a major issue, particularly for overseas markets. However, with the advent of app stores, this was no longer an issue, and their ability to reach users all over the world to receive feedback even led to a dedicated user in Sweden who offered to become an investor (out of the blue), just when the company was struggling in its initial startup phase.3

Cloud computing also enables distributed forms of sourcing and collaboration. Amazon’s Mechanical Turk and startups like TaskRabbit create a marketplace where companies and consumers post jobs that anyone can bid on - anything from computer programming to picking up dry cleaning. Crowd funding has also been seeing a surge, with Kickstarter helping to raise over US$1 billion to date to fund projects that include business and creative ventures.

For advanced industrial countries, cloud computing provides new opportunities for innovation and entrepreneurship, and promises substantial efficiency gains. For developing countries, cloud services open up new possibilities to enter international markets and find niches in global value networks. As with the previous computing platforms (mainframes, PCs and networks of PCs), cloud computing is becoming a baseline for national and

3 Phil Libin, CEO of Evernote, in his keynote speech at Jannovation Jam, Sunnyvale, California. 30 August 2012.

Setting a new baseline for performance

Moving from development to production

Cloud services become global marketplaces

for innovators

Benefits for advanced industrial and developing

countries

Cloud computing reduces the capital needed

for innovation

Cloud computing enablesdistributed forms

of sourcing andcollaboration



corporate IT infrastructure against which other forms of infrastructure and service delivery must be measured.

There are, of course, still unresolved questions and potential risks associated with cloud-computing services - business-resumption strategies in the event of a catastrophic failure, though unlikely, of the cloud providers’ services, for example. While many larger firms will choose to retain some on-premise capacity, smaller firms may not have that luxury, becoming entirely dependent on one or more cloud providers. Cloud-service reliability also critically depends on the network service provision; no network, no cloud. This raises serious implications about different national contexts, with different regulatory regimes governing who can provide network infrastructure and the rules under which they operate.

Risks attach to network availability and regulatory

context

Section 3: Driving disruptive impact Blue Books


Driving disruptive impact As the PC unbundled software and hardware, leading vendors pursued different paths to escape commoditization, embedding business logic and offering high-end, customized “solutions”. Cloud computing disintegrates the vertical stacks, further commoditizing the physical layers. A more radical form of disruption results from decoupling business logic from the database and infrastructure software, further undermining vendor “lock-in”.

A disruptor and driver of commoditization Cloud computing is poised as a major disruptor, not only for companies that are currently providers in the IT industry, but also for a wide range of users.

For providers, cloud computing commoditizes a large portion of computing activity that had been offered on a custom basis.

For users, cloud computing democratizes large-scale computing, enabling even the smallest of users to access large-scale computing capacity. This creates the possibility of disruptions in all sorts of user industries.

The evolution of computing paradigms In the early era of computing, all the layers in the stacks were vertically integrated. Purchasing an IBM mainframe meant that you would be using IBM’s operating system and software. They were interlocking components of a system that could not be broken apart.

The advent of the PC era occurred with the unbundling of this vertical integration and the decoupling between software and hardware layers. With the unbundling of vertical integration, each layer became a marketplace with different sets of competitive logic. The PC itself fragmented into its constituent components such as memory, processors and hard disks, linked by standard interfaces. Competition developed in each of the components, and value moved away from the final assembly, which became increasingly commoditized. Processors, in which Intel dominated, became the area within the physical infrastructure that retained value.

The decoupling between software and hardware occurred as a new logic of competition took hold: the operating system of Microsoft Windows emerging as its own platform layer. As a platform, Windows provided a common set of Application Protocol Interfaces (APIs) that freed software providers from writing the code to control specific basic hardware functions, such as accessing memory and storage file systems. The software would work on any machine that would run Windows regardless of who made the components of the underlying PC, as long as it had an Intel processor and could run Windows. Windows quickly became dominant following the principle of a positive feedback loop, in which the more users adopted Windows, the more valuable it became as a platform for future users.

In the vertically unbundled and software-hardware decoupled PC era, Microsoft and Intel captured a disproportionate amount of value from the rapidly commoditizing PCs. Microsoft licensed Windows and built an ecosystem of software application that required Windows to function. And within the PC, although almost all other components could be assembled from any number of manufacturers, the Intel processor architecture was necessary

As the PC era unbundled vertical integration,

layers began to compete

The era of “Wintelism” commoditized everything

else

Cloud computing commoditizes large-scale

computing



to run Windows, enabling Intel to avoid commoditization as well. We have called this an era of “Wintelism”, since much of the value has been captured by Windows and Intel, with everything else rapidly commoditizing.4

A major outgrowth of the unbundling of vertically integrated systems was the emergence of software as a discrete industry independent from hardware. Microsoft prevailed over independent competitors in personal productivity applications (such as Lotus 1-2-3 and WordPerfect) by integrating disparate spreadsheets and word-processing applications into a suite. It is notable that the independent competitors that failed also lacked the leverage that an OS offered to Microsoft.

Despite the failures of alternative OS and productivity-application providers to sustain meaningful share against a dominant Microsoft, we note that the software industry has seen independent application and infrastructure companies prosper and thrive at scale. The heyday of the Wintelist era, from the mid-1980s through the late 1990s, coincided with the time that Apple was reduced to a niche player. Apple did not decouple its hardware and operating system, and became increasingly reliant on Microsoft’s productivity suite in the 1990s. (Its comeback from the late 1990s after Steve Jobs returned, however, was fueled by its control over both hardware and operating systems.)

High-end, custom IT as a path to escape commoditization As corporate computing quickly moved to embrace commoditized Wintel PCs, one path to escape commoditization was to provide high-end, large enterprise-size IT “solutions” that provided custom functionality. IBM was the most notable firm to evolve into this form, selling its hardware by bundling it with the services it delivered - a shift from selling hardware servers with built-in functionality. IT integrators found value in vertically integrating the different layers of the stacks (infrastructure, platforms and applications) in a custom package for the user firm, charging millions of dollars upfront for an implementation. They were therefore responsible for building and operating corporate datacenters, with the corporate applications to provide functionality.

A consequence of the rise of client/server, then multi-tier architecture, was the growing complexity and overwhelming plethora of choices facing business users. At every layer of the stack - OS, database, middleware (application server) and application - vibrant competition helped reinforce the prima facie appeal of a best-of-breed strategy. That IT integrators possessed the specialized insight and expertise necessary to source and implement key software projects drove massive expenditures for customized “solutions” tailed to their clients’ unique needs. This approach proved highly lucrative for the first generation of Enterprise Resource Management (ERP), supply chain and Customer Relationship Management (CRM) applications, but over time the burdens of cost and maintenance have eroded the value proposition.

4 Borrus, Michael, and John Zysman. "Globalization With Borders." Industry and innovation 4, no. 2 (1997): 141-166.

One path for escaping commoditization was

high-end enterprise-class IT “solutions”

Unbundling vertically integrated systems

allowed the software industry to emerge

A consequence of client/server then multi-

tier architecture was growing complexity



Avoiding commodification by embedding business logic Another path was to provide high-end equipment to enterprise-class users, such as Sun Microsystem’s servers and Cisco Systems’ routers, or software solutions that had highly customized built-in functionality that could not be separated into constituent elements, such as Oracle’s enterprise solutions, in which Oracle Financials could not be separated out from Oracle Database. This strategy can be characterized as avoiding commodification by embedding business logic into the offering.

Figure 8

Evolution of stacks

Source: harikrish.net

The appeal of vertical integration is increasingly apparent with the emphasis on high-end engineered systems and hardware/software appliances. Appliances saw increased prevalence as a delivery medium over the past decade, most notably in network security where the integrated approach offered a hardened OS and system optimized for high performance and throughput. The appeal extended to data warehouse appliances and gained momentum following Oracle’s acquisition of Sun.

Oracle’s Exadata systems combined server, storage and networking hardware integrated and optimized with database, and subsequently middleware and analytics tools. These systems were initially targeted as a general-purpose database and hardware system to support both transactional applications and data warehousing scenarios. Over time, the company’s strategy reflected integration up the stack into middleware (Exalogic) and analytics (Exalytics). IBM’s Pure Systems take a modular approach towards providing systems with components optimized to specific applications. Oracle’s hardware business encapsulates this dynamic: as the commodity storage and Intel-based server businesses tail off, the company focused on driving growth from high-end engineered systems to offset its declining revenue streams.

Disintegrating the vertical “stack” With the advent of cloud computing, a further, more radical type of vertical disintegration and commoditization is occurring. The physical infrastructure itself is becoming unbundled from platform layer to an entirely new degree. The disruptive implication of this is that offerings in lower layers of the stack, such as storage, networking and even databases, are becoming commoditized

The layering of technology is giving way to the layering

by function

Vertical integration’sappeal is evidenced with

growth in hardware/software appliances

The physical infrastructure itself is

becoming unbundled from the platform layer

High-end equipment as another path to avert

commodification



more than ever before. “Value-added” management controls embedded in storage, networking or databases that optimize their resources and allow providers to charge premiums are falling by the wayside as value moves upwards in the stacks.

Figure 9

Value moves up in stacks

Source: Jonathan Murray (Murray's First Law of Platform Economics)

First, let us look at disintegration by taking a single-user example, then the context of corporate computing. In the PC era, you might not care who made your hardware as long as your operating system worked, but your copy of the operating system was tied to a specific computer. With the advent of cloud computing, for applications built on top of cloud-computing infrastructure, you don’t care where your hardware is, what it is or what it is running. The functionality delivered to you is usually through a web browser.

As a user, you really don’t care where or how the backend servers of your Gmail, Google Docs, Microsoft 360, Dropbox or Netflix services are running. To start a service, you can simply rent capacity from Amazon. In the corporate context, it is clear how value in the stack is moving up towards applications and business logic, commoditizing the lower infrastructure layers.

Disrupting the infrastructure layer Let’s take the infrastructure layer, which encompasses networking, storage servers and software for management and security. We will start with servers. Firms offering high-end, high-performance servers, such as Sun Microsystems, were once the darlings of Silicon Valley. Their fast and powerful servers gave users requiring massive computing power a competitive edge. However, with the advent of cloud computing, there is much less advantage for users to buy individual high-performance machines with high price premiums.

Google was the first paradigm shifter in using cheap, off-the-shelf computers rather than high-end servers to build datacenters. Its approach was that higher hardware failure rates could be built into the parameters of its software design, with the ability for databases and other tasks to be distributed among large numbers of physical hardware, unaffected by failures of particular hardware. Performance came from the algorithms for distributing the tasks rather than the performance of the hardware itself.5 This approach

5 For an excellent overview of the Google approach, see Levy, Steven. In the Plex: How Google Thinks, Works, and Shapes Our Lives. 1st Simon & Schuster hardcover ed. New York: Simon & Schuster, 2011. See also Barroso, Luiz André, and Urs Hölzle. The Datacenter as a Computer: An Introduction to the Design of Warehouse-Scale Machines. [San Rafael, Calif.]: Morgan & Claypool, 2009.

Commoditization erodes advantages of high-

performance machines

Users of cloud-basedapplications don’t care

where hardware is, whatit is or what it is running

There is less advantagefor users to buy individual

high-performancemachines

Google was the firstparadigm shifter in using

cheap, off-the-shelfcomputers

Value migrates from infrastructure to platform

to application



enabled Google to offer Gmail initially with 1 gigabyte (GB) of storage in 2004 (doubling it to 2GB a year later), at a time when Microsoft’s Hotmail offered 2 megabytes (MB) and Yahoo! Mail was 4MB - 500 times and 250 times, respectively, the amount of their competitors’ capacity - though initially to a limited number of users. Put simply, this was the first big step in physical infrastructure servers becoming unbundled and commoditized.

Cisco is also facing threats as networking hardware becomes commoditized. As investment in building out the internet took off in the mid-1990s, Cisco Systems became enormously successful in providing networking equipment. It dominated global markets for internet backbone equipment and was a strong presence in large corporate networking solutions. The physical need for routers to connect datacenters, servers and machines to each other and to the internet gave Cisco and its competitor Juniper Networks opportunities in the networking portion of the infrastructure layers.

Recently, Cisco is under an unprecedented level of pressure because its high-end networking infrastructure solutions are increasingly achievable through software. VMware’s acquisition of Nicira’s Software Defined Networking technology raised concerns that networking routers and switches could face disruption analogous to the impact of virtualization on the server market. The parallels are not exact, however; under-utilization and overcapacity of non-virtualized servers were a prime target for cost reduction. Nonetheless cheaper commodity hardware can replace both proprietary server and network hardware. For example, a major user of IT equipment that is currently in the process of revamping its entire IT infrastructure found that Tier 3, a Seattle-based company, was able to match Cisco’s high-end, hardware-based networking solution through software, importantly, at a far lower price.

Value moves from infrastructure to the platform Management functions within software are increasingly migrating to the platform rather than being encapsulated in individual components of datacenters, whether it be storage, networking or databases. This drives commoditization of the components.

Take databases, for example, which at the simplest level are comprised of the database itself, a business logic that manages the database and the user experience. These three used to be integrally tied together. Oracle, for example, in its offerings does not let you decouple Oracle Financials from Oracle Database. Oracle is therefore embedding a great deal of management logic into its database offerings.

Cloud platforms accelerate commoditization of the database However, with the advent of cloud-based platforms, it is rapidly becoming possible to procure just a pure database offering without any management logic embedded in the database. The management logic is taken care of at the platform level, managed through an open API by the database. From the user’s vantage, you can manage multiple resources through the platform, with multiple components open at the API level to control from the platform. Therefore, as a user, you want standalone components that can be mixed and matched, with the possibility of migration from one to another - your financial system will not be coupled with a particular storage engine, and either can be moved to a new provider. From the perspective of the providers, this increases the pressures of commoditization, since there are far fewer ways to lock in customers to their offering.

Cisco is under pressure as networking becomes

commoditized

Management capabilitiesevolve up the stack,

commoditizingcomponents

Cloud reduces ways to lock in customers

Cheaper commodity hardware can replace

both proprietary server and network hardware



Eroding stored procedures’ role in vendor lock-in As a current business reality, businesses in all layers have some business logic in each layer, including the lower layers, such as “stored procedures” in databases. But these stored procedures are what lock users into particular databases. The initial rationale for such stored procedures was for performance optimization and security. However, performance is now optimized by the cloud-computing economics of abundance, in which you can simply allocate more computing resources rather than pursue a customized performance database.

Users at the leading edge are pushing database commoditization Cutting-edge lead users that are adopting cloud-computing solutions on a major scale are beginning to move towards not allowing any stored procedures in the database, instead forcing the database to be controlled only by the platform layer through an open API. The implications are that those providing business functionality in the package and charging premium prices will face a different kind of competition.

Figure 10

Pre-cloud era value to cloud model commoditization


This disruption is so pervasive that it even affects some of the frontrunners of quite recent disruptions. There is such tension created that a firm may choose to compete with itself, VMware, for example. VMware is a company that provided virtualization - a layer of software that enabled the underlying physical infrastructure to be abstracted from the layers above. In late July 2012, it purchased Nicira, a company that provides software-based network virtualization. However, VMware is owned by EMC, which operates many of the virtualized datacenter component offerings such as storage. These are exactly the areas that VMware is helping to commoditize rapidly.

The leading technology vendors are forced to compete against themselves in order to mitigate the disruptive impact of the transition to the cloud. Oracle has launched both a private and public cloud, and has been acquiring Software-as-a-Service vendors aggressively, such as RightNow and Taleo, in order to stave off customer defections to cloud-based competition. Microsoft’s launch of Office365 and Windows Azure provide cloud-based alternatives to productivity suites, database, middleware and server offerings, while IBM is similarly investing in its own hybrid cloud offerings.

Disruption forces leaders like VMware to compete

with themselves

Stored procedures are what lock users into particular databases

Leading-edge users are freeing stored procedures from underlying database



New uses of technology lead the disruption The broader point that we need to make here is that disruptions are almost never just about the introduction of new technologies or equipment. Rather, they are about the experimentation by lead users, and the new uses to which they put the new technologies.

For example, in the case of electricity, it took almost 50 years for electrification to yield major productivity gains. This was because the factories were initially set up according to the logic of steam engines; the machines were connected to a centrally located steam engine with a system of belts. The first electronic motors simply replaced the steam engine. Only after the factories were reconfigured, with machines placed according to the logic of production rather than the previous logic of steam power, did productivity gains skyrocket.

Likewise, the earliest computers were simply replacements for calculators. Only after innovative lead users installed them to solve one set of problems - performing large numbers of calculations quickly - did they discover new uses. Airlines, for example, installed them to handle reservation systems, but then discovered they could manage and adjust routes based on reservation information. With the advent of databases, computers transformed from powerful calculators to “what-if” machines that could calculate probabilities and contingencies.6 Cloud computing is the next stage in this evolution of production.

6 See Cohen, Stephen S, John Zysman, and Bradford J DeLong. “Tools for Thought: What is New and Important about the “E-conomy”?” (2000). http://brie.berkeley.edu/publications/WP138.pdf

Disruptions are about theexperimentation by lead

users and their new uses

It took almost 50 yearsfor electrification to yieldmajor productivity gains

Cloud computing is the industrialization of IT

Section 4: Battle lines square off Blue Books


Battle lines square off Provision of the cloud involves four different sets of firms in the ecosystem, each with different competitive dynamics: cloud-service providers, connectors (network providers), device providers and arms suppliers. For service providers, global scale is an advantage, but regulations offer openings for localized providers. The starting points of Amazon, Microsoft, Google, Salesforce.com and Apple each define their strategies. Network providers must seek higher value to avoid commoditization. Arms suppliers Intel, Cisco and Oracle must adjust strategies to address the cloud’s existential threats to their core businesses.

We consider each set of firms separately since the logic of competition differs:

Cloud-service providers create, configure, run and distribute services from their cloud datacenters. Major cloud providers such as Microsoft, Google, Amazon, SalesForce.com and a plethora of smaller players are vying for consumers, firms and developers with an ever-widening and overlapping set of service offerings. The competitive game is winning the user.

Connectors (network providers) offer access networks to distribute cloud services from the cloud providers to access devices. Firms such as AT&T and Verizon face the serious threat of becoming commoditized. Global-scale cloud-service providers are building their own networks and new access devices place ever-increasing demands on network bandwidth and infrastructure. They are politically powerful actors, however, engaged in the search for value.

Device providers offer the access devices such as smartphones, tablets and PCs through which users access cloud services. A new generation of devices, and the operating systems on which they run, are pitted against each other with contrasting business models. These are the device wars.

Arms suppliers provide the equipment and infrastructure for cloud providers and connectors to provide and deliver cloud services. Firms such as Cisco for networking equipment and Oracle for databases are facing a “commoditization trap” as their hardware functionality is moving to software, and value-added functionality is moving into the cloud providers’ platforms.

We can put each of these players onto our technology stacks introduced earlier to get Figure 11. This will help us sort through their competitive positions.

Cloud-service providers’competitive game is

winning the user

Network providers are politically powerful actors

engaged in thesearch for value

Providers of accessdevices are engaged

in the device wars

Arms suppliers are facinga commoditization trap

Four sets of firms in the ecosystem, each with different competitive

dynamics



Figure 11

Competitive framework for providers


Cloud-service providers: Winning the user For cloud-service providers, there are a few big questions that will shape the industry on a global scale. Will global scale dominate, or is there room for firms that principally offer

local and more limited local deployments? Will provider strategies be limited by where they started?

The answers to these questions are not simply a matter of market dynamics. They also involve government policies that differ significantly around the world.

Scale offers advantages, to a point A major question for the cloud-provider industry is whether the logic of global scale will dominate, or whether there is room for local-scale players to carve out sustained market shares.

At first glance, it seems obvious - yes, global scale will prevail. Only a handful of companies can invest in numerous datacenters that exceed US$500 million per datacenter. Google and Microsoft are also aggressively purchasing physical networks to enhance their services and lower operating costs. The degree to which they can provide the illusion of infinite resources for instant scalability and offer platforms that take advantage of their databases, storage and other resources cannot be matched by smaller players.

However, when taking into consideration the importance of regulatory requirements and structures around the world, along with the fact that the largest global cloud providers are American firms, the answer becomes less obvious. Regulatory structures may provide room for smaller players with national or regional deployments. The very nature of global cloud services, in which the location of data is not inherently related to national borders, provides potentially strong disincentives for users - particularly large corporate and government users - from putting data in them.

Only a handful of companies can spend

US$500m on datacenters

Country regulatory requirements may benefit

smaller, local players

How we view the competitive dynamics

around cloud providers



The specter of the US Patriot Act, which allows the US government to access not only data that pass through the USA, but also all data owned by all multinational companies with a US presence, create serious concerns. Data held by Amazon, Google and Microsoft, for example, could be seized by a US government concerned by terrorist threats. (Indeed, the quick response of Amazon to jettison Wikileaks from using its storage service was likely an attempt to avoid a potential US government seizure of some of its data.)7

Regulations, particularly in Europe, stipulating that certain types of data must be stored domestically, can provide room for local storage providers. In some cases, for large corporate users, the slogan of “our data are subject to the US Patriot Act” may be enough to win a particular market niche.

The challenge for local or regional players, however, is that they are unlikely to have the capabilities for moving into areas of higher value, namely platforms. Yet, even though they face pressures for commoditization, the regulatory structures stipulating that certain types of data or functions remain in the domestic market, or corporate preferences to keep particular functions and data within national borders, will protect them from pure commodity price competition against the global-scale providers.

Deutsche Telekom, for example, marketed its cloud-storage service as being free from the US Patriot Act, and therefore more secure for large enterprises. Japan’s NTT Data also provides the datacenters to store the key databases for large users of Salesforce.com’s services platform, such as Japan Post, the postal service.

One can pose this question the other way around: does the fact that global-scale cloud providers operate on a global scale limit their ability to serve particular, reasonably large markets? This would imply that the logic of global cloud services provision is not a worldwide “winner-takes-all” market. That is, the benefits from low cost achieved by economies of scale would only reach a certain point before tapering off. It would allow a myriad of local, smaller players with offerings tailored to local regulatory, and possibly social and political, conditions to carve sustained niches. Given the disruptions to existing players’ business models brought by the cloud, the question is who will be the incumbent players, such as telcos and systems integrators, that successfully carve this niche? And which niche will it be? For example, some may specialize in healthcare-data storage and analytical tools, while others focus on finance or serving the government.

7 In the past few years, US court cases revealed some concerns. For example, the FBI used the Stored Communications Act to access a person’s email without a warrant or his consent, and the court was unable to determine whether they were subject to any expectation of privacy. In another event, the FBI raided and seized servers from two Texas datacenters with search-and-seizure warrants for a wide range of equipment. However, included in the equipment seized were servers that had data of a large number of clients unrelated to the investigation. The district court sided with the FBI. A major question is whether the Fourth Amendment, which protects against unreasonable searches and seizures, applied to data held by cloud providers. This has yet to be directly tested in courts. See James Urquhart, "Does the Fourth Amendment Cover 'the Cloud'?," in Wisdom of the Clouds (2009). David A. Couillard, "Defogging the Cloud: Applying Fourth Amendment Principles to Evolving Privacy Expectations in Cloud Computing," Minnesota Law Review 93, no. 6 (2009).

US government regulations can challenge

global cloud providers

Data-storage regulations,particularly in Europe,

can provide room forlocal providers

Local or regional players are unlikely to have the

capabilities to move into areas of higher value



Figure 12

Representative regulations affecting cloud-data deployment

Regulation Country Description Implications for cloud deployment

Health Insurance

Portability and

Accountability

Act (HIPAA)

USA Establishes standards for electronic healthcare transactions and encourages electronic data storage of healthcare information.

If health-related data are stored in the cloud, controls must be established to ensure compliance and security. Violators are subject to major fines.

Sarbanes Oxley

(SOX)

USA Requires management to certify accuracy of financial

information.

If financial data are stored in the cloud, controls must be established to ensure compliance and security. Management is subject to fines or imprisonment for violations.

Gram-Leach-

Bliley Act

USA Requires that financial information be properly

safeguarded and completely destroyed upon disposal.

If financial data are stored in the cloud, controls must be established to ensure compliance and security. Board of directors is subject to punishment for violations.

Breach

Notification

USA - CA Requires notification to residents of any breach of

computing system possessing personal information.

If a cloud provider’s security is breached, all of an

enterprise’s clients must be notified, and service must be

suspended until compliance is restored.

Data Protection

Directive

EU Regulates how personal data may be processed, shared, stored and accessed within the EU. Personal data cannot be transferred outside of countries not on a EU “safe list,” including the USA.

Data stored in the cloud that contains personal information of

EU residents must meet a series of security requirements

before transfer outside of the EU can be permitted.

Data Protection

Act

UK Defines lawful purposes for personal data collection and provides means for individuals to control information about themselves.

Personal data cannot be transferred to states outside of the

EU unless they are deemed to possess an adequate security

framework. Enterprises are responsible for securing their

data, even if storage is transferred to the cloud.

Data Protection

Act

France Established the French Data Protection Authority

(CNIL), which enforces privacy standards stipulated

by the law. CNIL has the power to investigate any

instance of data collection occurring in France.

Consent of data subjects required before processing

personal data.



framework.

Amendment to

Data Protection

Act

France Increases the authority of CNIL and establishes

additional sanctions for breaches of the law.



framework.

Personal

Information

Protection Act

Japan Applies to entities that access or store personal

information of 5,000 or more individuals.

Consent must be given before personal information can be

transferred to a third party, such as a cloud provider.

Source: SANS Institute, Technet, InfoLawGroup, MDB Avocat, IBM8

How will starting points limit their strategies? Today's cloud offerings emerged from major providers’ infrastructures for core businesses: Amazon's online retailing; Google's search and advertising; Microsoft's global web email and small business services; and Salesforce.com's online CRM. These firms shared the IT challenge of delivering continuously reliable, responsive service on a global scale and at low cost. Their distinct technology infrastructure starting points and

8 http://technet.microsoft.com/en-us/magazine/hh994647.aspx; http://www.sans.org/reading_room/whitepapers/legal/overview-sarbanes-oxley-information-security-professional_1426; http://www.mdb-avocat.com/index.php; http://www.infolawgroup.com/2010/10/articles/breach-notice/legal-implications-of-cloud-computing-part-five-ethics-or-why-all-lawyers-not-just-technogeek-lawyers-like-me-should-care-about-data-security/; https://www.ibm.com/developerworks/mydeveloperworks/blogs/CloudComputing/entry/eu_data_protection_directive_and_cloud_computing?lang=en

Attacking the market from different angles



development trajectories led to a set of common cloud services with quite distinct technical implementations. They compete on the basis of these differences in implementation.9 Note that all are moving towards the platform layer, including operating systems and PaaS.

Amazon The company pioneered infrastructure cloud services, initially out of its IT needs to serve retail demand peaks. Amazon prioritized building IT infrastructure capacity to comfortably handle the peak Christmas holiday season. To monetize the idle capacity for the rest of the year, it introduced Amazon's Elastic Compute Cloud (EC2) service in 2006. Amazon's rapid success with EC2 and S3, a data-storage service, opened up an entirely new cloud-computing business for the company. Moving well beyond excess capacity from retail, it now drives substantial new investments in datacenters and services.

Amazon has continued to build out from its Infrastructure-as-a-Service roots to higher layers within cloud datacenter and across into access networks (see Figure 13). In 2007, Amazon introduced CloudFront, a Content Delivery Network (CDN) service. (A CDN distributes copies of commonly requested information across several geographically dispersed datacenters, accelerating response times.)

Figure 13

Amazon’s strategy


9 This section draws heavily from The Gathering Storm: Analyzing the Cloud Computing Ecosystem and Implications for Public Policy (1 March 2012) by Kenji Kushida, Jonathan Murray and John Zysman.

Amazon has continued tobuild out from its IaaSroots to higher layers

within cloud datacenter

Amazon plays in the infrastructure and

platform layers, but not yet in applications

Amazon pioneeredinfrastructure cloud

services



By 2011, Amazon's major datacenter locations outside the USA included the UK, Ireland, Netherlands and Germany in Europe, and Singapore and Japan in Asia. Amazon strengthened its offerings to enterprises in 2009 with its Virtual Private Cloud (VPC) service, which facilitates linking users’ legacy IT system and new cloud-based services.

Increasing competition in the IaaS layer from smaller firms including RackSpace, CloudSigma and others intensifies cost pressures. From the top, Google and Microsoft are also broadening their Software-as-a-Service and PaaS offerings, expanding into IaaS. Amazon's “Elastic Beanstalk” PaaS in 2011 was a direct response, aiming to facilitate a developer ecosystem for complex cloud applications, pitting Amazon squarely against Google and Microsoft for cloud-application developers. Amazon is increasingly seeking to target higher-valued solutions with the announcement of the pending RedShift data warehouse as a service.

A key tenet of Amazon’s cloud strategy targets startups and departmental users. Amazon has become the de facto standard for internet startups including Foursquare and Pinterest. Its foray into access devices began with its Kindle e-book reader, first introduced in late 2007, tightly integrated into its online bookstore. In 2011, it offered a tablet, the Kindle Fire, with a proprietary customized version of the Android operating system modified by Amazon, capable of web browsing with the proprietary Silk browser, streaming video and running apps. Amazon runs its own app store and integrates the video streaming with its streaming-video online store, which utilizes its own IaaS infrastructure.

Microsoft Microsoft is moving into almost all areas of cloud services. Its need to decrease cost and increase scale in its (often free) online services such as Hotmail and MSN Messenger, which predate any conception of cloud computing by five to 10 years, catalyzed its cloud strategy. Until recently, these massively scalable services, with several hundred million global users, did not share a common infrastructure inside Microsoft. The drive to reduce operating costs and increase service flexibility led Microsoft to develop the underlying platform technologies for its current cloud-service strategy. Microsoft has been in the platform business since Windows (see Figure 14).

In the early 2000s, Microsoft introduced a new set of platform technologies for web and internet-based developers. In 2008, facing growing competitive pressure from Google and Amazon, the company introduced its Azure platform, enabling developers to build cloud-based applications. Windows Azure charges for consumption of underlying computing, storage and network resources on a per-unit basis - a significant change in Microsoft's business model since the company received a one-time payment for each sale of Windows without revenue streams from third-party applications. The new revenue streams in part drive Microsoft's strategic emphasis on the cloud.

Microsoft also continues to develop its SaaS suites such as Office 365 and Dynamics CRM Online, both to broaden its user audiences and as defensive strategies to retain existing customers migrating to the cloud. The company is extending downwards in the architecture stack, introducing an IaaS supporting virtual machines on Windows Azure that actually supports CentOS and SuSe Linux guests. It is also extending horizontally into network services, providing CDN services for Azure developers. Microsoft has invested billions of dollars in global cloud-services datacenter infrastructure in over 20 locations worldwide.

The drive to reduce costsand increase flexibility led

Microsoft to develop itscloud-services platform

A key tenet of Amazon’s cloud strategy targets

startups and departmental users

Windows Azure is asignificant change inMicrosoft's business

model

Microsoft also continuesto develop its SaaS suites



Figure 14

Microsoft's strategy


Finally, Microsoft's dominance in PC operating systems and applications faces serious pressure with access devices that do not run its operating systems - most notably Apple products and the plethora of devices running Google's Android. Microsoft was early in categories such as mobile phone operating systems and tablets running traditional Windows, but fell behind the competition. The success of its new attempts, Windows Phone 8, Windows RT and the SurfaceRT Tablet remains unclear, but they represent a major departure in Microsoft’s strategy towards hardware. In introducing its own tablet, the Surface, the company shifted away from its Wintel PC strategy of leaving hardware up to third-party providers. Its Windows Phone offering is in close partnership with troubled mobile handset manufacturer Nokia, but there have been speculation of Microsoft preparing its own smartphone. Microsoft’s mobile operating system, Windows RT, is designed to run on ARM architecture used on tablets. Some point to the potential disruption this may cause firms such as Intel, since a large number of cheaper ARM servers may provide higher-price performance than a smaller number of more costly Intel-designed processors - a new twist to commodifying hardware. In terms of apps, Microsoft launched its Windows Store as the sole avenue for users to download and deploy applications for Windows RT (the ARM-based OS).

The company increasingly emphasizes greater organizational integration across the disparate divisions to realize a “software plus services” strategy. Playing a key role is the extension of the Xbox brand beyond the gaming consoles and Xbox Live to include Music and other media content. With Xbox360, Kinect and the SurfaceRT, Microsoft is gradually expanding its footprint of access devices, which may be topped off by introducing a branded smartphone as well.

Microsoft has moved from platforms into

infrastructure and applications

Microsoft emphasizesgreater integration to

realize a “software plusservices” strategy

Its dominance in PC operating systems and

applications faces pressure



Google Google, though with a different starting point, now overlaps considerably with Microsoft in its range of cloud services, but with quite different business models. Its wild success in monetizing search through advertising provided the capital to experiment and expand. The company's massive investments into datacenters and access networks underpinning its search and advertising business provided the launch point for a range of additional cloud services. Google began with the introduction of Google Apps, which provided a cloud-based alternative to Microsoft's ubiquitous office applications (see Figure 15). Google Apps are true cloud-based applications, delivered from Google's datacenters, consumed via a web browser, with users' files stored in the cloud.

While originating as consumer offerings, Google has moved quickly to add features to attract firms of all sizes, including large enterprises and governments. Google, like Microsoft, is moving down the stack to attract the cloud-developer community. In 2008, Google introduced its App Engine PaaS offering, allowing developers to leverage some of the powerful underlying services Google developed for its own search and advertising businesses.

Figure 15

Google’s strategy


Its App Engine platform is more limited in its user configurability than PaaS services by Microsoft and Amazon. In 2011, it introduced an IaaS database service, Google Cloud SQL, for use with Google App Engines. It remains to be seen whether Google offers more generalized IaaS services, or whether it sees an advantage with keeping the IaaS element bundled with its App Engines.

In 2008, Google released its Android mobile operating system, which moved it firmly into the access-devices stack. From 2010, Google began producing Nexus-branded handsets and tablets, manufactured on an OEM basis from firms such as Asus, HTC, LG and Samsung. In the first of the line, the

Google now overlapsconsiderably with

Microsoft

Google has aggressively built from application to platform then to devices

App Engine is morelimited in its

configurability than Azureand AWS’s PaaS

It is moving down thestack to attract the cloud-

developer community



NexusOne smartphone, Google experimented by offering a new direct online retail channel to decouple handsets from network provider channels, though it folded the store later that year. In 2011, Google purchased Motorola's former mobile handset division, Motorola Mobility, for over US$10 billion, historically its largest acquisition, most likely to secure the intellectual-property underpinnings of Android and Nexus devices. In 2012, Google folded its Android Market into Google Play, an online store selling e-books, magazines, music and video.

Salesforce.com Salesforce.com pioneered the SaaS business model. It was founded in 1999, offering CRM over the web with a pay-as-you-go fee structure. Dominant competitor products, such as from Siebel Systems, cost several tens of millions of dollars for large customers and was largely driven by the technical complexity of integrating on-premise CRM software with customers' existing systems. Salesforce delivered the same functionality as a service, avoiding on-premise implementations and was wildly successful.

Figure 16

Salesforce.com’s strategy


Salesforce's CRM was initially attractive to smaller users, offering low upfront costs and the ability to quickly scale usage up or down. The company soon expanded to larger users as it grew functionality and is seeing the proportion of its business tilt increasingly towards large enterprises.10 Salesforce moved into a platform business by focusing on its core CRM services while encouraging the development of a supporting ecosystem of software developers for specific industries and niche markets. In 2006, Salesforce introduced its AppExchange

10 Benioff and Adler, 2009.

Salesforce.com pioneeredthe SaaS business model

It is seeing growingproportion of its business

from large enterprises

Salesforce moves from the app layer to the

platform layer



marketplace, which provided marketing, commerce and distribution service for third-party developers to extend the core CRM application. Its success has taken time to build momentum, but has become a major competitive strength, offering a rich portfolio of third-party applications.

In 2007, Saleforce.com formalized this platform with Force.com (Figure 16), which extended AppMarket with a set of tools and lower-level services for developers, making it easier to develop applications for the Salesforce.com ecosystem, and for which Salesforce.com would provide the hosting and distribution. The Force.com platform saw slower-than-expected adoption because of the requirement for developers to use the proprietary APEX programming language. An attempt to jumpstart appeal to Java developers through VMForce, a VMware joint venture, quickly failed. With the acquisition in 2010 of Heroku, a leader in multi-language technology, the company’s momentum in PaaS has steadily gained traction. Thus, Salesforce.com, like Google and Microsoft, started in the SaaS layer, moving to PaaS. The company recently passed the US$3 billion annual revenue run rate, making it by far the largest pure cloud-computing applications and platform vendor.

Apple Apple is now clearly a major player in the cloud ecosystem. It began by revolutionizing access devices with its vertically integrated offerings iPhone and iPad, and tying them to datacenter-based services and content. Many apps, whether news, music or certain types of games, initiate a connection with a cloud SaaS layer service, which provides core content or live information. The app is the user interface to a cloud service accessed through the internet, and is provided by AppStore's large-scale cloud service that leverages multiple layers of the cloud datacenter, access network and access-device stacks (see Figure 17).

Figure 17

Apple's cloud service architecture


With the acquisition ofHeroku, momentumin PaaS has steady

gained traction

It is now clearly a majorplayer in the cloud

ecosystem

Apple revolutionized access devices with its iPhone and iPad



Apple is unique as a cloud provider in using the cloud to primarily serve the needs of its own vertically integrated ecosystem. The AppStore is not available for non-Apple devices and therefore other operating systems. iTunes is also a very large-scale cloud service which enables you to browse, buy and download music and videos from multiple operating systems, including Windows.

In 2011, Apple revamped its MobileMe service into iCloud, providing a range of cloud-based capabilities including, email, music storage and device management in an offering closely tied to its devices. Apple began its resurgence in the late 1990s by focusing on the consumer market, but as the iPad and iPhone are increasingly adopted by corporate users, Apple has strengthened security measures. Its recent billion-dollar investment in a North Carolina cloud datacenter speaks loudly about its future ambitions.

Network providers (Connectors) The question for connectors is to what extent they are replaced by cloud providers and what potential value-added services they can offer? Can they provide anything beyond the commodity “pipes” linking cloud-service providers to access devices?

Traditional telcos are being squeezed at both ends. Global-scale cloud-service providers like Google and Microsoft have already made considerably investments into infrastructure. For example, in 2009 Google ranked third worldwide in the total volume of internet traffic carried over its own private networks, displacing AT&T, Verizon and Sprint.11 In November 2012, Google launched a pilot in the Kansas City area for Google Fiber, a high-speed internet service. Telcos such as Verizon have begun offering cloud-storage services. However, these are the most quickly commoditized areas.

CDNs and MSS as value-added service A potential value-added area for connectors that appeared relatively early as the internet became mainstream in the mid-2000s was the area of Content Delivery Networks (CDN). A CDN distributes copies of commonly requested information across several geographically dispersed datacenters, accelerating response times to compensate for network latency. This was a market that startup firm Akamai dominated and traditional telcos were not quick to move into. As the cloud datacenters become the locus of activity, with processes inside the datacenter taking the majority of the value, CDNs are also under pressure for commoditization. Notably, Akamai is developing a version of its own software to enable service providers to operate their own proprietary CDNs, and in November 2012 announced its first major partnership with Orange. Citrix acquired ByteMobile to pursue a similar strategy to Akamai’s.

Other value-added services include managed security services (MSS) and application hosting. The MSS market has undergone several rounds of consolidation, with connectors typically partnering with third-party specialists to white-label security services.

11 Labovitz, Craig, et al. "Atlas internet observatory 2009 annual report." Arbor Networks, the University of Michigan and Merit Network, Tech. Rep (2009).

Apple is unique in usingthe cloud to serve the

needs of its own verticallyintegrated ecosystem

Apple has investedbillions in a North

Carolina cloud datacenter

For connectors, thequestion is whether theyprovide services beyond

commodity “pipes”

A potential value-added area for connectors is

Content DeliveryNetworks



Telcos pursue strategies to outrun commoditization The major telecom firms, however, are not simply sitting back to die slowly. Their challenges are likely to come from regulatory strategies, utilizing their long history of working closely with governments (and in many countries, their origins as state-owned monopolies).

Political debates over network neutrality in various countries are a prime example. Cloud providers argue for equal treatment of all bits flowing through network providers' networks, but this commoditizes the latter. Network providers argue for deregulation enabling differential tariffs or service levels depending on the type of data and cloud providers’ willingness to pay.

Incumbent telcos have also begun offering infrastructure services. Terremark, owned by Verizon, offers virtualized datacenter capacity as blocks of resources, and is also focused on enabling firms to rely on multiple cloud deployments.

Access devices: The device wars Today's access devices derive their lineage from computing equipment, and they have become an intensely competitive battleground. Winners will not only define end-user experiences, but may also shape the nature and success of cloud services. The direct lineage from PCs suggests the likelihood of similar competitive battles: at the infrastructure layer over operating systems, processors and hardware; at the platform layer for developers; and at the application layer for end-user loyalty.

Tablets, PC, smartphones and everything in between The advent of cloud-service access has reshaped the nature of hardware competition, an ever-expanding category, currently including traditional PCs, netbooks, smartphones and tablet devices (such as the Apple iPad, Kindle Fire, Microsoft Surface and Samsung Galaxy Tab) as manufacturers experiment with new form factors that serve various cloud-enabled uses.

It is important to reflect that both PC hardware and traditional mobile handsets were rapidly becoming commoditized by the mid-2000s with a new generation of smartphones devices with cloud access, spearheaded by Apple's iPhone and then the iPad, which reignited interest in - and reimagined - the role of access devices. Below that surface, aggressive competition for the central processing units (CPUs) is unfolding. Unlike in PCs, Intel is the underdog against multiple vendors producing processors based on ARM Holdings’ architecture.

Apple’s control of the experience provides major differentiation Apple's control of its products over all three layers of architecture - from device hardware to platform to user experience - for both the iPhone and iPad differentiates it from other players in the market and clearly factors in its popularity. Others, including HP and RIM, are discovering their positions untenable in tablets due to their choice of weak platforms or limited services.

In operating systems as well, old battles from the PC era are playing out, and again, with the PC era's dominant player, in this case Microsoft, as the underdog. Google's Android and Apple's iOS are battling for dominance, and Nokia adopted Microsoft's Windows Phone 7 after abandoning its own.

The major telecom firmsmay use regulatoryleverage to outrun

commoditization

Incumbent telcos havealso begun offering

infrastructure services

Cloud access hasreignited interest in andreimagined the role for

access devices

Apple's control of itsproducts over all three

layers of architecturedifferentiates it

Access devices havebecome an intensely

competitive battleground



Figure 18

Comparing business models Companies Amazon Apple Google Microsoft Device/OEM strategy Own Own OEM, Own OEM, Own Revenue streams E-commerce, online store,

hardware, cloud services, AWS Hardware, online store, services,

Advertising, cloud services, hardware,

Licensing, cloud services, online store, hardware, advertising

Advertising Some in online store No Yes Yes Use of proprietary data Yes, online store Some, iCloud Yes, Search/ads Yes Proprietary content/games No No No Yes (Xbox Halo, etc) Source: John Zysman, Jonathan Murray, Kenji Kushida

Business models do battle Multiple platform business models are vying for market share.12 Apple's iOS operating system is integrated into its devices, only available on Apple products, and Apple takes revenue from third-party apps. Microsoft offers PC-style licensing with its operating system, charging manufacturers per unit, and has only recently expanded to offering its own tablets. Google's Android, by contrast, is free and largely open source, with Google extracting value through increased penetration of its cloud services and advertising revenue streams on Android-based devices. Amazon subsidizes the price of its Kindle tablets with the expectation that consumers will purchase more e-books and other products.

For PCs, Microsoft's dominance of operating systems became a policy issue in the USA and Europe. Microsoft's bundling of its Internet Explorer operating system into Windows PC was the focal point of antitrust action. In the cloud-computing ecosystem, policy issues may develop around linkages between operating systems and search. While a Windows/PC-style domination is unlikely, the ability for Google or Apple to leverage dominant positions in one market to gain advantage in another is already causing regulatory scrutiny.

Another platform layer service becoming the focal point of competition is the AppStore service, with offerings from each major device and operating-system provider, and even some network providers. Apple’s AppStore, Google Play and the Windows Store all offer apps built for their own platforms, while Amazon’s Kindle include a range of Android apps. Offerings entail a cloud service to browse applications, commerce infrastructure, distribution, license management and update services. This end-to-end capability requires services in each of the device, network and cloud layers. We can expect investments in platform layer common services, bound to backend cloud services, to become a major competitive point of leverage for device providers.

Content is another area of competitive focus. Although none own or operate their own content like creation firms such as film studios, recording labels or game-software companies, Apple offers a full range of music, media and ebooks as do Google and Amazon. Microsoft’s offerings are more fragmented with digital content offered through the Xbox brand, a nascent joint venture with Barnes & Noble around e-books and textbooks and of course Xbox gaming as a major point of differentiation.

12 For more on this, see an excellent paper by Kenney and Pon. Kenney, Martin, and Bryan Pon. "Structuring the Smartphone Industry: Is the Mobile Internet OS Platform the Key?". Journal of Industry, Competition and Trade 11, no. 3 (2011): 239-61.

Multiple business modelsare vying for market

share

The App Store platformlayer service is becoming

the focal point ofcompetition

Content is another area of competitive focus



Arms suppliers The core challenge facing arms suppliers - the providers of equipment for cloud-service providers - is that their functionality is rapidly being commoditized, often with software solutions replacing and exceeding the kind of performance previously only attainable through high-end hardware. There has been a wave of capital spending to build out cloud-computing infrastructure over the past few years, and this is expected to continue for the medium term. This has benefited holders of industrial real estate, particularly in rural areas near sources of cheap power. Other direct beneficiaries have included providers of networking equipment, servers, virtualization and open-source software. The industrialization of datacenters commoditizes foundational technology from the bottom up. The urgency of expanding capacity is partly the result of the explosive growth of internet-connected smartphones and tablets, which gain significant power from linkage to the cloud.

High-end Unix servers have been the first to give way to commodity x86 server farms. Google, which built its datacenters from off-the-shelf parts, is the paradigmatic example. Servers are therefore becoming commodities. Why invest in 10 high-end servers, when you can get the same or better functionality by deploying cloud architecture for 1,000 less powerful machines? Why not just buy however much capacity you need, when you need it, by accessing a cloud provider that uses hundreds of thousands of machines in their cloud datacenters?

Why invest in high-end Cisco or Juniper Network equipment, when the same functionality can be provided through Level 3’s software for a far lower price? Why invest in an Oracle system that contains management logic in the database that prevents you from separating out different functions, when you can put together a database from X, storage from Y, etc?

Cloud-computing infrastructure is disruptive on several levels. Virtualization improves hardware utilization and decouples workloads from physical hardware. Free, open-source software is foundational to the cloud, often supplanting expensive, proprietary software. Even the design of datacenters and servers is being open sourced and shared - Facebook has opened up its designs through Opencompute.org, where its cloud-computing peers and enterprises can freely benefit from best practices.

A notable aspect of the evolution of cloud stacks - the software that enables infrastructure as a service - is that leading technologies are mostly open source. OpenStack (a consortium of roughly 200 vendors spearheaded by RackSpace), CloudStack (purchased by Citrix and open-sourced to the Apache Foundation), Eucalpytus and Amazon’s Elastic Beanstalk are open source. VMware remains the proprietary enterprise standard for cloud infrastructure.

Who does the cloud disrupt? With the advent of cloud computing and its disruptive potential, many of the ways that major IT firms such as Oracle, Cisco and Intel earned their core revenue will be disrupted.

Arms suppliers facecommoditization with

software replacinghardware functions

Servers are becoming commodities

Even the design of datacenters and servers

is being open sourced and shared

It’s notable that leading cloud technologies are

mostly open source



Intel13 Intel’s high-margin and highly profitable market for server processors is currently under serious threat from the ARM processor architecture. With Microsoft’s decision to create a version of Windows designed to run on ARM, the Wintel era may be ending, as an era of ARMDows may be dawning. For years, manufacturing semiconductor chips has been moving towards commodification. The most significant threat to Intel has been that its x86 processor architectures would become commoditized. This was the threat it faced as AMD emerged as a viable competitor.

For Intel, Linux emerged as an additional and even potentially alternative platform to Microsoft. A significant merit was that Linux was potentially controllable by Intel since it was not under the control of any single player. This was not a popular move with Microsoft, which began focusing more and more on ARM.

ARM dominates tablet offerings from almost all device providers, and Intel has not been a powerhouse in this segment of low-margin, low-power, high-volume chips. Intel’s most profitable business comes from its high-end, server-targeted processors, for which it charges a high premium. It leverages volume economics and rode Microsoft’s growth of enterprise-software business as its competitors, such as Sun’s Sparc and IBM Power series, were killed off.

In the server market, the Wintel combination of Intel processors and Microsoft’s Windows Server stack continues to be dominant. IDC reports that in the third quarter of 2012, Intel servers running Microsoft’s Windows server operating system accounted for 51% of overall quarterly factory revenue. The runner up, Linux, was 22%.14

However, the question is what happens when ARM’s efforts to optimize processor design for server applications hits stride - in all likelihood with the Windows Server team engaged in the process. When these efforts mature and are combined with a major shift of enterprise-scale computing towards cloud computing, we are likely to see a new generation of low-power, highly efficient ARM-based servers to power many of the cloud datacenters. In Microsoft’s worldview, they will be running some version of Microsoft’s operating system. This will lead to Intel’s worst nightmare of a squeeze between the low end of high-volume, low-margin ARM-based processors and a very attractive ARM-based value proposition for high-end demand. Traditional middle-market PCs will rapidly get commoditized by more powerful low-end devices, and Intel will be left without the high volumes or margins to continue R&D at the high end.

Intel’s challenge moving forward will be to avoid this worst-case scenario, and investors should have this picture in mind when evaluating the company’s moves in the near future.

13 This section is drawn, with updated data, from Jonathan Murray’s analysis, “Intel’s ARMageddon” http://www.adamalthus.com/blog/2011/01/07/intels-armageddon/ 14 http://www.idc.com/getdoc.jsp?containerId=prUS23808612#.UND_L280V8E

Intel’s high-margin and highly profitable market

processors is currently under serious threat

Low-power, highlyefficient ARM-based

servers could power clouddatacenters

For Intel, Linux emerged as an additional and even

potentially alternative platform to Microsoft

Intel’s challenge moving forward will be to avoid

this worst-case scenario



Cisco The internet was largely built on Cisco routers, and enterprise-scale datacenters deployed high-end, high-performance Cisco routers and switches. However, with the growth of cloud datacenters and network virtualization, it is facing a serious commoditization challenge.

Cisco began facing a major challenge when the US$500-billion-plus datacenters built by the global cloud providers did not deploy premium Cisco network equipment. Google, for example, built its own, using chips from Fulcrum (later bought by Intel) or Broadcom, or bought lower-cost hardware from companies such as Arista Networks. As more firms choose to rely ever more heavily on external cloud-computing providers, Cisco will be left out of the most rapidly growing portion of datacenter traffic. In short, high-end network equipment with intelligence embedded in the hardware, and offering performance for a significant premium, will have far less of a market presence.

Network virtualization is the second major force commodifying Cisco’s premium offerings. Now, just as physical server infrastructure was virtualized into IaaS offerings, physical network infrastructure is being virtualized. With offerings such as Nicira’s Network Virtualization Platform, logical networking is being decoupled from the physical networks. Performance attributes traditionally attained through hardware solutions are increasingly achieved through software - known as Software Defined Networking.

As the CTO of a major lead user put it, ‘Today, I don’t care who makes the hard drives, CPUs and memory that my cloud runs on. Tomorrow, I’m not going to care who makes my routers and switches.’

Oracle The relentless progression of commoditization and unbundling, combined with the emergence of platforms, poses a serious threat to Oracle as well. We are hearing from lead users that a new degree of modularity is now possible - with all the management and control functions in the platform, all they want from components such as storage, networking and databases is an exposed API to allow the platform to control them. They do not want management functions embedded in the database, since everything will be controlled by the platform layer. Once this happens, databases become commoditized just as the other infrastructure elements. Oracle, offering optimized performance from a world of computing resource scarcity, now faces a world in which computing power abundance can be thrown at raw databases. Moreover, its offerings that are bundled, such as Oracle Database and Oracle Financials that cannot be separated and are not offered as distinct products, will be under pressure from modular components.

For all these three firms, the pressure will be coming from lead users that are in intense competition against one another to utilize cloud-computing resources not simply to gain efficiency, but as strategic weapons. We see a sea-change in demand patterns occurring as successful lead users drive all their competitors to adopt similar approaches. The platform-controlled enterprise IT fabric that is emerging commoditizes all the inputs below the platform layer, posing a major disruption to the suppliers - hitherto some of the most successful firms in IT - by offering premium performance in a world of computing power scarcity.

The world of computing has shifted, and the storm is about to hit. For user firms, providers and arms suppliers, only those that can ride the wave or are facing into the storm have a fighting chance.

Cisco faces challenges from cloud datacenters

that forgo premium network equipment

Oracle faces a world in which computing power

abundance can be thrown at raw databases

For all these three firms, the pressure will be

coming from lead users

The world of computing has shifted, and the storm

is about to hit

Important disclosures Blue Books


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 CLSA (which for the purpose of this disclosure includes subsidiaries of CLSA B.V. and Credit Agricole Securities Asia B.V., Tokyo Branch)/Credit Agricole Securities (USA) Inc ("Credit Agricole Securities (USA)")'s policy 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.

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. Unless specified otherwise, CLSA/Credit Agricole Securities (USA)'s did not receive investment banking/non-investment banking income from, and did not manage/co-manage public offering for, the listed company during the past 12 months, and it does not expect to receive investment banking relationship from the listed company within the coming three months. Unless mentioned otherwise, CLSA/Credit Agricole Securities (USA) does not own discloseable position, and does not make market, in the securities.

The analysts included herein hereby certify that the views expressed in this research report accurately reflect their own personal views about the securities and/or the issuers and that unless disclosure otherwise, no part of their compensation was, is, or will be directly or indirectly related to the specific recommendation or views contained in this research report or revenue from investment banking revenues. The analyst/s also states/s and confirm/s that he 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, non-public 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/Credit Agricole Securities (USA) investment rankings: BUY: Total return expected to exceed market return AND provide 20% or greater absolute return; O-PF: Total return expected to be greater than market return but less than 20% absolute return; U-PF: Total return expected to be less than market return but expected to provide a positive absolute return; SELL: Total return expected to be less than market return AND to provide a negative absolute return. For relative performance, we benchmark the 12-



month total return (including dividends) for the stock against the 12-month forecast return (including dividends) for the local market where the stock is traded. 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.

Overall rating distribution for Credit Agricole Securities (USA) Equity Universe: Buy / Outperform: 65%; Underperform / Sell: 35%; Restricted: 0%; data as of 31 December 2012. Investment banking clients as a % of rating category: Buy / Outperform: 75%; Underperform / Sell: 25%; Restricted: 0%; data for 12-month period ending 31 December 2012. 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) Credit Agricole Securities (USA), Compliance Department, 1301 Avenue of the Americas, 15th Floor, New York, New York 10019-6022; and/or (b) CLSA, Group Compliance, 18/F, One Pacific Place, 88 Queensway, Hong Kong.

© 2013 CLSA Asia-Pacific Markets ("CLSA") and/or Credit Agricole Securities (USA) Inc (“CAS”)

This publication/communication is subject to and incorporates the terms and conditions of use set out on the www.clsa.com website. Neither the publication/ communication nor any portion hereof may be reprinted, sold or redistributed without the written consent of CLSA and/or CAS, a broker-dealer registered with the Securities and Exchange Commission of US and an affiliate of CLSA.

CLSA and/or CAS has/have produced this publication/communication for private circulation to professional, institutional and/or wholesale clients only. 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 and/or CAS 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 and/or CAS 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 and/or CAS. 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 1 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 and/or CAS 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 and/or CAS 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 rating, 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 not rated companies.

Subject to any applicable laws and regulations at any given time CLSA, CAS, their respective affiliates or companies or individuals connected with CLSA/CAS 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, CAS 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 report. 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 the CLSA Group only and do not reflect those of Credit Agricole Corporate & Investment Bank and/or its affiliates. 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 analyst(s)) and /or CAS (for research compiled by US analyst(s)) in Australia by CLSA Australia Pty Ltd; in Hong Kong by CLSA Limited; in India by CLSA India Ltd. (Address: 8/F, Dalamal House, Nariman Point, Mumbai 400021. Tel No: +91-22-66505050. SEBI Registration No: BSE Capital Market Segment: INB010826432; BSE F&O Segment: INF010826432; NSE Capital Market Segment: INB230826436; NSE F&O Segment: INF230826436); in Indonesia by PT CLSA Indonesia; in Japan by Credit Agricole Securities Asia B.V., Tokyo Branch, a member of the JSDA licensed to use the "CLSA" logo in Japan; 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; and in Taiwan by CLSA Limited, Taipei Branch.

United States of America: Where any section of the research is compiled by US analyst(s), it is distributed by CAS. 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 U.S. Institutional Investors" as defined in Rule 15a-6 under the Securities and Exchange Act of 1934 and who deal with Credit Agricole Corporate & Investment Bank. 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 CAS.



United Kingdom: Notwithstanding anything to the contrary herein, the following applies where the publication/communication is distributed in and/or into the United Kingdom. This publication/communication is only for distribution and/or is only directed at persons ("permitted recipients") who are (i) persons falling within Article 19 of the Financial Services and Markets Act 2000 (Financial Promotion) Order 2001 (the "FPO") having professional experience in matters relating to investments or high net worth companies, unincorporated associations etc. falling within Article 49 of the FPO, and (ii) where an unregulated collective investment scheme (an "unregulated CIS") is the subject of the publication/communication, also persons of a kind to whom the unregulated CIS may lawfully be promoted by a person authorised under the Financial Services and Markets Act 2000 ("FSMA") by virtue of Section 238(5) of the FSMA. The investments or services to which this publication/communication relates are available only to permitted recipients and persons of any other description should not rely upon it. This publication/ communication may have been produced in circumstances such that it is not appropriate to categorise it as impartial in accordance with the FSA Rules.

Singapore: This publication/communication is distributed for and on behalf of CLSA Limited (for research compiled by non-US analyst(s)) and /or CAS (for research compiled by US analyst(s)) in Singapore through CLSA Singapore Pte Ltd solely to persons who qualify as Institutional, Accredited and Expert Investors only, 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 with regards to an Accredited Investor, Expert Investor or Overseas Investor, sections 25, 27 and 36 of the Financial Adviser Act shall not apply to CLSA Singapore Pte Ltd. Please contact CLSA Singapore Pte Ltd in connection with queries on the report. MICA (P) 138-12-2012

The analysts/contributors to this publication/communication may be employed by a Credit Agricole or a CLSA company 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 services marks of MSCI and its affiliates. The Global Industry Classification Standard (GICS) was developed by and is the exclusive property of Morgan Stanley Capital International Inc. and Standard & Poor's. GICS is a service mark of MSCI and S&P and has been licensed for use by CLSA Asia-Pacific Markets.

EVA® is a registered trademark of Stern, Stewart & Co. "CL" in charts and tables stands for CAS estimates unless otherwise noted in the source.

Research & sales offices www.clsa.com

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

India CLSA India Ltd 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 Two The Enterprise Center 6766 Ayala corner Paseo de Roxas Makati City Tel: (63) 2 860 4000 Fax: (63) 2 860 4051

USA - Boston Credit Agricole Securities (USA) Inc 99 Summer Street Suite 220 Boston, MA 02110 Tel: (1) 617 295 0100 Fax: (1) 617 295 0140

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

Indonesia PT CLSA 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 Credit Agricole Securities (USA) Inc 227 W. Monroe Street Suite 3800 Chicago, IL 60606 Tel: (1) 312 278 3604

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 5666

Japan Credit Agricole Securities Asia BV Tokyo Branch 16/F, Shiodome Sumitomo Building 1-9-2, Higashi-Shimbashi Minato-ku, Tokyo 105-0021 Tel: (81) 3 4580 5533 (General) (81) 3 4580 5171 (Trading) Fax: (81) 3 4580 5896

Taiwan CLSA Limited Taiwan Branch 27/F, 95 Dunhua South Road Section 2 Taipei 10682 Tel: (886) 2 2326 8188 Fax: (886) 2 2326 8166

USA - New York Credit Agricole Securities (USA) Inc 15/F, Credit Agricole Building 1301 Avenue of The Americas 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

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, Lumpini Pathumwan, Bangkok 10330 Tel: (66) 2 257 4600 Fax: (66) 2 253 0532

USA - San Francisco Credit Agricole Securities (USA) Inc Suite 850 50 California Street 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

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) 207 614 7000 Fax: (44) 207 614 7070

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

© 2013 CLSA Asia-Pacific Markets ("CLSA"). Key to CLSA/Credit Agricole Securities investment rankings: BUY: Total return expected to exceed market return AND provide 20% or greater absolute return; O-PF: Total return expected to be greater than market return but less than 20% absolute return; U-PF: Total return expected to be less than market return but expected to provide a positive absolute return; SELL: Total return expected to be less than market return AND to provide a negative absolute return. For relative performance, we benchmark the 12-month total return (including dividends) for the stock against the 12-month forecast return (including dividends) for the local market where the stock is traded. 01/01/2013


CLEANGREEN& TM

At CLSA we supportsustainable development.

We print on paper sourced fromenvironmentally conservativefactories that only use fibres

from plantation forests.Please recycle.

CLSA is certified ISO14001:2004