The Past, Present, and Future of Data Deduplication Toward Massive, Ultrareliable, and Low-Latency Wireless Communication With Short PacketsA Survey on Wireless Security

Point of View: Minting Money with Megawatts

Scanning Our Past: Light and Water in the 1929 Barcelona Exhibition

The Past, Present, and Future of Data DeduplicationToward Massive, Ultrareliable, and Low-Latency Wireless Communication With Short PacketsA Survey on Wireless Security

Point of View: Minting Money with Megawatts

Scanning Our Past: Light and Water in the 1929 Barcelona Exhibition

September 2016 | Volume 104 | Number 9

P O I N T O F V I E W

Minting Money WithMegawattsBY SVE INN VALFELLSFlux, Ltd.

JÓN HELGI EG ILSSONFaculty of Economics,University of Iceland

I . INTRODUCTION

Launched in January of 2009, Bitcoin is a rapidly growing peer-to-peer (P2P)online payment network with an annual transaction volume of $45.4 billion.The market cap of the bitcoin digital currency (BTC) is currently $10.2 bil-lion.1 For comparison, PayPal, which was founded nine years before Bitcoin

was launched, had payment volumeof $282 billion in 2015 and a marketcap of $45 billion at the end of lastyear. PayPal’s volume is still growingat an impressive 27% per annum,but the value and number of bitcointransactions have, on average, morethan doubled every year since incep-tion, a strong indication that thisnew payment network is on a fasttrajectory toward mainstream adop-tion (Fig. 1).

Unlike PayPal and other tradi-tional payment services, however,Bitcoin payments do not go throughfinancial institutions. Instead, theopen architecture first describedby the unknown founder, SatoshiNakamoto, allows users and otherparticipants to join the Bitcoinpayment network freely withoutgoing through traditional banks orpayment providers [1]. Through ajudicious selection of technologiesand incentives, Satoshi’s architec-ture has given rise to a newpayment network and digital cur-rency independent of existing finan-cial service providers and centralbanks.

II . BITCOIN MINING

A crucial part of this new financialnetwork is a computational processknown as mining. The Bitcointransaction ledger is distributed,and there are no central copiesDigital Object Identifier: 10.1109/JPROC.2016.2594558

1When capitalized, Bitcoin refers to the entire Bitcoin network and protocol; when not cap-italized, bitcoin refers to the digital currency used on the Bitcoin network. The currency is alsooften abbreviated as BTC or XBT.

0018-9219 Ó 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

1674 Proceedings of the IEEE | Vol. 104, No. 9, September 2016

maintained by trusted institutions.The ledger consists of a chain oftimestamped transaction blocks, theblockchain. The integrity of the block-chain is secured by cryptographichashes of the transaction blocks witheach block referencing the hash of theprevious block. Miners compete tocalculate or “mine” valid hashes andare rewarded with new bitcoins andtransaction fees. The miners consumehundreds of megawatts of powerminting bitcoins worth billions ofdollars.

In technical terms, the crypto-graphic hash which Bitcoin uses is aone-way transformation of an inputstring, the transaction block header,into a highly garbled output string.The block header consists of severaldata fields, including a signature ofthe enclosed transactions, the hashof the previous block, a timestamp,and a random data field called“nonce.” A valid output of the blockheader hash must belong to the sub-set of outputs which have a certainnumber of leading zeros. More zerosindicate a smaller subset and greaterdifficulty of finding a valid hash.The hash is computed repeatedlyvarying the nonce until a nonce isdiscovered which produces a validblock header hash. The target

difficulty is adjusted every 2016blocks to make new blocks appearevery 10 min, on average.

The block header hashes serve asdigital watermarks which are easy toauthenticate but difficult to reproduce.They remove the need for a central au-thority to maintain an official copy ofthe blockchain. The miners competein computing the hashes. Successfulminers claim the new block rewardwhich now stands at 12.5 BTC foreach block. User transaction fees areelective and currently average lessthan 1 BTC per block.

Miners’ annual revenues havereached over $545 million. Fiercecompetition has ensued, with minersmoving from central and graphicalprocessing units in desktops andservers to application-specific proces-sors and systems deployed in customdata centers. The largest five miningpools and companies control over75% of the total mining capacitywhich now stands at 1517 petahashper second (Fig. 2). The networkhashrate corresponds to power con-sumption of 160 MW if mined withthe latest processors and data cen-ters, but the actual consumption isprobably somewhat higher becauseof older systems still in use. The esti-mated replacement cost of the net-work is $800 million (Table 1).

Each miner’s share of total min-ing revenues is diluted when newcapacity is added. Consequently,miner profits drop as the networkgrows. Additionally, the new blockreward halves every 210000 blocks,or approximately every four years.Because transaction fees are cur-rently about 50 times smaller thanthe block reward, halving the blockreward effectively halves total min-ing revenues if bitcoin price or trans-action fees remain unchanged.

2Based on data from blockchain.info,Jul. 2016. All calculations performed by theauthors.

Fig. 1. Bitcoin market cap and trailing 365 day turnover (in billions of dollars).2

Fig. 2. The total hashrate of Bitcoin mining (in gigahash per second) has increased with

mining revenues (in thousands of dollars). The current hashrate is 1517 petahash per

second (peta denotes 1015).

Vol. 104, No. 9, September 2016 | Proceedings of the IEEE 1675

Point of View

Several miners have run into dif-ficulties, for example KnCMinerwhich declared bankruptcy in Mayciting the drop in transaction blockreward from 25 BTC to 12.5 BTCwhich took place on July 9. Capitalcosts of the latest systems and facili-ties are high which presents a barrierto entry to new miners. Consolida-tion in mining raises concerns aboutthe integrity of Bitcoin: the greaterthe concentration of mining capac-ity, the greater opportunity theminers have of colluding to attackthe blockchain. If any entity orgroup were to gain control over 51%or more of the mining capacity, theywould effectively control the transac-tion history of the blockchain, andthe decentralized, trustless quality ofBitcoin would be destroyed.

III . HOW TO MINEPROFITABLY

The evolution of mining is deter-mined by economic incentives andthe cost and performance of technol-ogy. In order to analyze the marketstructure we construct a simple eco-nomic model which captures the keyfactors influencing miners’ profit andloss: revenues, operational costs, andinvestment costs [see (1) in box].

As shown in Fig. 3, under theright circumstances new capacity canbe profitably added to the miningnetwork. Given market data, it ispossible to determine whether newentrants can profitably add capacity,how much, and the minimum cost ofprofitable entry. The shortest pay-back period can also be computed.

Using recent cost and performancenumbers of mining systems and de-ployment environments (Table 1), wecan map out the breakeven zonewithin which new miners can profit-ably add new capacity (Fig. 4). Wechoose an amortization period ofthree years which reflects the fact

3Based on information from miners, colo-cation centers, and power utilities.

Fig. 3.Miners compete to collect new supply and transaction fees. In dollar terms, their

total revenues are BðSþ FÞ where B is bitcoin price, S is supply, and F is transaction fees.

With total hashrate at h0, new capacity added must generate enough profit to cover

operational and investment costs within the amortization period T. For the right

combination of revenues, costs, performance parameters, and amortization period,

capacity can be profitably added within a range bounded by the lower and upper

breakeven points hlowerBE and hupper

BE . The point of maximum profitability h$ lies between the

breakeven points. For hashrates greater than hCAP, operational costs exceed revenues,

and miners shut off their equipment.

BITCOIN MINING PROFIT FUNCTIONThe profit !ðXÞ generated by new Bitcoin mining capacity X is definedas

!ðXÞ ¼ X

h0 þ XBðSþ FÞ & XC & 1

T

X

zþ NRE

! ": (1)

Here h0 is the preexisting hashrate, B is bitcoin price, S is supply, F istransaction fees, C is the operational cost, T is the amortization period,ð1=zÞ is the variable investment cost, and NRE equals the nonrecurringengineering cost of investment. Under the right conditions, capacity canbe profitably added within a range defined by the two breakeven points(Fig. 3). A detailed discussion of the profit function is available onGithub, including derivations of the breakeven points, the point of maxi-mum profitability, and the payback period [2].

Table 1 Price and Performance Data of Recent Mining Systems and Deployment Environments3

1676 Proceedings of the IEEE | Vol. 104, No. 9, September 2016

Point of View

that Moore’s Law doubles the peakoutput efficiency of processorsroughly every three years [3]. Withthe previous block reward of 25 BTCand at current network hashrate, thebreakeven zone started at just over$300 and fanned out with increasingprice. At current price of $648 andhashrate of 1517 petahash, there was

room for new entrants to more thandouble the hashrate. The smallestamount of capacity that could beadded profitably was just under 10petahash. However, when the blockreward halved to 12.5 BTC, the low-est price point at which capacity canbe profitably added jumped to over$600. Furthermore, the minimum

amount of capacity which can beadded profitably has risen to over60 petahash, and the maximumcapacity addition has declined to300 petahash, around 20% of theexisting hashrate. A significant barrierhas developed for new entrants, andthe range of profitable entry has con-tracted considerably.

Fig. 4. In the previous reward era of 25 BTC for each block mined, new miners could profitably add capacity in the breakeven zone

between hlowerBE and hupper

BE at bitcoin prices as low as $300. When the block reward halved to 12.5 BTC, the lowest price point of

profitable entry jumped to over $600, and the breakeven zone narrowed considerably. Existing miners will keep their systems

running for all values below hCAP. At any given price, h$ represents the most profitable capacity addition.

Fig. 5.Moore’s Law partially counteracts the reduced block reward. It expands the breakeven zone for new entrants and reduces the

lowest price of profitable entry from $610 to $530.

Vol. 104, No. 9, September 2016 | Proceedings of the IEEE 1677

Point of View

Unlike new miners, existingminers with installed capacity arenot constrained by the breakevenzone. Having already incurred in-vestment costs, they will continue tooperate their capacity until their op-erational costs exceed revenues, athCAP. As shown in Fig. 4, hCAP ismuch higher than the current net-work hashrate, and existing minershave no reason to turn off theirequipment even after the blockreward has halved to 12.5 BTC.Because hCAP far exceeds the valuesof the breakeven zone, incumbentsenjoy a clear advantage over newentrants.

While processors continue to ad-vance according to Moore’s Law,however, power efficiency will con-tinue to double approximately everythree years. As Fig. 5 shows, a newgeneration of processors with im-proved efficiency can expand thebreakeven zone and bump the lowestprofitable price point of new en-trants from $610 back to $530,counteracting, in part, the halving of

the block reward. As the power effi-ciency doubles, the smallest amountof capacity that can be added profit-ably at the current price and hash-rate declines to 30 petahash fromthe current value of 60 petahash,and the maximum profitable capacityaddition increases from 300 petahashto 600 petahash.

Finally, we note that halving ofthe block reward has more than dou-bled the shortest possible payback pe-riod of new entrants from 1.2 yearsto 2.8 years (Fig. 6). With Moore’sLaw, the posthalving payback periodcan reduce back to 2.4 years.

IV. CONCLUSION

Following the recent halving of theBitcoin block reward, new minersare close to being shut out from Bit-coin mining and unable to add newcapacity profitably. The ability of in-cumbents to continue mining is notsubstantially affected, however. Arise in bitcoin price, higher transac-tion fees, or a reduction in existing

mining capacity may improve the op-portunity for new miners to enterthe market, but each future halvingof the Bitcoin block reward will nar-row the range of profitable entry fornew Bitcoin miners considerably.The stronger relative position of in-cumbents compared to that of newentrants supports further consolida-tion of Bitcoin mining. Conse-quently, the distributed Bitcoinledger, the blockchain, may becomemore vulnerable to attack, and itstrustless nature may ultimately evenbe destroyed.

Continued improvement in mi-croprocessor efficiency will counter-act mining consolidation. WhileMoore’s Law remains valid, new gen-erations of processors will from timeto time open a window of entry intothe mining marketplace. Ultimately,the greatest disruption to the currentmarket structure may come throughadoption of new computing technol-ogies, such as graphene processors orquantum computers. h

REFERENCES

[1] S. Nakamoto, “Bitcoin: A peer-to-peerelectronic cash system,” 2008.

[2] S. Valfells and J. H. Egilsson, “Bitcoinmining profitability,” [Online]. Available:http://www.github/sweyn

[3] J. Koomey and S. Naffziger, “Moore’s Lawmight be slowing down, but not energyefficiency,” IEEE Spectrum, Mar. 31 2015.

Fig. 6. The shortest profitable payback period of new Bitcoin mining capacity increased when the block reward halved, but Moore’s

Law power efficiency gains will reduce the payback period and partially reverse the effect of the halved block reward.

1678 Proceedings of the IEEE | Vol. 104, No. 9, September 2016

Point of View