EFTA00797937.pdf

An (Institutional) Investor's Take on Cryptoassets December 24, 2017 • version 6' John Pfeffer Medium • • Linkedln John Pfeffer is an entrepreneur and investor. In the 2000s, he was a London-based partner at private equity firm Kohlberg Kravis Roberts, and in the 1990s, he was Chairman of the Executive Board of leading French IT company Groupe Allium S.A. Before that, he advised on turnarounds while with McKinsey in Europe and Latin America. IMPORTANT NOTICE: This document is intended for informational purposes only. The views expressed in this document are not, and should not be construed as, investment advice or recommendations. Recipients of this document should do their own due diligence, taking into account their specific financial circumstances, investment objectives and risk tolerance (which are not considered in this document) before investing. This document is not an offer, nor the solicitation of an offer, to buy or sell any of the assets mentioned herein. Amidst the indiscriminate speculation, sensationalist and mostly misguided media coverage and roller-coaster price volatility, this paper sets out to consider cryptoassets from the perspective of a rational, long-term investor. As investors, we look for things that generate sustainable, ideally growing economic rent—an economic surplus that will accrete to us. This paper evaluates the extent to which cryptoassets offer the foregoing. It aims to assess the potential future value of cryptoassets at mature equilibrium,2 on the assumption that they develop successfully and achieve widescale adoption. By design, it does not dwell on the significant risks that a given cryptoasset could fail, for technical, regulatory, political, or other reasons. These risks are very real, and are well documented elsewhere. Temporarily setting them aside allows for an objective analysis of the potential value of different kinds of cryptoassets and their use cases. I write not from the perspective of a trader, but from that of an investor who believes the long term is easier to predict than the short term. The paper thus focuses entirely on long-term equilibrium outcomes and investment strategy rather than short-term price movements. It also assumes the reader has some familiarity with the topic. Blockchain technology has the potential to disrupt a number of industries and to create significant economic surplus. The open-source nature of public blockchain protocols, Earlier versions of this paper were drafted beginning in June 2017. 2 The notion of mature equilibrium as I use it here is admittedly imprecise. Conceptually I mean once the speculative phase has passed and (i) in the case of monetary store of value, once there is a mainstream, institutional view that crypto is a core monetary store of value like gold is today and (ii) in the context of infrastructure and applications, once markets are valuing cryptoassets based on significant realised user penetration. The obvious analogy is the intemet. Internet penetration and internet-enabled businesses are still growing today but growth is slowing. Today, large internet•enabled businesses are valued based on financial ratios such as PEG and EBITDA multiples rather than clicks or eyeballs as was the case in the late 1990s. That's the end point I'm thinking about. For shorthand, let's assume 10 years from now. 1 EFTA00797937 combined with intrinsic mechanisms to break down monopoly effects, mean that the vast majority of this economic surplus will accrue to users. While tens or perhaps hundreds of billions of dollars of value will also likely accrue to the cryptoassets underlying these protocols and therefore to investors in them, this potential value will be fragmented across many different protocols and is generally insufficient in relation to current valuations to offer a long-term investor attractive returns relative to the inherent risks. The one key exception is the potential for a cryptoasset to emerge as a dominant, non-sovereign monetary store of value, which could be worth many trillions of dollars. While also risky, this potential value and the probability that it might develop for the current leading candidate for this use case (Bitcoin) would appear to be sufficiently high to make it rational for many investors to allocate a small portion of their assets to Bitcoin with a long-term investment horizon. We can break cryptographic token use cases into three broad categories: I. Network backbone / Virtual Machine (e.g., Ethereum) 2. Distributed applications (Dapps) 3. Money, and in particular: a. Payments b. Monetary store of value. I will start by looking at the first two use cases from a general perspective and then dive deeper in analysing the largest current example of the first one, Ethereum. I'll then turn to a discussion of the different functions of money, the potential for cryptoassets to perform them and the implications for the value of such cryptoassets, including Bitcoin. The economics and valuation of utility protocols Use cases I and 2 can be grouped into what I call utility protocols. I will start with some general observations on utility protocols and the implications for their network valuation at equilibrium and then specifically consider the network value of Ethereum at mature equilibrium. General observations A blockchain protocol is a database maintained by a decentralised consensus mechanism operated by its nodes. Utility protocol tokens serve to provision scarce network resources: the processing power, memory, and bandwidth necessary for maintaining the blockchain in question. These resources have a real-world cost in terms of energy and the equipment employed, and these costs are borne by the miners who maintain the blockchain by providing computational services. The miners may be remunerated for their service with block rewards, paid in protocol tokens, and/or transaction fees, paid in protocol tokens or some other means of exchange. While protocol developers may claim that tokens are the basis for other kinds of exchange among users and not just a means of allocating and paying for computing resources, it is my argument that, at mature equilibrium, tokens will do no more than allocate computing resource, with the exception of the special case of a cryptoasset that serves as a monetary store of value. A given protocol is analogous to a simplified economy. The GDP of such an economy would be the aggregate cost of the computing resources necessary to maintain the blockchain, based on the quantity of processing power, memory and bandwidth consumed, multiplied by the unit cost of each. The token is typically the currency used to pay for those resources. The total network value is analogous to the money supply M (i.e., all tokens in issuance), where M = PQ/V; PQ (Price x Quantity) is the total cost of the computing resources consumed, V is 2 EFTA00797938 a measure of how frequently a token is used and reused in the system (its velocity, V). The value of a single token is therefore M/T, where T is the total number of tokens. If a given utility protocol does not have a built-in mechanism, such as Ethereum's GASPRICE, to ensure that the cost of using the network does not arbitrarily and sustainably diverge from the underlying cost of the computational resources it consumes, one of three things happens: (a) the token's price trades to a level such that there is no premium cost to using the network (i.e., there is no economic rent); (b) the chain forks into a functionally identical but less rent-seeking chains until any premium usage cost and economic rent on the network declines to a level at which it is no longer worthwhile to arbitrage; (c) the protocol's adoption is temporarily limited to the highest-value use cases until (a) or (b) occurs. In all cases, the equilibrium result must be at or near marginal revenue = marginal cost for the mining industry maintaining the blockchain in question, so that the token's value cannot materially decouple from the underlying computing resource cost. PQ, the cost of computing resources required to maintain a blockchain, is not only low relative to the current network values being attributed to cryptoassets; it is also inflated by the prevalence of proof-of-work consensus mechanisms, which mean that the vast majority of computing resource consumed is make-work. To the extent that new scaling technologies such as proof-of-stake, sharding, Segregated Witness, Lightning, Raiden and Plasma become prevalent, the amount of computing resource consumed may become quite small. Note also that in the context of cryptoassets, V could go very high at equilibrium. Even if a significant portion of a given cryptoasset has a low velocity because it is being hodl'd by speculators or because it is staked by miners under a proof-of-stake consensus mechanism, the circulating portion of the tokens can circulate at the speed of computer processing and bandwidth—i.e., fast and accelerating. The implication is that average velocities can and are likely to be high, regardless of how many tokens are actually actively circulating for utility purposes to allocate network resources.' The combined effect of low and falling PQ and potentially very high V is that the utility value of utility cryptoassets at equilibrium should in fact be relatively low. Clearly, scaling solutions such as proof-of-stake, etc. are bullish for adoption/users but bearish for token value/investors. Even without those technology shifts, the cost of using decentralised protocols is deflationary, since the cost of processing power, storage and bandwidth are deflationary. This is also bullish for adoption and users and bearish for token value and investors?' Whatever scaling solutions are developed, the inherent redundancy of the consensus mechanism means that there may be fewer use cases than many decentralised revolutionaries think in which a decentralised solution displaces a centralised solution. Use cases will be limited to dematerialised networks where the value of decentralisation, censorship-resistance and trustlessness is high enough to justify the inherent inefficiency and redundancy of the consensus mechanism. Is it worth the cost for payments? Yes for some, but not for all. Consider Twitter -- what is the added value to the user of a massively redundant, trustless, 3 Chris Burniske's recent blog post "Cryptoasset Valuations" (https://medium.comftacburniske/cryptoasset- valuations-ac83479ffca7) estimates an average V of 7, after adjusting for hodlers, stakers, etc. This assumption may be optimistic (meaning, it is probably a low value of V to assume at equilibrium and therefore an optimistic number to be using to estimate the potential equilibrium value of a given cryptoasset), but his framework is useful for thinking about the different drivers of V for a given cryptoasset. I have yet to come across any examples of a protocol where I have been persuaded that when all is said and done the underlying scarce resource being provisioned is something other than computing resources, or at least where that is what it will boil down to at competitive equilibrium after competition in mining, forks, etc. Please alert me to any counter examples you have seen or can think of. 3 EFTA00797939 decentralised Twitter? Is that added value enough to offset its inefficiency compared to the incumbent centralised Twitter? Would Token Twitter offer compellingly higher utility compared to centralised Twitter, including enough surplus utility to offset the cost of operating the consensus mechanism? I'm not so sure. People often make the mistake of conflating the monopoly network effects of, say, Facebook to blockchain protocols. This notion is fallacious on several levels: • Blockchain protocols can be forked to a functionally identical blockchain with the same history and users up to that moment if a parent chain persists in being arbitrarily expensive to use (i.e., rent-seeking). Like TCP/IP but unlike Facebook, blockchain protocols are open-source software that anyone can copy or fork freely. A protocol fork is analogous to a team of Facebook developers who decide one Tuesday morning that Zuck is not paying them enough; they could simply flip a switch and use the servers and software that run Facebook to run a new Facebook that is functionally identical, with all the same users and data up to that point. That can, does, and will happen all the time in protocol-land, but would be theft in the context of private companies that own their code, data, intellectual property, etc. Those property rights are why Zuck is rich, and their absence in the protocol economy has profound implications. The ability to fork protocols maximises utility for users but suppresses economic rent for token holders. • When people talk about the potential value of cryptoassets, they often refer to Metcalfe's Law. Metcalfe's Law asserts that network value = - n(n-I), where A is a constant that captures the differences in the economics built into the business model of each network and where n is the number of nodes in the network. It's not enough to focus on n(n-1). You must also consider what 0 is. Wikipedia has a lot of contributors and users but not a lot of monetary value because it doesn't charge users or have advertisers or attract any other sources of revenue apart from donations. Facebook's 0 is higher than Twitter's because its advertising business model is stronger. TCP/IP lacks financial value not only because no one owns it but also because it doesn't have a revenue model. The problem for utility protocols is that the in question is driven by the cost of the computing resources to maintain the network, which is relatively low and deflationary and which must remain low for their adoption to be successful vs. non-distributed technologies. • When thinking about whether a protocol's token can capture and sustain economic rent, what is relevant is whether the mining industry maintaining the protocol's blockchain is competitive, not the stickiness of users. The mining industry supporting any decentralised protocol must be a competitive market; otherwise the protocol isn't decentralised. It is the economic competition amongst miners that will ultimately drive the cost of using the protocol and therefore the value of the token. No mechanisms for monopoly rents there. • Not only must protocols compete against their own potential forks; competition amongst protocols is also fierce. Witness, for example, recent press reports that Kik is considering migrating its token network from the Ethereum backbone to another blockchain because the Ethereum network is becoming too expensive to use.' S https://www.coindesk.com/kik-might-move-its-ico-tokens-to-a-new-blockchain/ 4 EFTA00797940  • The network value of a tokenised version of a dematerialised network business (a social network, Uber, AirBnB, a betting exchange, etc.) will by construction be a small fraction of the enterprise value of its centralised, joint-stock-company equivalent. Holding the number of users constant, you basically take the fully-loaded IT budget (including energy and a capital charge) of those companies (representing PQ) and divide by some (likely high) velocity V. The disruption of traditional networked businesses by decentralised protocol challengers will represent an enormous transfer of utility to users and an enormous destruction of market value. Great for users, the economy and society; bad for investors. The next topic to address is the impact of a move to proof-of-stake mining and of staking models in general on the network values of Ethereum and other protocols. The idea is that miners are compensated for maintaining the network either in a native cryptoasset or another cryptoasset (such as ETH or BTC), in proportion to the amount of the native network cryptoasset that they stake (i.e., effectively put into escrow and at risk of loss if they attempt to validate false transactions and the like). The promoters of this idea hope that it will reduce the actual computing costs of maintaining the network, by eliminating the costly proof-of-work mechanism, while at the same time creating an alchemic virtuous cycle wherein miners buy and lock up significant amounts of the native cryptoasset as an investment conveying them a right to a mining revenue stream, thereby reducing the velocity of the native cryptoasset and causing its value to rise to a level representing some multiple of their mining profits, much as taxi medallions or shares in a company are valued based on the net present value of future cash flows. Let's think through how this plays out. First, before staking is introduced into the equation, we've established that forks and competition in mining and among protocols lead us to an equilibrium outcome where PQ equals the aggregate cost of the computational resources (capital charge on or usage cost of processing and storage hardware, cost of bandwidth and energy) of maintaining the network. Second, recall that the impetus for moving from proof-of-work to proof-of-stake is to reduce the amount of computational resource and energy required to maintain the network by a couple orders of magnitude. That's good for scalability and potential adoption, but also means a commensurate reduction in the PQ of the network. Third, let's layer on the idea that in order to participate in mining and the associated revenues, on top of paying for processing power, storage, bandwidth and energy, you must now bear an additional cost in the form of a capital charge from acquiring and immobilising an amount of the native cryptoasset. This capital charge on immobilised cryptoasset is added to PQ, making the protocol in question more expensive to use than an equivalent utility protocol that doesn't require staking (or where staking is less expensive because the native cryptoasset is cheaper). This system operates a bit like a taxi medallion system: an authority issues a finite number of licenses, and you must buy one from another medallion holder if you want to operate a taxi. The value of the license captures the discounted value of any economic profits that are expected to accrue from operation. Whoever owned the license first is the primary beneficiary of this monopoly, and he receives that value when he sells the license to someone. The buyer of the license does not enjoy any economic rent because he paid the discounted present value of it to the previous license holder, and so on as the license changes hands. Passengers pay higher fares because the taxi driver's capital cost of buying the license must be compensated for, all for the benefit of the first owner of the license. 5 EFTA00797941 Imagine there are several different taxi companies operating that have acquired a number of licenses. Now imagine that a new entrant decides it would like to take market share. In the world of a taxi medallion monopoly created by an issuing public authority, they would have no option other than to buy medallions from other medallion owners. But here is where protocol-land is different from real-world taxi medallion schemes. Protocols are open source software and can be freely forked. In protocol-land, all the upstart taxi company needs to do is to fork the protocol, effectively issuing an identical number of new taxi medallions and reallocating medallions owned by existing large taxi companies to itself and perhaps a few other friends. Because the upstart taxi company didn't have to pay for its taxi medallions, it and the other recipients of the new medallions can charge its passengers lower fares. Passengers thus flock to the upstart company, and the monopoly value embedded in the original taxi medallions vanishes. Everyone in the system except for the large taxi company wins. If necessary, this process can be repeated indefinitely. The result is that the medallions have low values (as would the analogous native cryptoasset).6 Another mechanism for utility protocols is mine and burn. In this system, new coins are minted and allocated to miners based on the network services they provide, and users must buy these coins and burn them to pay for transaction processing. This is a perfectly fine mechanism, but it simply ensures that the network value equals PQ/V, where PQ is the actual fiat cost of maintaining the network and V is the average time from minting to burning. That gets you to the same low equilibrium network value more simply and quickly. Other general observations: • Analysts often use a working capital analogy in order to assess how much of a given cryptoasset a user will stock to facilitate actual use of a given blockchain's utility function. Fair enough, but digging further into that line of thinking, the way optimal inventories of a good are set is based on the relationship of the volume and volatility of demand, optimal order sizes, communication and delivery latency and production times. Since cryptoassets are generally highly divisible and may circulate very fast (as fast as processor speed and bandwidth allow), it would seem to me that a user would, by the same maths as those used to determine optimal inventory quantities, conclude that he needs to hold very little inventory of a given cryptoasset. Friction moving among cryptoassets is already low and will quickly disappear entirely with technologies like atomic swaps. Consequently, one would expect velocity to be very high at equilibrium. It would make no more sense for users to hoard utility cryptoassets beyond the minimum they need to carry out their desired operations than it would be for individuals to hoard petrol or for companies to hoard giant warehouses full of whatever goods they sell. Companies need inventories of goods to run a business and those inventories have a value on their balance sheet, but they try to minimise such holdings, as they are unproductive assets that are costly to finance and carry. They certainly don't try to accumulate more inventory than necessary as a way to store their retained earnings. Similarly, individuals have petrol in the tanks of their cars, but they don't stockpile petrol in their basements as a form of savings.' 6 The competitive forces to eliminate economic rent would function in largely the same way whether the staking system involves payment for services in cryptoassets that are native or external to the protocol at hand. 7 See also Vitalik Buterin's recent blog post: "On Medium of Exchange Token Valuations" (http://vitalik.ca/general/2017/10/17/moe.htmil 6 EFTA00797942  • For every successful utility protocol (certainly for every successful Dapp), there will ben failed versions. In fact, one of the advantages of the protocol economy is that it facilitates open and inexpensive experimentation, which will mean that there will be many more attempts and many more failures, and that each success will be individually smaller in its value and reach. The open-source, forkable nature of this kind of software will likely drive toward a fragmentation of use cases and protocol functionality; businesses built on top of the protocols will be protocol agnostic and capable of using and combining modularly a changing array of protocols to deliver whatever service or value chain they are trying to deliver. These dynamics are great for users and generate lots of positive economic and social externalities, but they are bad dynamics for investors.8 The problem of making money by investing in utility protocols is aggravated by: (a) the fact that this is a fragmented space with very high failure rates, so selecting winners a priori will be very difficult; and (b) the fact that most of the long-term winning protocols probably haven't even been launched yet (witness the fact that the most valuable internet businesses were founded after 2001). • Developer incentives over time are a fundamental issue in crypto. For most protocols, such incentives are heavily front-ended around launch and insufficiently provided for over time. The more ambitious and long-term a protocol's development roadmap is, the more problematic this failure of incentives becomes. The incentives to improve an existing protocol by forking it may be strong if some tokens are reallocated at the fork to the developers making the improvements. For example, where tokens have been retained by a foundation linked to the original protocol developers, an aggressive group of forking developers could reallocate the foundation's tokens to their own new entity in their fork, leaving all other users in the same position and letting the market decide which fork to support. The incentives for a developer to create a new, competing protocol are also strong, but network effects do make it harder to displace an existing protocol than to improve or fork it. Miners and perhaps large users have a strong economic incentive to invest in development of the protocol they are mining either through changes to the protocol or by forking it. The foregoing suggests that we're likely to see (a) more success with protocols focused on simple use-cases that require less ambitious future development; (b) future protocols launched with better long-term developer incentive schemes (easier said than done)9; (c) aggressive forks that transfer value from incumbent to challenger developers1); and (d) large miners/users or groups of miners/users acting together employing or paying developers to improve legacy protocols either directly or via forks. The implication of this section is not that utility protocols won't have any network value. PQN does represent positive value. The implication is that network value of a utility See also Teemu Paivenen's blog post "Thin Protocols" (httos://blotzeppelin.solutions/thin-protocols- cc872258379f). g Tezos proposes an interesting potential solution to the developer incentive problem. Tezos combines a PoS consensus mechanism with a system whereby token-holders can vote on improvements to the protocol proposed by developers and reward the developers for their contribution. We'll see if it works, but the problem for Tezos remains that the mature equilibrium value of the Tezos token will be Trews = PQ/VM where PQ is the cost of the computing resource maintaining the Tezos blockchain, i.e., Tiezos probably won't have a high value when the dust settles. 1D See also Fred Ersham's blogpost "Accelerating Evolution through Forking" (https://medium.com/OFEhrsam/accelerating-evolution-through-forking-6b0bba8Sa2ba) 7 EFTA00797943 protocol will converge on or near an equilibrium, where it is a fraction (denominator V) of the actual cost of the computing resources consumed to maintain the networks. For a fork to succeed, there needs to be enough value available to arbitrage to incentivise users, some miners and a sufficiently credible developer group to support the fork. It should therefore be acknowledged that, to the extent the equilibrium outcome is arrived by way of one or more forks, there could be a sustainable level of network value economic rent premium above computing cost that is too small to provide adequate incentives for a fork to succeed. I would not, however, consider it to be a very compelling investment thesis when the best I can hope for is to keep an amount of value corresponding to an economic rent that's too small for anyone to bother arbitraging it away from me despite relatively low barriers to doing so. While a protocol's core development team may be bound by various soft ties, in protocol-land (unlike in a traditional software business), the work product is all open-source; intellectual property isn't generally owned or protected; and developers have little or nothing in the way of contractual ties or limitations (e.g., no non-compete, no non-disclosure, no non- solicit). That means developers can defect or take the work of others. At a minimum, these factors place a low ceiling on how much economic rent can be created and sustained." As illustrated in the ETH valuation example to follow, it is likely that the combined network values of all utility protocol cryptoassets together will total between tens of billions and hundreds of billions of dollars. That is significant value, but not when compared to the current -$250 billion combined network value of protocols other than Bitcoin. Investing in utility protocol cryptoassets could make sense if their current network values were one or two orders of magnitude lower than they currently are, but at current valuations, the risk/return to investors is not attractive. The Network Value of ETH ETH, the Ethereum token, is an interesting case to explore because of its significant current network value and Ethereum's potential as the ultimate utility protocol. Ethereum could serve as the backbone for processing smart contract operations for (hopefully) untold numbers of decentralised applications, DAOs, etc., and perhaps one day maybe even something like the fabled Ethereum Virtual Machine (EVM). Ethereum's developers understood that for Ethereum to fulfil its potential, the cost of using it as a smart-contract-executing utility must be as low as possible and must not depart at equilibrium from the actual cost of the computational resources consumed. To ensure this will be the case, they built the GAS mechanism into Ethereum to decouple the use of the network (and the cost thereof) from the value of the ETH token. Each possible type of computing operation has a pre-defined GASCOST, measured in units of GAS. GAS may then be paid for using ETH (or another token or currency) based on the GASPRICE `exchange rate', which is freely set among users and miners.12 11 An interesting business idea that someone could logically pursue at some point would be to raise capital to fund a crack team of mercenary blockchain developers and systematically target technically-mature or maturing protocols where there is still a significant economic rent premium and arbitrage that value via hostile forks of those protocols in a way that reduces cost and/or improves functionality to users and reassigns network tokens held by the incumbent developer team and backers to the insurgent team and backers. 12 Note that because GASPRICE is fully-flexible, GASCOST might only need to be updated in the system from time to time if and to the extent the relative cost of certain sub-components of computing costs changes, for example the cost of processing power vs storage. 8 EFTA00797944 The Ethereum Homestead Documentation makes this all clear: "Gas Price is how much Gas costs in terms of another currency or token like Ether. To stabilise the value of gas, the Gas Price is a floating value such that if the cost of tokens or currency fluctuates, the Gas Price changes to keep the same real value. The Gas Price is set by the equilibrium price of how much users are willing to spend, and how much processing nodes are willing to accept '3." (Ethereum Homestead Documentation Release 0.1, p49) "Gas and ether are decoupled deliberately since units of gas align with computation units having a natural cost, while the price of ether generally fluctuates as a result of market forces. The two are mediated by a free market: the price of gas is actually decided by the miners, who can refuse to process a transaction with a lower gas price than their minimum limit." (Ethereum Homestead Documentation Release 0.1, p68) This is all logical in the sense that GAS, and by extension the ETH token itself, is a metering device meant to ensure correct economic allocation and remuneration of the network's resources. In the long term, the GASPRICE (and through it the value of ETH) should therefore tend toward the actual marginal cost of computing resource on the network. It could not possibly be otherwise, since if the cost of running operations on the Ethereum blockchain became materially more expensive than the actual underlying cost of computing resources consumed by it, people would simply use another blockchain where that premium doesn't exist (or fork to create a cheaper Ethereum network that has identical functionality and users at that moment)? Also, if the GASPRICE were to decouple sustainably from the actual computing cost of operations, then mining would be the only perfectly competitive industry in history to earn sustainably positive economic rent. There is no reason for this to be the case in an industry where capacity can be freely added and withdrawn and the market price freely set. Since the value of ETH is decoupled from GAS and therefore from the volume of transactions on the Ethereum protocol, an ETH bull could argue that ETH tokens could have an arbitrarily high value without compromising the cost-efficiency of operations on the chain. But let's first agree that because of the GASPRICE mechanism14 the volume of transactions on the ETH blockchain and the scale of its adoption are not transitive to a high ETH token value. This point is important as observers often erroneously assume that a high volume of network transaction volume driven by all of the different potential uses of the Ethereum protocol will necessarily give the ETH token high value. Let's work through some numbers to see what in fact the utility value of ETH might be. Ethereum GDP (i.e., PQ) is the total `revenue' of the computing network performing the 13 Today in practice it seems that the vast majority of transactions use the default 0.02 microETH price, but that most likely reflects the incipient nature of activity on the network. GASPRICE can be expected to become more market-driven as use of the Ethereum network grows. From a basic microeconomic perspective, if the GASPRICE (in fiat terms converted via the GASPRICE to ETH exchange rate and the fiat value of ETH) exceeds from time to time the actual fiat cost of providing the requisite computing resources, you would expect users to reduce GASPRICE offered or miners to add competing computing resources to the network until the marginal cost again equals the marginal revenue, driving a decline in the GASPRICE. This relationship should hold no matter what the scale of the operations being performed on the blockchain. The market will just keep allocating more computing and storage resources to the network as long as it is profitable to do so. 14 Note that the GASPRICE mechanism helps to reduce the incentives to fork the chain because economic rent can be eliminated quickly through it without necessitating a fork. Protocols without a GAS mechanism can be expected to end up at a similar economic equilibrium through forks as Ethereum will reach through the GAS mechanism. Ethereum may still fork for other reasons. 9 EFTA00797945 underlying operations, which can be directly measured as GAS used multiplied by the average GASPRICE. On 23 December 2017, the total amount of ETH used to fuel (pay miners for) transactions on the Ethereum network was ETH 1,388 (derived from the total GAS used's multiplied by the average GASPRICE that day16)17. ETH 1,388 is worth about $1 million at $700 per ETH. Annualised (simplistically multiplying by 365), this is about $355m per year. We can then play with different assumptions for how fast the Ethereum network will grow vs the declining computing and energy costs. For example, let's assume Ethereum network traffic grows from here at the same rate internet traffic grew from 1995 to 2005 (roughly 150% growth per year)'8 and that the combined offsetting impact of declining computing costs is -20% per year (optimistic as this approximates only the effects of the average rate of decline in computing costs without a change in the consensus mechanism; implementation of proof-of-stake or other scaling solutions could represent a step change down in the computing costs of the network by orders of magnitude). The combined net effect would imply `Ethereum GDP' (PQ) doubles each year. At this rate, Ethereum GDP would grow from $355 million to $363 billion in ten years, an over thousand-fold increase. If we assume an ETH velocity of 7, the network value of ETH would be $52 billion in 10 years, about 24% less than its current network value of approximately $68 billion. Of course, in order to provide an attractive return to investors buying ETH today, its current network value would have to be significantly lower than $52 billion (assuming investors would expect to make a 30 - 40% annual rate of return over that period, the current network value would need to be in the range of $1.8 - 3.8 billion). The foregoing calculation implicitly assumes that GASPRICE is already set at the level where miners are making zero economic rent and that Ethereum does not change its proof-of- work system, for example to proof-of-stake. As it's early days for Ethereum and mining computing resources are still catching up with demand, miners are probably still temporarily making positive economic rent, which means this back-of-the-envelope calculation in fact overstates PQ even if proof-of-work is maintained. More significantly, if Ethereum successfully moves to a proof-of-stake mining system and thereby substantially reduces the computational inefficiency inherent in proof-of-work where 99% of the computing power goes to proof-of-work and only a very small portion to actually maintaining the ledger, the PQ of the blockchain would fall massively and along with it the Ethereum network value. Recall also the analysis in the previous section explaining why staking of tokens for mining under proof-of-stake won't allow Ethereum to sustain a network monopoly premium. Another way to look at this is to relate the Ethereum GDP to the total revenue of Amazon Web Services. AWS total revenue in 2017 is estimated to be $16.8b, growing to $40b in 2021 (according to JP Morgan), an order of magnitude smaller than our 10-year estimation for ETH GDP in the previous paragraph. If the velocity of ETH is 7, the Ethereum GDP (PQ of computational resources running the network) would need to reach approximately $476 billion or 28 times AWS' current revenue to justify its current network value and excluding any return on investment during the years while Ethereum grows to reach that scale. Now, of course, AWS is just one provider of cloud services, but Ethereum is just one blockchain. Even if we assume that Ethereum will have some greater market share of blockchain than https://etherscan.io/chart/gasused https://etherscan.io/chart/gasorice 17 On 23 December 2017: GAS Used 41,686.74 million x Average GASPRICE 0.000000033285710975 ETH = 1,388 ETH. https://blogs.cisco.com/so/the-historv-and-future-of-internet-traffic 10 EFTA00797946 AWS has of cloud, it is still hard to see how the current Ethereum network value can be remotely justified on this basis. Note that in my reasoning about the future value of ETH at equilibrium, I have so far not taken into account the mined tokens that miners receive for performing computational services for the network as that value does not accrue to token holders. Rather the opposite. There are in fact two negative impacts of mining rewards on the value per token: • Issuance of new tokens doesn't increase the total network value, just as printing fiat money doesn't make people collectively richer in real terms. The new issuance goes to the miners at a one-for-one cost of dilution spread across the value of all pre-existing tokens. This must also be true for the interest rate (BIR) paid on ETH tokens deposited in a proof-of-stake system. The new tokens generated to pay the interest dilute all existing tokens such that the effect on the overall total value of ETH tokens is neutral. Those engaged in mining will benefit from the interest earned while those not engaged in mining will suffer from the corresponding dilution. But the existence of this system does not drive growth in the network value of ETH and in fact drives devaluation of each ETH token at the rate of total interest paid in ETH divided by the total issuance of ETH.19 • There is a second, subtler negative effect of this new token issuance. It subsidises the cost of operating the network, which at competitive equilibrium puts downward pressure on GASPRICE, which in turn puts downward pressure on the value of ETH at a constant GAS <> ETH exchange rate. The paradoxical combined effect is that the cost of new token issuance through mining rewards is effectively borne twice by non-miner token holders. The implication of all of the foregoing is that, even if Ethereum is hugely successful, the value implied by its use as a backbone utility protocol is likely a small fraction of its current value. All of this raises a question for ETH bulls: why would ETH be arbitrarily valuable if it's not some scarcity in relation to the volume of transactions and operations on the chain? One proposed reason has been that people will hoard ETH as a currency with which to make financial investments, for example in ERC20 token ICOs or DAOs built on the Ethereum protocol. In his blog post "Platform Currencies May Soon Be Obsolete" 20, Aleksandr Bulkin articulates why it is unlikely that a single blockchain will host a large number of Dapps and at the same time function as a major monetary store of value. Also, if utility protocols turn out to be poor financial investments as the foregoing analysis suggests, how much investment demand will there be? Finally, in a frictionless, multi-protocol future, why stockpile a particular token specifically to make a particular type of investment rather than store your value in the best pure store of value protocol (or in productive investment assets) and acquire the amount of ETH or any other currency for a particular purpose (including a subset of investment purposes) at the time of need? So that leaves the possibility that ETH replaces Bitcoin and becomes the dominant non- sovereign monetary store of value simply on pure store-of-value merits. We'll go deeper into the topic of monetary store of value below, but from where we are today, an objective observer would give Bitcoin significantly higher odds than ETH of becoming such a store of value. And as for those who argue that you can recreate Bitcoin on top of Ethereum, the 19 Vitalik Buterin Incentives in Casper the Friendly Finality Gadget (v 27 August 20171, p6. Aleksandr Bulkin https://blog.coinfund.io/platform-currencies-may-soon-be-obsolete-78d9b263d902. 11 EFTA00797947 question is, why would you? Why substitute a new sub-token on top of a more complex protocol with a larger attack surface, shorter track record, less decentralised governance and propensity to make backwards incompatible protocol changes, for a hugely robust, stable, proven, and widely accepted protocol that already performs that narrow function very well? Cryptoassets as Money Money is a debt ledger with three sub-functions: I. Store of value 2. Means of payment 3. Unit of account. Cryptocurrency's performance advantage over incumbent forms of money is (a) strongest and most obvious as a monetary store of value; (b) stronger for some, but far from all, payments; and (c) differentiated as a unit of account for a few select purposes. Cryptocurrency is overwhelmingly better as a monetary store of value than, say, gold. (I won't enumerate the reasons why, as it's pretty intuitive and has been written about widely.) As a means of payment, it can perform better than incumbent technologies in specific instances (think international payments), but Visa, Apple Pay, Google Pay, PayPal and fiat currency work well and better than cryptocurrency for most day-to-day payments. As a unit of account, a non-sovereign cryptocurrency could be most useful in international trade, global commodity markets, foreign reserves, and jurisdictions with unstable domestic currencies. Before addressing the question of how to think about valuing the payment and the monetary store of value functions of a cryptocurrency, I'll first examine the link between payments and monetary store of value. Many observers presume this link to be very strong, but the reality is more nuanced. First, let me draw a distinction between a monetary store of value and a run-of-the-mill asset. A monetary store of value is characterised by having a value that is decoupled from its utility for other purposes and from the cost of making/extracting and storing it. A warehouse full of goods, a stockpile of copper and a tankful of petrol are all assets and have value (determined by the market at the equilibrium point where their marginal utility meets their marginal cost of manufacture/extraction, i.e., MR = MC). Inventories of assets such as these appear on a company's balance sheet, but companies seek to minimise how much they have to hold to carry out their business, given the capital carrying cost. They don't try to accumulate these inventories to store their retained earnings. Gold, by contrast, is a monetary store of value. Its value is decoupled at equilibrium from the cost of extracting and storing it. While we may also use it for jewellery (an ancient way of signalling our wealth to other members of society), and we use a bit of it for manufacturing electronic goods and other industrial uses, we also store tonnes of it at great expense in giant inert lumps as a form of savings-a store of value—with no intent of ever using those lumps for any other purpose. Gold is therefore arbitrarily expensive relative to its extraction and storage cost. Its value is subjective. Consider some examples of the things we use as means of payment versus those we use as monetary stores of value today21: 31 Note that I exclude here things like pre-paid debit cards, gift cards, pre-paid telephone plans and air miles as they are relatively immaterial to the financial system. As it happens, these can all be used for payments and are assets but people treat them as working capital (immobilised balance sheet assets with a carrying cost) rather than a form of savings, so if anything, they are more payment rails than monetary stores of value. 12 EFTA00797948  • Means of payment: Visa (credit and debit), SWIFT, PayPal, Apple Pay, Google Pay, Western Union, physical cash • Monetary stores of value: Gold, fixed and demand bank deposits, physical cash. What's interesting is that the only thing that appears as both a means of payment and a monetary store of value is physical cash. Yet even though physical cash is c