The State of Crypto Valuations

December 17, 2018

There has already been a bit of ink spilt on crypto-currency valuations and it’s something that I’ve spent some time thinking about. Some of the methods that have been proposed are reasonable as relative valuation models and I’m intrigued by the monetary equation exchange method that was proposed by Chris Burniske and Brett Winton. The problem I have with the current approaches is that, at the moment, they aren’t rooted in first principles and are reliant on inputs which are little more than guess work. While I think that we’ve started to lay a solid foundation based on traditional equity valuation methods as approach the valuation of crypto-currencies generally, it’s likely we still have a way to go.

When I refer to crypto-currencies in this piece, I’m referring specifically to utility tokens. Not security tokens or monetary tokens (aka stable coins). I’m referring to the tokens that are asking you to buy a ticket that will give you a chance to take a ride on an as-yet undeveloped ride at an as-yet undeveloped theme park. While there have been a lot of spams and scam in the space, I believe that there will be a massive amount of creativity and ingenuity (other than in the financial engineering space) that will be unleashed as a result of crypto-currencies as frameworks are created that incentivise developers to create new protocols, new ways of communicating and new ways of doing things on the internet.

Investing in these protocols at this stage is probably going to be similar to investing in start-ups prior to the turn of the century. The one key difference, in my mind, is that then we had relatively few venture capitalists or funds focused on investing in start-ups. Previously, the venture capitalists were focused on investing in projects that would take an immense amount of capital and time to build out, like a semi-conductor foundry. However, today we have numerous experts and everything from hedge funds to venture capital funds that are investing in the space. There are a lot more people that are willing to take a risk on something that’s unknown and a surprisingly small handful of people that are willing to take a look at how we can develop the tools to look at the emerging asset class.

Let’s look at some of the models that have been developed so far. I’m doing this as much to make a record (so people can point out if I’m missing any methods that have been developed) as I am to see where the short-comings or assumptions are in the approaches that have been developed to date. I don’t mean to be completely critical of the approaches because they are somewhat novel and we need more people thinking about how we can approach valuations as applied to crypto-currencies. My fear is that we don’t have the tools available to us yet. We need the industry to develop further and see some of the projects and use cases reach scale to get an idea of what they could potentially be worth.

Quantity Theory of Money

Firstly, the quantity theory of money, where MV = PQ

M = The nominal money supply in circulation or average supply of coins,

V = Velocity of money or the amount of times a coin circulates in a given year,

P = Price level of the resource being provisioned,

Q = Real expenditures or the quantity of the digital resource being provisioned.

To go into the weeds, you can read Burniske’s article here. To give a brief overview of what this means, we need to determine what M = PQ/V as M is the monetary base needed to support the economy. P represents the price of the good being provisioned by the network. When talking about Filecoin it would be $/GB, when talking about compute power it could be $/GHz etc. Q is the corresponding amount of that underlying resource that’s consumed in the crypto-economy.

I like the way this model is applied generally. The issue that I have with this model is that it’s hard to determine what the velocity will be. It cannot be dismissed out of hand because we could use a range of velocities to determine what a range of reasonable valuations could look like. However, with crypto-currencies, I’m uncertain how people will interact with them. Will people hold a number of crypto-currencies for all of the activities that they need to pursue? I think that this is unlikely, I believe that it’s more likely that people will hold a reserve currency or two and use an interface to pick up the currency that they need at any given time.

The value of a token is inversely proportional to the velocity, such that, the longer people hold the token, the more valuable it becomes. This intuitively makes sense if you consider that people might hold desirable tokens as they appreciate in value restricting the supply and creating a reflexive feedback loop and continuing to drive prices higher. This could explain what we saw in the crypto complex in late 2017.

The point that I want to make here is that it’s difficult to know what assumption we should use for velocity. It’s likely going to be different for different types of tokens and I’m unsure of whether the historical examples that we have could be applied to all types of crypto-currencies in the future. The velocity will depend on whether the crypto-currency is just a means of exchange or perceived store of value — it could lie on a continuum between these two ends as well. I believe that this is a start but probably not a conclusive way to look at crypto-currency valuations.

Network Value/Transactions (NVT) and Network Value/Future Transactions (NVFT) Ratios

The NVT ratio looks at the Network value divided by the daily Transaction volume on a network. This is an attempt at creating an equivalent of the P/E ratio that’s used to make relative value comparisons among stocks. Arguably, there’s less fudging that can happen because there are fewer accounting tricks that can be played on the denominator. As a result of the volatility of the transactions on any given day a smoothing function is typically used, which could for all intents and purposes be one year, which would link with one year’s earnings (or trailing twelve months earnings) that are used in the P/E ratio. HASH CIB has pointed out that there’s an inherent link between velocity and NVT, the more transactions that occur on-chain, the higher the velocity and the lower the multiple regardless of what smoothing mechanism is used.

HASH CIB also introduced the Network Value to Future Transactions (NVFT) ratio. This ratio replaces the trailing transaction multiple with a potential future transaction multiple. They’re quick to point out that the measure is subject to the same vagarities as the prior multiple because we’d require a mature market to understand where future transaction volumes could fall.

Rational Network Value (RNV)

Another approach proposed by Hash CIB is the Rational Network Value (RNV) approach, which applies a value to the current utility value (CUV) of the network (today) and adds the future discounted utility values (Additional Current Utility Values — ACUV). They make a reasonable argument that a utility network is more similar to a financial institution and its value should be determined by working out the excess returns on equity (returns to equity over and above the cost of equity).

The benefit of this model over the Burniske’s initial model is that it values the incremental additions to equity each year (discounted back to today) whereas Burniske’s model only looks at the terminal value. The model is still reliant on a terminal growth rate, which is likely as equally unknowable as the potential velocity at the end of a project. Typically, in Discount Cash Flow (DCF) models, I’ve seen terminal values that account for everything from 50–80% of a company’s value. Changing the growth rate only slightly can have significant changes to the terminal and thus the overall value of a company.

While I’ve pointed out some of the short-comings with the currently proposed valuation models, the issues are largely still relevant to traditional financial models as well. In writing this, I’m trying to look for a framework that I could apply to crypto-currencies to better understand where values should lie. My gut feeling is that we don’t have the tools that we need (e.g. understanding where the velocity of a mature project will sit) to be able to properly value crypto-currencies and crypto-networks — YET. We’re getting closer and each of these methods has built upon prior methods.

I believe that it’s likely that we’ll apply different models and standards to different crypto-currencies depending on the utility that they provide. This is similar to how an institutional investor would be willing to accept a range of EV/EBITDA multiples for utility companies but may prefer to look at a different range of P/BV ratios for airlines (if they are so bold) or P/PPOP for banks (noting these ranges are typically built on rules of thumb that are determined by market norms or more substantive financial models).

The Intelligent Investor

Personally, I’m more drawn to a back-of-the-envelope approach to valuing crypto-networks at the moment. This method has some reliable input and is less reliant on unknowable inputs, at least until networks mature. John Pfeffermade a pretty compelling case when he compared the value of Ethereum to Amazon Web Services (AWS). Both propositions are providing computing services for a price.

A simpler approach that he proposes compares AWS’s 2017 revenue of $17.5bn to the potential ETH GDP then at the time, the value of the Ethereum network would have had to reach $476bn or 27x the size of Amazon’s AWS revenue (at a velocity of 7) to justify the then value. If we apply these numbers to Amazon today, ETH would be turning over $21m p.a. (ETH Price at $112 and daily gas usage at 36.5bn WEI annualised with 365 days). This compares with a network value of $11.5bn.

If we assume that 10.6% of Amazon’s $773bn in market capitalisation is attributable to AWS, in line with the TTM revenue split, then the AWS portion of the business is currently worth $81.7bn. AWS is valued at 7x the value of the ETH network when it’s generating 1,000x the amount of revenue that the ETH network is generating annually. The argument could be made that 60% of Amazon’s value should be attributed to the AWS portion of the business in line with the earnings split but this would make the contract even starker.

Now, it could be argued that there should be further value in the network because of the other uses of the tokens, such as a means of money (store of value, means of payment and means of account). This could add some potential value to the network but it’s hard to argue that the price justifies the value as the technology is still maturing.

Conclusion

Through the diligent work of a couple of analysts we’ve started to lay a foundation to understand how we could look at crypto-currency valuations. However, I believe that we still don’t have the tools that we need to adequately apply thorough financial models. We can make some reasonable assumptions and stress test the models which might be sufficient for now. As projects mature and find their use cases, which I don’t really expect to happen for another five or so years, we’ll be able to build a toolkit to better understand how to approach crypto-currencies.

Coming from a fundamental equity investing background, I’ve noticed that it’s relatively rare that institutional investors (other than in private equity) put much reliance on DCF models. The assumptions are too easy to get wrong and terminal growth rates can have too much influence on the overall value of a company. I believe that starting with first principles or potentially drawing analogies to existing, dare I say “old world”, tech companies to the new blockchain and crypto paradigm we can get some fundamental underpinnings to draw our assumptions.

From these first principles valuations, we can draw some conclusions about where the end valuations should sit and then add some multiples on it to determine the potential that new technologies could add to the space. I’ll admit that this doesn’t feel like standing on the shoulders of giants stuff and I’m not building on the models that have been created recently. I’m talking about adapting tools that have been used on the buy-side for security analysis for a few decades. This doesn’t need to be rocket science, even though there’s a fair amount of computer science underlying the whole topic.