Revenue is not a claim: the limits of fundamental analysis in tokens

It is worth beginning with the strongest version of the argument, because it is stronger than it was and dismissing it would be lazy.

A handful of protocols now generate revenue at a scale that is not in dispute and is verifiable on public infrastructure. In the first half of 2026 Hyperliquid produced $419.3 million of gross fees, of which $305.3 million was retained as core protocol revenue — a capture rate of roughly 73%, which very few businesses of any kind achieve. The direction of those two numbers is the more interesting fact: gross fees rose 31% against the first half of 2025 while core protocol revenue fell slightly, because its HIP-3 markets let external builders keep half the trading fees they generate. Revenue can grow at the venue and shrink at the protocol at the same time, and only one of those is what a tokenholder is being invited to value. Over the twelve months to 30 June 2026 Aave generated $930.1 million of fees and $122.8 million of revenue on DefiLlama's daily series — a capture rate near 13%. Uniswap has generated fees on a comparable scale while capturing a small fraction of them — but that gap is closing rather than chosen: its UNIfication proposal passed in December 2025 with near-unanimous support, and protocol fees were switched on across selected v4 pools on seven networks on 27 July 2026, lifting daily protocol revenue several-fold from a low base. Sector-wide, reported protocol revenue has run into the billions year to date.

What is new is not the revenue but the plumbing attached to it. Several protocols now route part of that revenue into buying their own tokens, a mechanism that barely existed two years ago. Uniswap passed a governance proposal to enable fee capture with near-unanimous support; Aave has run buybacks since 2025 and expanded the programme in 2026; Hyperliquid directs the overwhelming majority of its fees to buying and burning its token. Serious market participants have argued that as this spreads, valuations could expand materially without any growth in revenue at all.

That is a coherent argument and parts of it are correct. Fee capture rising from near zero to the majority of revenue is a genuine change in how these systems work, and an analyst who ignores it is not being rigorous, merely incurious.

A word on what follows. The protocols named in this note are cited because their figures are public and their governance decisions are documented, which is what makes them usable as case studies. Naming one is not a view on it, an indication that we hold it, or a recommendation of any kind. This note is written for information purposes, and nothing in it should be read as saying that Block Asset Management invests in, or would invest in, any asset mentioned.

The argument for applying a price-to-earnings ratio to a token rests on an analogy, and analogies are worth examining at the joints.

When you buy a share you acquire three things that are frequently taken for granted precisely because they are so reliable. You acquire a legal right to the residual profits of the enterprise after everyone senior to you has been paid. You acquire a defined position in the capital structure, so you know who ranks ahead of you in a liquidation. And you acquire a mechanism of accountability: directors owe you duties, you vote on their appointment, and if they misapply your money there is a body of law and a court that will hear you.

The price-to-earnings ratio is a shorthand for the first of those, and it is only meaningful because the other two stand behind it. It is not a measure of how much money an enterprise makes. It is a measure of the price of a claim on that money — and a claim is a legal relationship, not a cash flow.

This is why the multiple is doing more work than it appears to. When an analyst says a company trades at fifteen times earnings, the earnings figure is the visible part and the claim is the invisible part. Remove the claim and the ratio does not become a worse valuation tool. It becomes a category error dressed as one.

Against that, consider what a typical protocol token confers. The specifics vary and some designs are more thoughtful than others, but the general position holds more often than the fundamental-analysis framing acknowledges.

None of the above would matter much if tokens with revenue and buybacks reliably behaved like the equities they are being compared to. The record so far does not support that, and the most instructive case is the one that did everything the argument asks for.

Pump.fun generated well over $900 million of revenue and, for roughly nine months, directed effectively all of it into buying back and burning its own token. On 29 April 2026 it burned approximately $370 million of supply in a single action, reducing tokens in circulation by around 36% — and, in the same move, replaced the all-revenue policy with a contract committing half of net revenue to ongoing buybacks. By almost any reading, the first version was the thesis executed in full: real revenue, near-total capture, and an aggressive, verifiable reduction in supply. It was halved anyway.

The token spent most of 2026 below its launch valuation regardless. And then the decisive thing happened — the protocol changed the policy, moving from directing all revenue to buybacks to splitting it roughly in half between buybacks and operations, on the reported basis that the model was not supporting the price.

That sequence is the whole argument of this note, compressed. The distribution was real while it lasted. It did not produce the outcome it was expected to produce. And when it disappointed, the people controlling it changed it — which is precisely what a discretionary policy is, and precisely what a dividend backed by a legal claim is not. An equity investor whose dividend is cut has a governance grievance and a set of rights. A tokenholder in the same position has an announcement.

The comparison is not confined to one protocol. Ripple's private share price has risen substantially, marked around $136.90 in secondary markets after a $750 million buyback in March 2026 that valued the company near $50 billion, while XRP fell over the same period. The most revealing detail is arithmetic, and it holds at the prices of that moment rather than permanently: at the XRP price prevailing around the buyback, the company's own holding of roughly 38 to 40 billion tokens would have been worth more than the entire company was valued at. Sophisticated investors, pricing the equity directly, marked those tokens well below the screen price — in part because escrow arrangements cap how much can be released in any month, so the holding cannot in practice be sold at the price quoted for it.

That is worth sitting with. In the same enterprise, the instrument with enforceable rights and the instrument without them were valued on entirely different bases, and the market that priced both took the view that the token was worth materially less than its own quotation.

A second omission runs through most of this analysis, and it is the one an equity analyst would notice first.

When a company announces a buyback, the analyst asks about the share count — issuance, options, convertibles, anything that offsets the repurchase. Net of dilution is the only figure that matters, and everyone knows it. In tokens, the equivalent question is asked far less often, despite the answer usually being published years in advance.

Most protocol tokens have contracted emission and unlock schedules: tranches releasing to teams, investors and incentive programmes on dates set at launch. A buyback absorbing a given amount per month against a schedule releasing more than that is not reducing supply in any economically meaningful sense; it is slowing an increase. The protocol discussed above has faced exactly this, with a substantial unlock scheduled against a buyback programme that had already been reduced.

There is also a subtler version. Where buybacks are funded by recycling tokens back into incentive programmes rather than by permanently removing them, the reported buyback figure and the change in circulating supply are different numbers, and only one of them affects a holder. The distinction between a burn and a recirculation is the difference between a repurchase and a transfer, and it is not always obvious from the announcement.

There is a complication in this argument that its proponents rarely address, and it is the one an institutional allocator should notice.

The more explicitly a token is presented as conferring a share in the profits of a common enterprise, the more closely it resembles the thing that securities law exists to regulate. The tests differ between jurisdictions, but the direction is consistent: an instrument marketed on the expectation of profits derived from the efforts of others sits closer to the regulatory perimeter than one marketed as a utility or a governance right.

This creates a genuine tension rather than a rhetorical one. The clearer and more binding the value-accrual mechanism becomes — the more it looks like the enforceable claim that would justify a price-to-earnings ratio — the harder it becomes to maintain that the instrument is outside the securities framework. And the more discretionary and revocable it remains in order to stay outside that framework, the less it supports the valuation argument being made for it.

This is the same boundary our notes on the CLARITY Act and on tokenisation approached from other directions. Where an instrument is a security, existing securities law applies to it and the crypto-asset regimes do not; the technology does not determine the classification, the nature of the instrument does. An allocator being shown a fundamental valuation of a token is entitled to ask which side of that line the issuer believes it is on, and what happens to the thesis if a regulator disagrees.

None of this means fundamental analysis has no place here. It means the analysis has to be of the right object. Revenue, margin, competitive durability and reinvestment discipline are all worth examining — they tell you about the health of the business. What they do not tell you, on their own, is what your instrument is worth, because your instrument may not be a claim on that business.

Five questions do most of the work, and they precede any multiple: