Crypto asset valuation: what you are actually buying when you buy a blockchain

Hyperliquid is worth nineteen times less than Ethereum, and yet it collected four times more fees than Ethereum over the past twelve months.

Both figures check out. As of 10 August 2026, Ethereum is capitalised at 232.3 billion dollars, Hyperliquid at 12.2 billion. Hyperliquid, for anyone unfamiliar with the name, is a crypto derivatives exchange that opened in late 2024. Over twelve months, the Ethereum chain collected 244.1 million dollars in fees. Hyperliquid, adding up its perpetuals, its spot order book and its own chain, collected 1,010.4 million. Sources: DefiLlama and CoinGecko.

One correction is needed before going further. Only 22 % of Hyperliquid tokens are in circulation, out of a maximum supply of one billion. When circulating supply is a minority of the total, market capitalisation flatters the asset, and fully diluted valuation is the honest base. On that basis, the gap narrows from 19 times to 4.25 times. It shrinks. It does not vanish.

Which leaves a simple question: what actually sets the price of an asset like this?

The reflex borrowed from equity markets, and why it fails

The reflex imported from public markets is to divide market capitalisation by fees collected. Everyone does it, and it is what produces the initial surprise. I have already written about why that number says nothing about what a holder actually receives once issuance is deducted. The problem I want to address here sits further upstream: the ratio does not even describe the network it claims to measure.

Applied to the major chains at the same moment, it yields 12 for Hyperliquid, 81 for Tron, 168 for Solana, 374 for BNB Chain and 951 for Ethereum. A spread of 79 times between the extremes, across assets doing broadly the same job.

When a measure produces that kind of dispersion across a homogeneous family, the explanation is rarely that millions of buyers are irrational. It is usually that the wrong thing is being measured.

Three frameworks tried in ten years, three dead ends

Not for lack of trying.

The monetary framework, 2014 to 2016. The token is treated as a network currency and the equation of exchange is applied: its value depends on the volume transacted and the speed at which it changes hands. The model runs into its own paradox. The more the token is actually used to pay for things, the faster it circulates, and the less it is worth. A framework that punishes usage was never going to survive.

The fat protocol thesis, from 8 August 2016. Joel Monegro published at Union Square Ventures the idea that would finance an entire cycle: unlike the internet, where value settled in applications while protocols stayed free, blockchains would invert the stack, and value would concentrate at the shared protocol layer. The facts largely said otherwise. Applications captured value, protocols rented out their space at cut rates. With one exception, which I will come back to.

Deflationary money, from August 2021. An upgrade introduced the burning of part of the fees, and the token finally acquired something resembling a return to holders. The narrative held for two and a half years. Then the chain decided to break its own fee base, and the reasoning collapsed with it.

Three frameworks borrowed from elsewhere: monetary economics, platform economics, corporate finance. Each broke on the same reef. A blockchain is not a company, not quite a currency, and not exactly a platform either.

Meanwhile, the market quoted a price every single day for ten years. Based on what?

What a blockchain actually sells: space, not computation

Step away from the ratios for two minutes and look at the object itself.

On Ethereum, as of 10 August 2026, there are 1,831 deployed protocols. The chain carries 148 billion dollars of stablecoins, or 49.7 % of every digital dollar in circulation worldwide, and 42 billion dollars locked in its applications. In total, roughly 190 billion dollars sit on top of it.

On Hyperliquid, same date: 185 protocols, 6.21 billion in stablecoins, 1.21 billion locked. Roughly 7.4 billion.

Ten times fewer applications. Twenty-five times less value carried. And four times more fees collected.

This is where the object shows itself for what it is. Ethereum does not sell computation. It sells space, and it rents that space out at a derisory price relative to what it shelters: 244 million dollars to secure 190 billion, an annual rent of 0.13 % of the value hosted.

Applications see that same rent from the other side of the counter. When a major lending protocol costed out what a chain actually earns it, some of its deployments were running below 5,000 dollars per quarter.

And the compression was not suffered, it was chosen. In the wake of the upgrade that introduced blobs in March 2024, chain fees fell from 2.47 billion dollars in 2024 to 523 million in 2025, down 79 %. Ethereum deliberately gave away its own revenue to lower the cost of using its ecosystem. A listed company that cut its own turnover fivefold on purpose would be torn apart. Here, it was the strategy.

The measure that closes the gap

Since what matters is what the network carries, let us measure that.

By value carried, I mean the sum of stablecoins issued on the chain and capital locked in its applications. This is not an industry standard, it is a convention I am setting for this exercise, and everything that follows depends on it.

Market capitalisation divided by value carried, same source, same day:

Ethereum: 232.3 / 190.0 = 1.22.
Hyperliquid: 12.2 / 7.4 = 1.64.

On fully diluted valuation, the only defensible base for Hyperliquid given its float, the ratio rises to 7.35 (54.6 / 7.4).

In other words: the nineteen-fold gap closes by an order of magnitude as soon as you look at what the two networks carry rather than what they charge. It falls to 6 times on the harshest base, and to 1.3 times on the most generous one. It does not close entirely.

Nothing changed in the world between the first calculation and this one. Only the denominator changed.

The test across seven networks

A coincidence across two assets proves nothing.

A warning before the numbers, otherwise the list becomes dangerous. What follows is not a ranking from cheapest to most expensive. Tron shows the lowest ratio and secures proportionally almost four times more dollars than Ethereum, which says nothing whatsoever about the quality of that security, the degree of centralisation of the network, or how it would behave under stress. The measure describes an exposure, never a merit.

Seven chains measured on 10 August 2026, by descending market capitalisation. Price to value carried, then price to fees in brackets:

Ethereum 232.3bn · value carried 190.0bn · 1.22 (fees: 951)
BNB Chain 80.4bn · 18.3bn · 4.39 (fees: 374)
Solana 44.8bn · 20.4bn · 2.19 (fees: 168)
Tron 31.4bn · 96.6bn · 0.32 (fees: 81)
Hyperliquid 12.2bn · 7.4bn · 1.64 (fees: 12)
Avalanche 2.8bn · 2.0bn · 1.40 (fees: not measured)
Aptos 0.5bn · 1.3bn · 0.38 (fees: not measured)

These ratios are computed on market capitalisation for all seven. Hyperliquid is the only one whose float is low enough for the distinction to matter: on a fully diluted basis it moves from 1.64 to 7.35.

Measured against value carried, the seven sit within a range of 1 to 14, or 1 to 23 if the diluted base is used for Hyperliquid. Measured against fees, the five for which the measure exists spread from 1 to 79.

My reading, and I offer it as such: the market does not first ask how much a network collects, it asks how much value agrees to live on it. This is a regularity observed across seven networks on a single date, not a law. It does, however, sort those prices into a band three to five times tighter than the ratio everyone reaches for.

Platform or product: the category error that distorts every ratio

The heart of the matter is a distinction the market applies without ever stating it.

A platform is worth what it hosts. Its fees are a by-product, sometimes even an adjustment variable that gets crushed on purpose to bring people in. What gives it value is the number of applications that depend on it, the capital willing to sit on it, the depth of its developer ecosystem, its role as a settlement layer. Ethereum is a landlord, and you do not value a landlord on one year of rent, but on the city that has been built on their land.

A product is worth what it collects. Its surface is narrow by construction: one function, one market, one clientele. Hyperliquid belongs to that second family: a narrow surface, concentrated revenue, a directly measurable market share.

Which brings back the exception left hanging earlier. Hyperliquid did not refute the fat protocol thesis, it made it true in the only way that has worked so far: by collapsing the stack. The application and the protocol are the same object, on the same chain, with the same token, so there is no intermediary left for value to leak to. That is what explains its fees, not technical superiority. An application captures protocol-level value only by becoming the protocol itself.

Comparing a platform's fee multiple to a product's is comparing the price per square metre of an entire district to that of a single shop. Both numbers are correct. They answer different questions.

So the opening anomaly was never a market anomaly. It was a category error. Mine, before I went and checked.

What this changes when you look at an asset

Three things I take away, and they hold well beyond these two networks.

First, before reaching for a ratio, work out which family the object belongs to. Platform, product, or infrastructure that others depend on without paying it much. The right denominator follows from that answer, never the other way round.

Second, low fees are not always a sign of weakness. They can be the price paid for an ecosystem to be built. The useful question is not "how much does it collect", but "what would happen if this network stopped tomorrow". For 190 billion dollars sitting on Ethereum, the answer takes a while to write.

Third, none of these ratios tells you whether an asset is expensive. They only tell you where the market is looking. Price remains a collective opinion about the future, and no division has ever replaced it. Which stops no one from trying: a major index provider built its admission criterion on protocol revenue verified directly on chain. The debate is anything but theoretical.

Next time a number strikes you as absurd, the useful question is probably not who is getting it wrong. It is what that number is actually measuring.

Frequently asked questions

What determines the valuation of a crypto asset?
For a blockchain, the measure that best sorts observed prices is the value sitting on the network: stablecoins issued on it, capital locked in its applications, and the number of applications that depend on it. Network fees sort them far less well. Across the seven chains measured on 10 August 2026, dispersion runs from 1 to 14 on value carried, against 1 to 79 on fees.

Why is a revenue multiple not enough?
Because a blockchain can choose to give away its own fees to attract users, which Ethereum did from March 2024 onwards. Falling fees can therefore signal a strategy rather than lost traction. The ratio cannot tell the two apart.

How is value carried calculated here?
Stablecoins issued on the chain plus capital locked in its applications, recorded on DefiLlama as of 10 August 2026. This is a convention set for this analysis, not an industry standard. The scope excludes activity on layer 2 networks.

Does this ratio rank blockchains from cheapest to most expensive?
No, and that would be the worst possible reading. Tron shows the lowest ratio on the list while being markedly more centralised than Ethereum. The measure describes an exposure, never a quality or a level of security.

Is the count of deployed applications exact?
It is the DefiLlama listing as of 10 August 2026, which structurally understates newer chains. The order of magnitude is solid, the precise figure is not.