ANTH on Hyperliquid: How a $14.3M Order Book Priced Anthropic at $2.33 Trillion
The Print
On September 9, a market called ANTH printed $2,327.
Two other numbers sat behind that print, both small enough to fit on a single line: 24-hour volume of $14.32 million, and open interest of $28.25 million.
From those three figures, an arithmetic chain produced a fourth. An implied Anthropic valuation of roughly $2.327 trillion.
That fourth number is the one that traveled. It was screenshotted, quoted and repeated as though it were a measurement. It is not a measurement. It is a product โ and the constant being multiplied was chosen by whoever wrote the contract, not by Anthropic, not by a court, not by any cap table that has ever been made public.
I have audited markets built exactly this way. In the summer of 2020 I spent a week instrumenting Uniswap V2 pools for a protocol I will call YieldFarm X, which reported nine figures of total value locked. The contracts did hold tokens. But the same 500 ETH of collateral was rotating through five pools simultaneously, and the headline figure described nothing about the protocol's actual capital base. It collapsed inside 72 hours. A market's headline number is only as honest as the mechanism that generates it. Almost nobody covering ANTH has separated the number from the mechanism, and that separation is the entire story.
What the Feed Says, and What It Refuses To
ANTH is a perpetual futures market on Hyperliquid, deployed under HIP-3 โ the framework that allows third-party builders to spin up perpetual contracts against Hyperliquid's order book rather than waiting for the core protocol to list them. The reported deployer is a builder called Entropy. I cannot verify that from the public record, and I have not seen the deployment's parameter file.
Before any analysis, a disclosure ledger. I have kept this deliberately flat, because the shape of the gaps matters more than the shape of the filled cells.
| Item | Status | |---|---| | 24h volume | $14.32M (disclosed) | | Open interest | $28.25M (disclosed) | | Last price | $2,327 (disclosed) | | Funding rate | Not disclosed | | Long/short composition | Not disclosed | | Oracle source and update cadence | Not disclosed | | Settlement rule at IPO | Not disclosed | | Share-count multiplier | Not disclosed | | Deployer's emergency delisting authority | Not disclosed | | Audit | Not disclosed | | Legal or regulatory registration | Not disclosed | | Team, ownership, jurisdiction | Not disclosed |
Eight of the fields that would let you price this market are absent. What remains is a price, a volume and an open interest. On their own, those describe a tape, not a security.
Chain links don't lie. But a chain link only carries what it was built to carry. A perpetual with no deliverable is not a share. There is no custodian, no transfer agent, no borrow market, no dividend, no voting right, and โ critically โ no share certificate that will change hands at settlement. What a holder owns is a claim on a number that resolves when an event occurs, and the rules for resolving that event were written by the counterparty to the trade.
This is where ANTH diverges from the tokenized-equity products that multiplied over the past two years. Those instruments carry a legal wrapper: a broker-dealer or a special-purpose vehicle holds the underlying, and the token is a claim on that wrapper. ANTH has no wrapper. It is a pure synthetic โ a derivative on an equity that has no public market, no ISIN, and no reference price observable outside Anthropic's cap table.
There is precedent for that structure, and it is not encouraging. FTX listed pre-IPO equity tokens in 2020. They were synthetic in much the same way, they were popular, and they were discontinued once regulators made their position clear. The lesson was not that the product was technically impossible. The lesson was that it was impossible to operate without a legal answer to the question of what the token actually was.
Two Questions Nobody Has Answered
The first is jurisdictional. ANTH is a derivative whose underlying is a private company's equity. In the United States, that places it somewhere between a security-based swap and a contract on an unregistered security, and both roads lead to registration requirements that a permissionless deployment cannot satisfy. Offshore venues have historically argued that they fall outside that perimeter because they do not serve US persons. That argument has held until it hasn't โ repeatedly.
The practical question is not whether ANTH is legal. It is whether Hyperliquid or the deployer retains a governance lever to halt the market if a regulator or Anthropic itself objects. That lever almost certainly exists at the smart-contract level. Whether it exists at the operational level is unknown, and unknown is the answer you least want when you are holding a position.
The second question is settlement. A pre-IPO perpetual has no final anchor until an IPO occurs. If Anthropic lists at $400 billion, what does ANTH resolve at? If it lists at $300 billion? If it never lists? None of these are answered in the published material. An instrument without a defined terminal condition is not a forecast of anything. It is a wager on a definition.
The Multiplier Nobody Disclosed
Now the arithmetic.
The $2,327 per-share print and the $2.327 trillion implied valuation are not two independent facts. They are the same fact. The chain only closes if the market applies a multiplier of exactly one billion shares:
$2,327 ร 1,000,000,000 = $2,327,000,000,000
That is a suspiciously round correspondence. It implies a modeled share count, hard-coded into how the number is displayed or how the reference price is derived. The "$2.327 trillion valuation" is not an observation about Anthropic. It is the output of a constant that nobody published, applied to a price that a thin order book printed.
This matters because the constant is not knowable. Anthropic's issued share count is private. The error bar on the headline number is therefore inherited entirely from the error bar on the multiplier. Move the multiplier to 700 million shares and the implied valuation falls to $1.63 trillion. Move it to 1.4 billion and it rises to $3.26 trillion. Same tape, same price, a range spanning $1.6 trillion โ a spread wider than the annual output of most nations, produced by a single unmade disclosure.
When I audited Project Aether in 2017, the entire fraud lived inside a constant. A hidden minting function, callable from a developer-controlled address, meant the supply shown to buyers was not the supply that existed. A 12,000 ETH discrepancy between stated and actual token counts was enough to get the project delisted from three exchanges. I am not alleging anything of that kind here โ there is no evidence of it, and the comparison is structural, not moral. But the weakness is the same shape: a public number whose most important input is a private assumption. Code is the only witness, and in this case the witness has not been called.
The Ratio Test
If the multiplier is unverifiable, the liquidity is verifiable โ and it is the more interesting number.
Open interest of $28.25 million against an implied valuation of $2.327 trillion gives a ratio of 0.00121%. Written plainly: the market's total committed capital equals 1.2 parts per 100,000 of the valuation it is asserting.
For context, the cleanest analogue I have comes from work I did last year. I built a tracking model for a family office that compared daily net inflows into BlackRock's IBIT against on-chain exchange reserves, and the finding that mattered was structural: ETF demand was not moving a price in a vacuum. It was interacting with a float of measurable size, and the interaction was visible in reserve balances. That model worked because Bitcoin's float is observable. BTC perpetual open interest across major venues typically runs in the low single digits as a percentage of Bitcoin's market capitalisation. Call it 1% for the sake of argument.
ANTH's ratio is 0.00121%.
That is roughly two to three orders of magnitude thinner. I want to be careful with the comparison, because it is not like-for-like โ BTC's figure is a market cap and ANTH's is an implied one. But the direction of the gap does not depend on the multiplier's precision. Even if the true share count halves the headline number, ANTH's open-interest-to-valuation ratio remains below anything a functioning price-discovery market produces.
Turnover makes the same point from another angle. $14.32 million of volume against $28.25 million of open interest is a 50.7% daily turnover ratio. For an instrument whose entire premise is holding exposure to a company that will not be tradeable for months or years, 50% daily churn is not investment flow. It is a trading book wearing an equity ticker as a costume.
The capital-to-claim ratio is starker still. $14.32 million of notional traded against a $2.327 trillion assertion means that for every dollar of implied valuation, roughly six ten-thousandths of a cent actually changed hands in a day.
Wallets connect the dots, and here there are few dots to connect. With $28.25 million of open interest, a single $1 million position is 3.5% of the entire market. One account rebalancing once is a material price event. Two accounts moving in the same direction are a trend. That is not a market. That is a room.
The Illiquidity Discount, Inverted
Here is the detail that reframed this story for me.
In traditional finance, restricted pre-IPO equity trades at a discount to the last primary round. The mechanism has a name โ discount for lack of marketability โ and it exists because a buyer who cannot exit for eighteen months demands compensation for that illiquidity. Depending on the restriction period and the company's profile, that discount typically runs 10% to 30% against the most recent mark. It is one of the few relationships in private markets that holds with reasonable consistency.
ANTH inverts it. Against a reported primary range of roughly $200 billion to $350 billion, the implied $2.327 trillion is a premium of about 6.6x to 11.6x. The same underlying asset, in two markets, carries opposite signs on the same risk factor. Private holders are compensated for being trapped. ANTH holders pay enormously for the privilege of being able to exit continuously.
That is the actual content of the $2,327 print. It is not a valuation of Anthropic's business. It is the price of exit optionality on an asset that has none, expressed in a venue where the exit is instantaneous and the reference price is unobservable. Those are two different numbers, and the market has priced the second one while everyone quotes it as the first.
Once you see it that way, the premium stops looking like a bubble and starts looking like a liquidity derivative. It is still fragile. It is still unhedgeable. But it is not necessarily irrational โ and that distinction is precisely what most coverage has missed.
Where the Price Actually Comes From
If the multiplier is unverified and the book is thin, what is the print telling us?
The honest answer is narrower than the headline. It is telling us what the marginal trader with a Hyperliquid account was willing to pay, in a book with no arbitrage link to any reference price. That is the only statement the data supports.
The private-market anchor sits somewhere in the $200 billion to $350 billion range based on reported primary rounds. Against a $2.327 trillion implied figure, that is a 6.6x to 11.6x premium. In a normal market, a spread that wide gets closed by someone buying the cheap leg and selling the expensive one. Here, no mechanism exists to do it.
You cannot buy Anthropic shares โ the round is closed, allocations are gated, and secondary transactions are restricted and illiquid. You cannot borrow them either. The only expression of a bearish view is a short perpetual, which pays funding to the crowd for as long as the premium persists, and can be liquidated long before the thesis resolves. In 2022 I watched UST's reserve addresses deteriorate three days before the public announcement, and that trade worked because Curve pools gave me an executable route to express the view. ANTH offers no equivalent route. A price with no arbitrage mechanism is neither overvalued nor undervalued. It is unconstrained. Those are different words, and conflating them is the most common analytical error in this story.
The missing field that would resolve everything is the funding rate. Funding is the only real-time measurement of who is paying to hold their position. If ANTH's funding is deeply positive โ say, north of 50% annualised โ longs are crowded and the premium is being actively financed, meaning the position pays rent to exist. If funding sits near zero or negative, the premium is not financed at all, which means the print is a stale quote carried by an empty book.
I cannot see the funding rate in the published material. As far as I can tell, neither can anyone else reading the same coverage. A pre-IPO perpetual story without a funding rate is a story with its most important number missing.
The Contrarian Case Against My Own Skepticism
The obvious reading is that ANTH is a bubble printing a fantasy number. The contrarian reading is that the market is doing exactly what it was designed to do, and the number only looks absurd because we are comparing it to a private valuation that is itself inaccessible.
Follow the gas, not the hype. The reason ANTH trades at 6.6x to 11.6x the last primary mark is not that traders are irrational. It is that the private mark is not a price. It is a negotiated allocation in a closed round, available only to a set of institutions with relationships and capital. ANTH is the only continuously tradeable expression of AI equity exposure a crypto-native participant has ever had. In a world where the private market is gated and the public market offers no entry, a large premium for access is not a malfunction.
The premium may be a scarcity rent, priced in earnest, on a market that happens to be very small.
There is a second version of this argument that is harder to dismiss. Thin liquidity cuts both ways. Yes, a small book means a small capital base can set a large implied number. It also means the number is fragile in the other direction: one large seller could collapse it without any change in Anthropic's fundamentals. That fragility is a property of the experiment, not evidence of manipulation.
Where I part ways is on the composition of the marginal buyer. The natural buyer of AI equity exposure is an AI investor. The natural participant in a Hyperliquid perpetual is a crypto perp trader. These populations overlap far less than the ticker suggests, and the second one is not running a discounted cash flow model on a frontier lab's enterprise contracts.
This is testable, and I would want to test it before accepting the price as an AI signal. If the print were being set by fundamental investors using a synthetic rail, I would expect volume to correlate with Anthropic's news cycle. I would instead expect it to correlate with crypto's own volatility regime. Correlation is not causation. A number that moves with crypto beta is not a valuation of an AI company. It is a crypto position that borrowed an AI company's name.
Risk Disclosure
Any position in a market like ANTH should be underwritten against the specific on-chain and off-chain metrics that would invalidate the thesis. Four would trigger reassessment for me.
A sustained annualised funding rate above 50% would indicate crowded longs financing an unsustainably wide premium, and would shift my expectation toward a liquidation cascade rather than a gradual convergence. Any public statement from Anthropic disclaiming authorisation or requesting cessation would remove the market's only external anchor, and would likely force a delisting whose timing and price are both unknowable in advance. A published settlement specification that ties the contract to a price determined at the deployer's discretion would convert what looks like a market into what is functionally a counterparty claim. And a visible concentration of open interest in fewer than ten addresses would confirm that the price is a position, not a discovery.
None of these are predictions. They are thresholds. The distinction matters, because the failure mode in thin synthetic markets is rarely a slow repricing. It is a single block of liquidity leaving, and the last person to notice pays for everyone who noticed first.
What to Watch
Three fields would tell me more than any price forecast.
Funding rate. If ANTH's funding goes persistently and steeply positive, the premium is being financed by longs who cannot afford to be wrong for long. That is the condition under which a thin book becomes a stampede.
The settlement rule. A pre-IPO perpetual has no final anchor until an IPO happens. If the listing is delayed, repriced lower, or cancelled, the contract resolves against a definition nobody has published. I have watched what happens when resolution criteria are vague: the ambiguity gets settled by whoever holds the largest position. Read the spec before trading the ticker, not after.
The next deployment. HIP-3's real significance is not ANTH. It is that the mechanism exists at all. If builder-deployed perpetuals on non-crypto assets prove operable โ operationally and legally โ the pipeline that follows carries names like OpenAI, SpaceX and Stripe, and the volume that comes with them goes somewhere. That is a materially larger story than one $2.3 trillion number on a small book, and it is the one I would track over the next two quarters.
One question to carry forward. When a company's price is discovered on-chain before its shares exist, who is the counterparty, and what exactly are they on the hook for?
The tape will answer that eventually. It usually does, and usually later than the people holding the position would prefer.