I pulled the numbers before I read the announcement. ANTH quoted at $2,327, up 5.4% on the day. Twenty-four-hour notional volume: $14.32 million. Open interest: $28.25 million. Deployed on Hyperliquid L1 through the HIP-3 mechanism by an entity called Entropy, framed as an Anthropic pre-IPO perpetual market.
Three of those figures are ordinary. The fourth is the whole article. Open interest is running at roughly twice daily volume, which inverts the normal signature of a liquid perpetual. Deep BTC markets turn their open interest over several times a day — basis traders, market makers, funding harvesters cycling continuously. When open interest sits above daily volume, the average outstanding position is being held longer than 24 hours. On a venue whose entire product is continuous leverage, a market where positions sit still is a market where nobody is arbitraging the basis.
A perpetual contract nobody arbitrages is not a price. It is an opinion with a funding rate attached. That distinction is where the risk lives.
HIP-3 is Hyperliquid's market-deployment pathway: new perp markets enter through the L1's governance and validator layer, with designated deployers executing configuration. Entropy is the deployer here. Mechanically, almost nothing about this is novel. Hyperliquid already runs dozens of perps, and adding another is a governance operation rather than a protocol upgrade. No new cryptography. No new consensus. No new settlement primitive. Whatever innovation exists sits entirely in the choice of underlying.
Which is where it gets interesting. A perpetual has three moving parts: an index price, a mark price, and a funding rate that tethers one to the other. For BTC, the index is a composite of spot prices from venues with real order books holding real coins. Arbitrageurs buy spot and short the perp when the basis widens, and that arbitrage is what gives the funding rate meaning. The whole machine runs on the assumption that someone, somewhere, can settle the difference in the underlying asset.
Anthropic is a private company. There is no spot ANTH. There is no order book holding anything. The closest thing to a reference is a set of periodic secondary-market indications — broker-quoted marks from platforms that facilitate tender offers and employee liquidity events — which print monthly at best and describe transactions that already happened. A perp trading continuously cannot be tethered to a reference that updates every thirty days without the funding mechanism doing nearly all the work.
So the question worth asking is not whether ANTH goes up. It is: what exactly is the index price of this market, who computes it, how often, and what happens when it disagrees with the last price the perp traded at. Those are questions a functioning contract must answer. None of the public material answers them.
The circular reference problem. If the index is imported from a stale private-market mark, funding has to carry the entire basis between a monthly print and a live market. Suppose the last secondary indication implies $1,800 and the perp quotes $2,327 — a 29% premium. With funding clamped at a typical hourly ceiling, the carry required to close that gap over any reasonable horizon becomes enormous, and the clamp becomes the binding constraint. Once the cap binds, funding stops doing its job. The perp decouples. From that point forward the mark is validated by nothing except the venue's own last trade, and you have a circular reference: the price is correct because the price says so. That is not a hypothetical scenario. It is arithmetic. It is also where the oracle-latency problem that haunts every DeFi derivative stops being an abstraction and becomes a line item. I spent three weeks in 2020 reverse-engineering dYdX's flash-loan plumbing for a pre-mortem, and the lesson generalizes: price feeds do not fail loudly. They fail quietly, by being right about the wrong timestamp.
Funding is the only honest signal here. Which means the most informative number in this market is not $2,327. It is the funding rate. If funding sits persistently positive and near the clamp, longs are paying heavily to hold a claim they cannot redeem, and the basis between the perp and any external reference is wide. If funding is flat or oscillating, the market is either genuinely well-arbitraged — unlikely, given the open-interest-to-volume ratio — or the index is computed internally in a way that makes convergence automatic by construction. Both cases deserve scrutiny, for opposite reasons.
The liquidation engine's balance sheet. Open interest of $28.25M sounds small until you ask who absorbs a cascade. Hyperliquid routes liquidations through a backstop vault, which means venue capital rather than a third-party clearinghouse is the buyer of last resort when positions fail. Across $28.25M of notional, an average effective leverage of even 5x implies something on the order of $5.6M of posted margin in the entire market. A 20% adverse move vaporizes essentially all of it in one candle. In a market whose reference price updates monthly, a 20% candle is not an edge case — it is what happens the first time a real headline crosses the tape. The cascade then clears into an order book whose daily notional is half the open interest, meaning the book is thinner than the exposure it must absorb. Liquidity, in that configuration, is just trust with a price tag.
What "pre-IPO" actually denotes. The label implies a claim on future equity. Mechanically, that is almost certainly not what trades. A cash-settled perpetual with no deliverable conveys no shares, no pre-emptive rights, and no conversion at listing. It is an exposure expression, not an ownership instrument. The distinction changes the legal analysis entirely: an instrument settling against a reference price published by a venue is a derivative, full stop, whatever the underlying company is called in the marketing. If that reference price derives from secondary-market transactions in actual equity, the payoff is tied to a security's price — historically the configuration regulators find easiest to characterize. Not a prediction about enforcement. A statement about what the contract does.
The token vacuum. Here the disclosures stop. No supply figures. No vesting schedule. No team allocation. No treasury. No protocol revenue routing. ANTH could be a Hyperliquid-ecosystem utility token, a pure derivative with no cash-flow claim, or something else entirely. The material does not say, and that absence is itself the finding. Yield is a function of risk, not just time, and you cannot price a risk you cannot enumerate. A market printing $2,327 with zero published supply data is a market where the float — and therefore the price — is unconstrained by anything mechanical.
Deployer authority. HIP-3 designates a deployer to configure the market, and in a perpetual that configuration includes oracle sourcing, leverage caps, funding parameters and, depending on implementation, the ability to modify or wind down a market. That is an administrative surface with unilateral authority over market structure, and its scope is not documented in anything I have reviewed. Audit reports are promises, not guarantees — and here there is not even a report doing the promising. In 2024 I audited MPC threshold signing for an Indian exchange preparing institutional custody. The finding that mattered was not in the signing algorithm. It was a side channel in key generation, in the part of the system nobody had thought to specify. Markets with undescribed admin surfaces have the same shape.
The bull case writes itself: tokenized private markets, AI narrative meets on-chain leverage, price discovery for the most-watched private company on earth. It is a good story and it is structurally wrong on one point. Price discovery requires that the people holding information be able to trade. Employees with Anthropic equity, funds on the cap table, anyone in possession of material non-public information, are legally constrained from expressing a view in a derivative on the same issuer. The marginal participant in this market is therefore systematically uninformed. What ANTH discovers is not Anthropic's valuation. It discovers what retail sentiment about Anthropic's valuation looks like when it is levered.
There is a second inversion. Most commentary treats the missing token economics as an oversight. It is more plausibly a design feature. An instrument with no supply schedule, no revenue claim and no settlement path cannot be valued against anything, which makes it maximally responsive to narrative and minimally accountable to fundamentals. The absence of fundamentals is not a gap in the market. It is the product.
That is the part I would want written into an audit. Not "the contract is safe." Rather: state the reference price, state the update frequency, state the fallback when the reference is stale, state the deployer's authority. Four sentences. None of them currently exist in public.
The number to watch is not $2,327. It is the funding rate, and behind it, the oracle configuration — because that is the only mechanism holding a continuously traded instrument to a monthly-updated private mark. If funding sits pinned at its cap for more than a few days, the tether has already broken and the market is pricing sentiment rather than equity. The right question for anyone sizing this position is not whether Anthropic goes public. It is what happens to $28M of open interest in a half-volume book the first time a real headline lands.
