At 09:00 UTC on September 9, a ticker most traders could not have named a week earlier printed $2,327.
Multiply that by the share count implied in the quote and you land on $2.327 trillion. That is the number Hyperliquid's ANTH perpetual is assigning to Anthropic — a company with no registration statement on file, no public transfer agent, and a last private round reportedly struck an order of magnitude below this mark. Twenty-four-hour volume: $14.32 million. Open interest: $28.25 million. Those three figures, in that order, are the entire story. None of them are about Anthropic.
The contract is a pre-IPO perpetual future deployed under HIP-3, Hyperliquid's builder-deployed market framework. Entropy is reported as the deployer. And the deployment, not the price, is the event.
Until HIP-3, the venue's perpetual listings tracked crypto assets and a narrow band of proxy instruments. HIP-3 lets a third-party builder define the oracle, the margin asset, the leverage caps, and — critically — the settlement rule, then split fees with the protocol. On-chain, ANTH looks like any other perpetual. Off-chain, it is a contract template applied to a $2.3 trillion claim on unissued equity.
The structural novelty is not the asset. It is the governance surface. A perpetual is only as sound as its settlement specification, and here the specification is authored by whoever deploys the market. No clearinghouse. No exchange rulebook filing. No CFTC Part 38 core principles to satisfy. A smart contract with a human behind the oracle, and a UI that renders a number with four significant figures.
Start with the denominator, because almost nobody does. $2,327 implies roughly one billion shares outstanding. Anthropic has not disclosed its diluted cap table. If the true figure is 1.4 billion, the implied valuation is $3.26 trillion. If it is 700 million, it is $1.63 trillion. A two-fold swing in "valuation" is fully absorbed by a variable that is, literally, unknown to the people quoting the price. That is not price discovery. That is a ratio with a missing input, marked to the last trade.
Now the liquidity. Open interest of $28.25 million backing a claim on $2.327 trillion of equity is a ratio of roughly 82,000 to 1. CME equity index futures carry open interest that represents a meaningful fraction of the underlying exposure they reference — because a deliverable, or a cash-settlement index, exists to anchor them. Here the underlying does not exist in any transferable form. The turnover ratio (volume over open interest) comes to about 0.51, meaning the contract churns half its book daily. For a $2.3 trillion notional, that is not a market. That is a signal generator with a liquidation engine attached.
So what is actually being traded? Not equity. Not a forward on equity. It is a convexity claim on a narrative — closer in payoff shape to a long-dated option than to a stock, except it has no expiry, no strike, no deliverable, and a funding rate that quietly reprices it every hour.
That funding rate is the single most important missing number in the public data. A perpetual with no borrow market, no dividend, no custodied underlying, and no natural short — because who voluntarily shorts a story in a bull market — runs structurally positive funding. If it annualizes above 50%, longs are paying a premium to hold a lottery ticket, and the quoted price is a financing artifact rather than a valuation. Based on the token emission audits I ran during the 2021 Axie cycle, where a 72-hour reward-versus-inflation window was the entire trade, I have learned to treat the financing leg as the trade and the headline price as the marketing. Here, the financing leg is undisclosed.
Settlement is where pre-IPO perps die. Equity index futures cash-settle against an observable, published average. "Anthropic's IPO price" is observable in principle, but the contract must define which price: the bookbuild print, the first-day secondary close, a volume-weighted average over the first week, a direct listing reference, and whether a lockup discount applies. When I organized the filing-timeline tracker ahead of the 2024 spot Bitcoin ETF approvals, the entire edge was legal-timeline probability — the same discipline applies here. The rulebook is two disclosure tiers below what I would require before sizing a position, and nothing in the parsed data suggests it has been published at all.
The consensus read on $2.327 trillion is that the market is wrong. I read the number as the correct output of a market with no arbitrage channel — and that is the more uncomfortable conclusion. When there is no borrow, no deliverable, and no convergence mechanism, a mispricing does not get corrected; it gets financed. Arbitrage isn't available to the desk that spots the gap. It is available only to the desk that can hold the gap until the funding rate flips, which may be never.
The second-order effect matters more than the quote. HIP-3 converts the absence of an underlying into a product. Once that precedent holds, the wishlist writes itself — OpenAI, SpaceX, Stripe, whichever private name carries the loudest narrative. And at that point Hyperliquid's regulatory perimeter shifts from "crypto derivatives venue" to "venue listing derivatives on unregistered securities," which is a different legal animal with a different enforcement history. During the Terra collapse I built a failure model around algorithmic stablecoin decay rates; the relevant analogue here is not UST. It is the 2021-era SPV stacks. Those vehicles were not fraudulent. They were unpriceable. The difference is that they carried quarterly marks. This carries a mark and a margin call.
Watch three things, in this order. First, ANTH funding: sustained annualized funding above 50% means the quote is a financing artifact, and the unwind arrives through liquidations rather than convergence. Second, Anthropic's own response — a legal statement or a denial of authorization would force the venue to choose between delisting and defending a market the issuer never sanctioned, and the old precedent from the Tornado Cash designations suggests regulators do not need a working theory of harm to act. Third, the next HIP-3 deployment. If a second marquee private name lists within two quarters, the story stops being about Anthropic entirely.
We don't get to call something price discovery when the underlying has no transfer agent, no cap table, and no settlement rule anyone outside the deployer has read. What is happening is infrastructure learning to quote things that do not exist yet — and that capability will be aimed at something real long before Anthropic rings a bell.
