Over the past seven days, a market that did not exist two weeks ago repriced Anthropic — a company whose last private round valued it near $184 billion — at roughly $2.159 trillion, then watched that number slide 6.1 percent in three sessions. The venue is not a secondary desk in Menlo Park. It is Hyperliquid, and the instrument is not equity. It is a contract, deployed by an entity calling itself Entropy, through a mechanism designated HIP-3, carrying $28.19 million in open interest against that thirteen-figure mark.
Nothing in that sentence would survive a first-pass diligence review. The striking part is not the valuation itself. It is the ratio. A market whose entire open interest equals 0.13 percent of the value it claims to represent is not a market in the ordinary sense. It is a signal generator wearing a market's clothes — and over the next few months, that signal will shape how a great many people think about the price of artificial intelligence.
Private secondary markets have always been structurally strange. SecondMarket and SharesPost built the first serious venues for pre-IPO equity in the late 2000s; Forge Global and a sprawling SPV ecosystem inherited the model and industrialized it. Every one of them shares three properties: gated access, monthly marks, and a legal apparatus mediating each transfer. The illiquidity is not a defect. It is the load-bearing wall — the reason a private valuation can sit unchanged for a year while the underlying business churns beneath it.
The 2024 real-world-asset narrative pushed tokenization downstream of that wall, but only for the easy assets. Tokenized Treasury funds, money-market sleeves, real estate: all of them have a published reference price and a settlement convention. What nobody had touched was the hardest RWA class of all — equity in a company that has never filed, has no ticker, and has publicly suggested it may never list.
HIP-3 is that touch. Entropy deployed an Anthropic pre-IPO market on Hyperliquid; an OpenAI market followed, currently marked near $164 million. Hyperliquid itself needs little introduction — a perp DEX that has spent two years converting trading volume into infrastructure ambitions. HIP-3, read in context, looks like a market-creation primitive: a template for pointing the same contract structure at any underlying a deployer selects. Whether that reading is correct depends on documentation that does not appear to exist.
The institutional framing I worked on in 2024, translating crypto primitives for asset managers in Washington, taught me one durable thing: institutions do not buy narratives, they buy reference prices. A pension consultant will accept a novel instrument if it can be marked against something auditable. They will not accept a novel instrument whose mark is produced by the instrument itself. That is the entire distance between tokenized Treasuries, which cleared institutional scrutiny, and a pre-IPO contract with no index.
Every token is a vote for a future we haven't defined. This one is voting on a company that has not agreed to be voted on.
Begin with the number, because the number is the trap.
$2.159 trillion is not Anthropic's valuation. It is the aggregate notional value of open contracts referencing something labeled as Anthropic equity. The distinction is not pedantic. In crude oil futures, notional dwarfs the physical market by orders of magnitude, and no one mistakes it for the price of oil — because there is a reference index, a delivery mechanism, and a settlement process that forces the contract to converge on reality. Strip those three things away and "notional value" degrades into a measure of how much people have agreed to pretend.
The tell is the divergence between the two markets sharing the same venue. If both were pricing genuine pre-IPO exposure to two companies whose real private valuations differ by a factor of roughly 0.85, their marks should differ by roughly 0.85. They differ by a factor of about 13,000. That is not a pricing disagreement. That is two disconnected pools, each with its own shallow order book, neither anchored to anything external. Under those conditions the first marginal buyer sets the mark, and every subsequent participant inherits whatever fiction that buyer chose.
I have seen this shape before, from the other side of the screen. In 2018 I spent three months auditing the 0x protocol's v2 contracts line by line — not because I expected to find fraud, but because I wanted to know whether the math held at its boundaries. I submitted seven edge-case findings, including a reentrancy flaw in the filler function. The lesson was not that the code was bad. The lesson was that a contract's honesty is fully determined by what it does at the moment it is asked to settle, and nowhere else. Everything before that moment is narrative.
So the question worth asking about HIP-3 is not what it is worth. It is: what happens when this contract is asked to settle?
Nobody has published an answer. There is no stated expiry, no published reference index, no methodology for the mark. Dario Amodei has said on the record that Anthropic may never go public. If no listing occurs, does the contract converge on a funding round, a 409A mark, a committee's judgment, or nothing at all? If the answer is nothing, then this is not a pre-IPO market. It is a perpetual swap with a story attached.
The microstructure agrees. Open interest of $28.19 million against $6.74 million in volume implies a turnover ratio near 23.9 percent in a market roughly three days old. Genuine price discovery produces churn, because participants disagree and trade against each other. A shallow book produces almost none, because there is no one to disagree with. A 6.1 percent drawdown across three sessions looks dramatic until you remember that private secondary marks move on a quarterly cadence, in single-digit percentages, for reasons that get documented. A three-day move of that size is not repricing. It is one position finding the exit.
The structural problem compounds at the securities layer. Applying the Howey framework here is not a legal opinion; it is an architectural observation. Money is invested, in a common enterprise, with an expectation of profit derived from the efforts of others — specifically, from Anthropic's management and the timing of its public offering. Three of four prongs are satisfied on inspection, and the fourth, the profit expectation, is satisfied by the fact that no one would buy these contracts for any other reason. An instrument that only makes sense as a bet on someone else's future filing is functionally a security in search of a registrar. There is no registrar. There is also no identified counterparty: Entropy, the deployer, has published no team, no jurisdiction, no contact, and no audit. A market with anonymous operators and anonymous participants, referencing a company that does not know it is being traded, is not a gray area. It is the absence of an area.
The core insight is this: HIP-3 is not pricing Anthropic. It is pricing the AI narrative and using Anthropic as its ticker, because Anthropic is the only private AI company with a name recognizable enough to function as one. The $2.159 trillion figure is not a methodological error. It is an accurate measurement of something else entirely — the intensity of demand for exposure to a story that has no listed instrument anywhere in the world.
A corollary matters more. A market's legitimacy is established at settlement, not at listing. Every claim made about HIP-3 over the next quarter will be beside the point until the first contract is asked a question about the real world and has to answer it.
The prevailing reads on HIP-3 split into two camps: bold RWA innovation on one side, transparent scam on the other. Both assume the market is attempting to be a pre-IPO market and is either succeeding or failing at the attempt. Consider instead that the equity is incidental.
HIP-3, if it functions at all, is a market-factory template — a way to spin up a tradeable contract around any underlying a deployer chooses. Anthropic is the test payload. If the primitive survives contact with regulators and with a corporate event, the same template gets pointed at private credit, pre-revenue biotech, litigation outcomes, anything with a compelling story and no published reference price. Hyperliquid's downside if Anthropic blows up is a week of bad press and one dead market. Its upside if the template works is a general-purpose venue for assets that currently cannot be traded at all. That asymmetry explains the launch far better than any theory of misplaced equity demand.
The regulatory read is worth inverting too. The reflex is that the SEC will shut this down. My reading of a decade of regulation-by-enforcement is that it will not — not quickly, and not before the exit. Enforcement has operated as an implicit subsidy to offshore experimentation: instruments too small to prosecute and too novel to fit existing categories accumulate until a loss is large enough to force action. The action then arrives after retail has funded the exit, which makes the enforcement a post-mortem rather than a deterrent. Every token is a vote for a future we haven't priced, and the price of this one is being paid by whoever holds last.
The tidiest story available — pumped on the ninth, dumped by the twelfth — is also the least likely to be right. A 6.1 percent glide on $28 million of open interest is what a shallow book does when a single participant leaves. The more unsettling possibility is that nobody involved knows what the price should be, because nothing anchors it. When there is no anchor, the first buyer does not discover a price. He authors one.
The question is not whether HIP-3 survives. It is what the first settlement event looks like — the first time one of these contracts is asked about the real world and discovers it has no answer. Watch four things: whether Anthropic issues a statement at all; whether an independent audit of the HIP-3 contracts appears; whether open interest crosses $100 million or falls below $10 million; and whether the template is next pointed at a company that actually intends to file.
If that last one happens, the experiment stops being a curiosity and becomes a question about what a price is for.
Every token is a vote for a future we haven't seen. Most are cast before anyone has read the terms.
