At 09:41 UTC on September 12, a new ticker went live on Hyperliquid.
It read: Anthropic. Pre-IPO. Perpetual.
No S-1. No roadshow. No SEC registration. Just a contract market, spun up by a builder called Entropy under Hyperliquid's HIP-3 standard, letting anyone with a wallet take a long or short position on the private valuation of a company that has never printed a single public share.
Two hours later, the feed attached to that market showed Anthropic at $2.159 trillion.
I have spent the better part of five years front-running these venues from the front lines of the hype cycle. I know what a fat finger looks like. I know what a mislabeled data field looks like. And I know what narrative inflation looks like when somebody hides it behind a decimal point. This was all three, stacked on top of each other.
So before anyone sizes a position, let's do the boring work and read the tape.
Context: the plumbing is the story
HIP-3 is Hyperliquid's builder-deployed market standard. Third parties stake HYPE, define a contract specification, and launch a custom perpetual or expiry market on top of Hyperliquid's order book. Hyperliquid supplies the rail — sub-second finality, an on-chain matching engine, and the HLP vault standing behind it as a backstop market maker. The deployer supplies the product.
That division matters more than it first appears. Hyperliquid's core perpetuals have processed hundreds of billions in volume and have been stress-tested by every ugly weekend of the last two years. HIP-3 markets inherit the execution layer, but they inherit none of the credibility. Everything specific to this Anthropic ticker — the margin schedule, the funding curve, the expiry mechanics, and most critically the settlement reference — was authored by Entropy, a deployer whose team composition is not disclosed anywhere I could locate.
Building a Pre-IPO market is not technically difficult. You need three things: a book, a funding mechanism, and a number to settle against. The first two are commodity infrastructure in 2025. The third is the entire problem.
Anthropic's actual private valuation sits in the $180 to $200 billion band based on its last primary rounds. Secondary platforms like Forge Global and SharesPost have brokered paper in that zone for years, carrying all the friction that implies: accredited investor gates, transfer restrictions, rights of first refusal, and settlement windows measured in weeks rather than blocks.
Strip that away and you get a clean cash-settled contract paying out on a published number. Faster. Cheaper. And entirely dependent on who is doing the publishing.
Core: two numbers, same venue, same day
Now the forensics.
Track two figures that appeared on the same platform, on the same date, in the same field.
Anthropic: $2.159 trillion. OpenAI: $164 million.
If those carry the same unit, Entropy has discovered that OpenAI is worth roughly 0.008 percent of Anthropic. That would make Anthropic the most valuable enterprise in human history by a multiple of five over anything that currently exists, and it would make OpenAI a modest regional credit union.
Neither number survives contact with reality. The $2.159 trillion figure sits at least ten times above Anthropic's last-round valuation band. The $164 million figure is off by four orders of magnitude in the opposite direction and is almost certainly open interest or rolling turnover that got mislabeled as valuation somewhere in the data pipeline.
That is not a rounding error. That is a broken instrument, and it is the first thing anyone evaluating this market needs to understand.
Here is what the picture looks like once you strip the corrupted field and read volume instead. The Anthropic market carried roughly $28.19 million in contract notional and about $6.74 million in rolling turnover. The OpenAI market, deployed by the same builder, carried slightly more turnover at roughly $7.67 million.
Two markets. Under eight million dollars of daily turnover each. That is a rounding error inside a rounding error. For scale, a single mid-tier token listing on a tier-one exchange clears more than that in an hour.
I opened the book myself to check depth. I sized a modest clip — four figures — and watched the spread widen past the point where any honest price discovery can occur. The quote you see is not a consensus about Anthropic. It is the outermost resting order in a book that cannot absorb real flow. Speed is the only currency that matters across most of crypto, but in a market this shallow, size is what kills you, and there is almost none of it to go around.
Which brings us to the technical heart of the matter, and the thing almost nobody is writing about.
A Pre-IPO contract's risk surface is not the price of the underlying. It is the definition of the underlying. Whoever writes the settlement formula writes the P&L.
Every DeFi veteran carries a scar from price feeds. In 2020 the wound was latency — an oracle updating four seconds late, a liquidation cascade triggered on stale data, a lending market drained because the feed quoted a price that no longer existed. That problem has been engineered down into a manageable nuisance.
The 2025 version is semantic. There is no public market price for Anthropic. There is no consolidated tape. There is no last trade to print. So Entropy must define one: some blend of last primary round, secondary transfer marks, and analyst composite, refreshed on some schedule, published through some mechanism, adjudicated by someone.
None of that is disclosed. From what is publicly visible, the market documentation does not state which reference source anchors settlement, how frequently it refreshes, what happens if Anthropic closes a new round at a different price mid-contract, or who arbitrates a dispute at expiry.
That is not a footnote. That is the product.
Notice also the language in the market's own promotional framing: valuation that "peaks and retracts." In a book this thin, a peak and a retraction is not a market repricing a company. It is one participant moving the outermost quote and another participant reacting to it. You are watching two people in a hallway, not an index.
There is a structural cost buried here too. HIP-3 deployers stake HYPE to launch markets, which means Entropy has capital locked into this specific ticker. That cuts in two directions. It signals skin in the game. It also means the deployer's economic interest lies in the market continuing to trade, not necessarily in the market being priced correctly. When the entity defining settlement is also the entity that needs the market to survive, you have a governance problem before you have a trading problem.
One more consumer-protection point that keeps getting lost. If you buy this contract, you do not own anything. No equity, no voting rights, no claim on future cash flows, no dividend-equivalent exposure. You hold a cash-settled derivative that pays out on a number. Traders who believe they are "getting in early on Anthropic" are not. They are betting on a reference value they cannot audit, published by an entity they cannot identify, settled under rules they cannot read.
Contrarian: the question everyone is asking is the wrong one
Everyone following this story is chasing a single question: is Anthropic close to an IPO? Wrong question. The interesting signal is hiding in the relationship between the two markets.
OpenAI's book is deeper than Anthropic's — slightly more turnover, on a ticker that generated a fraction of the headline energy this week. If flow were thesis-driven, if real money were positioning around Anthropic's listing timeline, the Anthropic market should dominate. It does not.
That tells you the flow is novelty-driven, not thesis-driven. Capital is not expressing a view on private market valuations. It is rotating toward whatever ticker is freshest. That is a rotation trade wearing a fundamental costume, and it will rotate again the moment a new name — SpaceX, Stripe, whoever gets the nod next — comes online. Chasing the alpha, one block at a time, looks identical to chasing the novelty, one ticker at a time.
Second blind spot: almost everyone is modeling this as trading risk. It is labeling risk.
Strip the mechanics back and ask what this contract actually is. Money is contributed. There is a common enterprise. There is an expectation of profit. And the profit derives from the efforts of others — specifically, from the operating performance of a company whose management team is doing the work. That structure sits squarely inside the Howey frame.
US regulators looking at a cash-settled derivative on an unlisted equity valuation have two plausible homes for it: a security-based swap under the SEC, or an event contract under the CFTC. Neither is a comfortable place to be, and the market sits in the gap between them.
What matters is the blast radius. If a regulator rules that HIP-3 Pre-IPO markets are unregistered security-based swaps, the entire category gets pulled at once — not one ticker. And there is a second, quieter problem. Anthropic and OpenAI have almost certainly not authorized anyone to run derivatives on their own valuation. A company preparing a real listing does not want an offshore perpetual market printing a fabricated number next to its name, and the legal machinery available to make that stop is well-worn.
Takeaway: watch three things, ignore the rest
First, whether Entropy publishes a settlement specification — the reference source, the refresh cadence, the dispute process, the expiry adjudication. Until that page exists in public, no position in this market is investable. It is only gamblable.
Second, whether US regulators say anything at all. Silence here is not tolerance. It is lag, and lag resolves.
Third, and most useful: watch contract notional. If it crosses into the hundreds of millions, real demand exists and the category has legs. If it keeps oscillating in the single-digit millions, then this was never a market. It was a headline with a matching engine attached.
Surviving the winter to plant for spring applies to products as much as to portfolios. The platforms that win the next cycle will not be the ones that listed the flashiest private ticker first. They will be the ones that published the boring settlement rulebook before anyone thought to ask for it — live from the edge of the unknown, one page of documentation at a time.
