There is a number on Binance right now that reads approximately 2,147.
It is quoted against a company that has never sold a public share. It carries no unit I can verify, no stated assumption about shares outstanding, no index methodology. It is the price of an Anthropic pre-IPO perpetual contract, and roughly $79 million of open interest sits behind it. Add the OpenAI equivalent and the figure crosses $160 million.
Six months ago, this market was worth about one million dollars. I have spent twenty-four years reading code and reading the stories people tell about code, and I have learned that when a number grows two orders of magnitude faster than its justification, the justification is usually the last thing anyone checked. So I went looking for the specification. There isn't one.
Trust the protocol, not the pitch. Here there is no protocol. Only a pitch.
Anthropic is private. In June it filed a confidential draft registration statement with the SEC. The Wall Street Journal reported an IPO timeline that slipped from October to November. There is no prospectus, no audited public financials, no valuation any regulator has signed off on. The company has said nothing about Binance's contract โ not a denial, not an endorsement. Silence is the loudest audit.
Into that silence, a centralized exchange built a financial product. Not an on-chain asset. Not a token. A cash-settled perpetual futures contract that references the eventual value of a private AI company. Binance Research describes it plainly: no underlying shares are required, and traders are taking opposing positions on where the company's final value lands.
That sentence is the entire product. Read it twice.
It is worth being precise about what "crypto" means here, because the word is doing a great deal of work. There is no chain, no validator set, no smart contract open to inspection. There is an order book, a matching engine, and a risk desk. The blockchain is not in the product. It is in the brand.
Here is what I want to establish technically, because the marketing is running well ahead of the mechanism.
A perpetual futures contract normally works because of a tension between two prices. There is the contract price, and there is the spot price of the thing it tracks. When they diverge, arbitrageurs do the boring, unglamorous work of hauling them back together, and the funding rate โ a periodic payment between longs and shorts โ accelerates that convergence by making the wrong side expensive to hold. That mechanism is not decorative. It is the entire reason a perpetual can be called a price at all. Strip out the spot leg and the funding rate has nothing to arbitrage against. It becomes a sentiment poll with leverage.
Anthropic has no spot leg. There is no share to buy, no token to redeem, no deliverable at settlement. CoinGlass lists the contract with zero circulating supply and no spot trading. Every position is matched against another position, marked to a number that exists only because two counterparties agreed on it this morning.
Here is the missing signal. In a healthy perpetual, the funding rate is the loudest thing on the screen โ it tells you which side is crowded and how badly. In this contract, funding has nothing to correct toward, so it stops being a measurement of imbalance and becomes a measurement of how much one side will pay for exposure to a story. That is still information. It is simply not the information anyone claims to be publishing. The coverage does not disclose the funding rate at all, which is the first number I would have asked for.
Then there is scale. Roughly $79 million in open interest against an anticipated valuation in the trillion-dollar range is a ratio of about four hundredths of one percent. A market that small does not discover the price of anything. It issues an opinion and calls it a price. I have watched that arithmetic justify "the market is pricing in X" for years, and it is almost always the wrong denominator being ignored on purpose.
The open interest argument has also been read backwards in most of the coverage I have seen. Open interest is not a directional opinion. Every contract has a long and a short. When OI climbs from near zero to nine figures, it tells you participation and leverage increased. It says nothing about whether anyone is bullish. One of the sources in the original reporting made exactly this point, and it deserves more credit than it received โ it is the most honest sentence in the whole story.
I want to flag a data problem too, because I audit and audits are unkind. One widely repeated figure states that open interest grew from about $1 million in April to more than $160 million with a month-over-month increase of 179 percent. One million to $160 million is a multiple of 160. Either the base figure is wrong or the percentage describes a different slice entirely. When a market's headline growth number contradicts itself inside the same paragraph, every other self-reported metric deserves the same scrutiny. I would not build an argument on it.
There is one more thing worth borrowing from the traditional world. Pre-IPO secondaries on venues like Forge or EquityZen are ugly, illiquid, and heavily restricted โ and they possess something this contract does not. They trade actual shares. They carry information rights, transfer restrictions, a preference stack, a real cap table behind them. Their prices move slowly and unreliably, which is exactly why they are not exciting. The friction is the feature. It is what separates a valuation from a vibe.
Structurally, this is a cash-settled contract with no underlying, cleared by a centralized exchange that acts as counterparty, sets the margin rules, picks the price source, and can change or delist the instrument unilaterally. That is not a knock on Binance specifically. It is simply what the architecture is: trust, relocated from a blockchain to a company. Anyone who has been here long enough should recognize the trade.
It also raises the settlement question nobody has answered. When this contract settles, against what reference? There is no registered benchmark for Anthropic's equity, no published share count, no disclosure schedule. Someone has to define the number, and that someone is the venue that profits from the market existing.
On the regulatory side, the Howey analysis is not close. Money invested, a common enterprise, expectation of profit, derived from the efforts of others โ the four prongs land on this product with uncomfortable neatness. The referenced asset is an unregistered security in a company that has registered nothing publicly. A synthetic derivative tracking a private company's future equity is, functionally, a wager on an event the securities regulator has not permitted anyone to wager on yet.
Contrast the tokenized equities that already exist โ Backed, xStocks, and their peers. Those instruments are backed by actual securities held in custody, and they inherit a compliance chain with them. The Anthropic perpetual inherits nothing. It belongs to the same genre only to a casual eye.
Meanwhile Binance holds roughly 40 percent of activity in this category. Concentration that high in an instrument with no anchor is a liquidity advantage and a fragility at once. One exchange's listing decision is the difference between a market and a footnote.
The template generalizes trivially, which is the part that should worry people. Any private company with a loud enough story can be wrapped in the same structure: list a perpetual, take the other side of both crowds, collect fees, and let the referenced entity remain a stranger. The first one is a novelty. The fifth is infrastructure for a parallel valuation layer that answers to no disclosure regime.
Here is where I part company with most of the commentary, including the sympathetic kind.
The prevailing worry is that retail traders will lose money. They will. Zero-sum with fees attached is negative-sum, and the losing half of this trade will not appear in any headline celebrating the open interest record. That is the visible harm. It is real, and it is not the worst thing here.
The worse thing is epistemic.
This contract manufactures a number. The number is denominated in a familiar-looking figure. Journalists, analysts, and eventually some AI newsletter will cite it as evidence of "what the market thinks Anthropic is worth." They will do it because it is convenient and because nobody will stop them. What they will be citing is the residue of two crowds of leverage agreeing with themselves โ no balance sheet, no disclosed terms, no anchor, no audit. The danger of a synthetic market is not that traders lose. It is that its output becomes indistinguishable, in the public record, from information.
Code doesn't care whether a number means anything. It executes the assumption it was handed and waits.
I have watched this movie before. In 2017 I spent three months auditing the Ethereum Classic fork's immutable ledger logic and filed twelve critiques on GitHub, almost none of them about bugs. They were about governance โ about what a system commits to when it refuses to change. The lesson was not that code lies. It was that code executes whatever assumptions are buried inside it, silently, and at scale. Here the buried assumption is that a number floating free of any reference is still a measurement. It is not. It is a quote. "Price discovery" is a claim about mechanism. This instrument has a mechanism for expression, not for discovery.
The last time I sat with a family office in Abu Dhabi, walking them through custody and compliance on a $10 million allocation, the entire conversation was provenance. Where did the asset come from, who holds it, what happens if it fails. That is the correct instinct. This market is engineered to route around it.
So I am not arguing the contract should not exist. I am arguing about what we call it. Call it speculation, because that is honest. Do not call it price discovery, because that name requires an anchor it will never have.
The real question lands in November. If Anthropic lists, this contract settles into something with a prospectus behind it, and the synthetic era for that name ends in anticlimax. If the IPO slips again, its holders learn what it means to hold a claim on nothing. Either way, the pattern โ exchanges wrapping the most anticipated private companies in derivatives before those companies have chosen to be public โ will not stop with Anthropic.
Which leaves the thing I keep returning to. Anthropic never said yes to this. Somewhere in a database, a market is pricing a company that has not consented to being priced, and doing it out loud. What happens to a business, and to a public, when a shadow price arrives before the real one? That is not a trading question.
That is the question.