The 8,700% Ghost: Anthropic's $2 Trillion IPO and the Derivative That Doesn't Exist

The funding rate on the Anthropic pre-IPO contract printed at 8,700% annualized. Let that settle for a second.

Most people read a number like that as hype — a headline, a joke, a rounding error in a casino. They are wrong. A funding rate is not a sentiment poll. It is the price of borrowing conviction. When it annualizes to four digits, the market is telling you that one side of the book has run out of counterparties. Not liquidity. Counterparties.

There is a product trading right now — on a venue that has published no settlement currency, no custodian, no legal wrapper, no expiry schedule — that lets you take exposure to equity in a company that has not priced its offering yet. The shares do not exist as a transferable security. The contract does. And the contract is charging 8,700% a year just to keep the long side breathing.

That is the story. Not Anthropic's model. Not Claude. The story is that synthetic exposure to a private AI lab has become the most crowded, least hedged trade in the entire crypto derivatives complex — and almost nobody who is long it understands what they actually own.

Context: Two Markets Wearing One Ticker

Anthropic is a Delaware corporation. Traditional equity. No token, no chain, no validator set. It is not a blockchain project, and I want that on the record before anything else, because the reporting around this story keeps fusing two things that share nothing but a name.

The company closed its Series H on May 28 at a $965 billion post-money valuation, led by Altimeter, Dragoneer, Greenoaks, and Sequoia. Four months later, the market is being asked to underwrite a $2 trillion IPO with a $100 billion raise — the largest primary offering ever attempted, larger than SpaceX's $85.7 billion, larger than Saudi Aramco. The filing itself is a 261-page S-1, roughly 80 pages of which are risk factors. That is a 30.6% disclosure ratio, conservative by big-tech standards. The lawyers did their job. Keep that in mind, because the lawyers are not the problem here.

The problem is the second market.

Market one: Anthropic's equity. Governed by the SEC, priced by institutional book-builders, settled in a regulated clearing system. Fundamentals matter. Diligence matters. The $2 trillion number lives here.

Market two: the pre-IPO derivative complex. Governed by nobody. Priced by whoever shows up. Settled — allegedly — by a platform that has disclosed nothing about how it would deliver equity in a company that, at the time of writing, has no public shares to deliver. The 8,700% number lives here.

These two markets share a ticker in people's heads. They share nothing else. Confusing them is how capital gets destroyed, and it is happening at scale right now. Crypto media are covering this as a blockchain story. It is not a blockchain story. It is a securities story with a blockchain costume, and the costume is thin.

Core: The Arithmetic, Then the Order Flow

Start with the arithmetic, because the arithmetic is the only thing that does not lie to you.

Anthropic's 2025 revenue is $4.6 billion. Its 2025 net loss is $42 billion. That is a revenue-to-loss ratio of roughly 0.11 — for every dollar of revenue, the company burns about nine dollars of cost. This is not unusual for a frontier lab mid-buildout; compute is a monster and the buildout is real. What is unusual is the price attached to it.

At the Series H mark of $965 billion, the company is priced at 210× trailing revenue. At the $2 trillion IPO ask, it is priced at 435× revenue.

Put that against public comparables and watch the regime break. Snowflake at its absolute peak never cleared 50× revenue. Datadog never cleared 40×. The most expensive software franchise in the history of the Nasdaq topped out at roughly one-ninth of what Anthropic is asking today. The gap between 50× and 435× is not a premium. It is a different pricing regime entirely — narrative pricing, not cash-flow pricing. There is no discounted cash flow model that produces 435×. There is only a story that does.

Now watch how the story gets built.

The headline is a "run-rate" of $47 billion, representing "1,000% growth." That run-rate is derived by annualizing a single month — May. Not Q2. Not the trailing six months. May, the one month the company chose to highlight, multiplied by twelve.

I have traded this move before. In the 2017 ICO cycle, projects annualized their best week and called it traction. The mechanics are identical: pick the window that flatters the slope, extrapolate linearly, and let the market price the extrapolation as if it were a signed contract. It is legal. It is also a signal. When a company reaches for the best single month rather than the quarter, you should assume May was the peak and Q2 was the fade. The selective window is the disclosure. The number they did not show you is the number that matters.

Then there is Q3. It does not exist in the filing. It will surface during the November roadshow — unaudited, at the moment of maximum pricing pressure. That timing is not accidental. If Q3's slope comes in below the May-implied run-rate, the book-building narrative has a hole in it, and the hole opens exactly when the company needs the price to hold. Watch for it. It is the single most important data point between now and pricing.

Layer in Nvidia. Nvidia is reportedly prepared to anchor $10 billion of the offering. Read that twice. Nvidia sells Anthropic the compute. Anthropic spends the raise on that compute. Nvidia takes an equity position funded by the same capital flow. This is a circular trade wearing a strategic-investment costume — upstream vendor, downstream customer, and anchor shareholder, all stacked on one balance sheet. Related-party disclosure obligations on that configuration are non-trivial, and any accountant who has modeled it knows the revenue-recognition questions write themselves. When the same dollar flows out of Nvidia, into Anthropic, and back into Nvidia, the income statement is doing something that the footnotes will eventually have to explain.

Now to the 8,700%.

A funding rate is the periodic payment exchanged between longs and shorts to keep a perpetual contract pegged to its reference. When it annualizes to 8,700%, it means longs are paying shorts 8,700% a year — except there are almost no shorts. The rate is high precisely because nobody will take the other side. That is not a liquid market. That is a one-way crowd paying a toll to stay in a room with a locked exit.

Here is the mechanical read, and I want to be precise, because this is where retail gets liquidated:

First — 8,700% annualized is almost certainly the artifact of a perpetual, not a borrow rate. Perpetuals have no maturity, so the rate floats to whatever clears the book. When the book is one-sided, the rate goes vertical. This tells you the instrument has no natural short supply — no market maker willing to warehouse the risk, no arbitrage desk able to hedge, because there is no underlying to hedge with.

Second — you cannot delta-hedge a contract whose underlying has no secondary market. A real pre-IPO position can be hedged with a forward, a total-return swap, an SPV interest. This contract cannot. There is no borrow, no locate, no shares to short. So the long side is structurally naked, and the only exit is to sell to a greater fool at a worse funding cost.

Third — the settlement mechanism is undisclosed. I went looking for it. Currency of settlement: not stated. Oracle: not stated. Expiry and roll schedule: not stated. Liquidation waterfall: not stated. Counterparty: not stated. When every one of those fields is blank, you are not trading a derivative. You are trading a promise.

The floor didn't hold — because there was never a floor. There is a reference price and a crowd, and between them sits a platform with admin keys and no published delivery rules. That is the entire architecture.

Let me put numbers on the crowding, because the crowding is the whole trade. At 8,700% annualized, a long pays roughly 23.8% of notional per month, or about 0.8% per day, simply to hold. A position that is directionally correct but flat for six weeks bleeds 33% of notional to funding alone. That is a trade that must be right quickly and cannot be right slowly. It is the signature of a momentum crowd, not an investment base. And momentum crowds do not exit — they stampede.

Now the regulatory read, because this is where the two markets collide hardest.

Apply the Howey test to the on-chain product. Money invested: yes. Common enterprise: yes — the enterprise is Anthropic. Expectation of profit: yes — an 8,700% funding rate is the market explicitly stating it. Profit from the efforts of others: yes — entirely dependent on a third party's IPO. All four prongs land. The "blockchain stock" is, on its face, an unregistered security wrapped in an unregistered derivative — a double exposure with no compliance wrapper of any kind. No custody arrangement disclosed. No legal settlement entity named. No KYC framework published. The SEC and CFTC both have clean jurisdiction here, and pre-IPO derivatives are among the first categories global regulators move on — MiCA in Europe, enforcement actions in the US. The product exists in the gap between the two regimes, and gaps close.

Contrarian: Who Is Actually On The Other Side

Here is where I part ways with the crowd — including the crowd that is bearish.

Everyone is now screaming that Anthropic is overvalued. Fine. That is the easy call, and on a five-year horizon it is probably right. But it is not the trade. The trade is understanding who sits on the other side of each market, because the two sides are not the same animal.

In the equity market, the smart money is not buying the $2 trillion print. It is the anchor allocators — the sovereign funds and crossover funds who get allocation at the IPO price, not at the open. They will flip into retail demand on day one, capture the first-day pop, and be gone long before the lockup math matters. The retail buyer at the open is buying the exit liquidity of the allocator. That is not a conspiracy; it is the standard mechanics of every large IPO ever printed. The asymmetry is structural.

In the derivatives market, the smart money is not long at 8,700%. The smart money is short the funding — if it can find a venue that will let it, and if it can survive the platform risk of a counterparty that might simply stop honoring withdrawals. That is the real asymmetry: the rate is fat enough to pay a short, but the venue risk is fat enough to wipe the short. When both sides of a trade are uninvestable for different reasons, the market is not a market. It is a trap with two doors.

The blind spot is this: retail thinks it is trading Anthropic. It is not. It is trading the gap between Anthropic's private valuation and whatever the platform decides the reference price is — a gap the platform controls, on a schedule the platform sets, with settlement terms the platform can rewrite. You are not long Anthropic. You are long a promise about Anthropic, issued by an undisclosed counterparty, priced by a crowd that cannot hedge. The equity story and the derivative story are related by marketing, not by mechanics.

Takeaway: Watch Three Numbers, Not The Narrative

Watch three numbers. Ignore everything else.

One — the Q3 unaudited slope in the November roadshow. If it prints below the May-implied run-rate, expect the pricing to be guided down, and expect the 8,700% contract to gap through whatever thin bid exists.

Two — the funding rate itself. A collapse from 8,700% toward sub-1,000% is not relief. It is the crowd exiting, and it usually prints before the reference asset turns. Extremes in funding are leading indicators of tops, not coincident ones. When the toll gets this expensive, the room is already on fire.

Three — any S-1 amendment disclosing the Nvidia anchor terms. Confirmation of the circular structure turns a strategy story into a related-party disclosure story, and that is when institutional demand starts asking questions it did not ask in the roadshow.

The equity may well price. The derivative will not survive contact with a real secondary market. When the shares actually trade and a genuine hedge becomes possible, a contract charging 8,700% to hold a long has exactly one destination, and it is not up.

The floor didn't hold because it was never load-bearing.