The Two-Trillion-Dollar Ghost: Debugging Hyperliquid's ANTH Pre-IPO Perpetual

On September 9, a market called ANTH printed $2,327. Multiply that by what a naive reader believes to be Anthropic's share count and you arrive at a $2.327 trillion valuation for a company that has never filed an S-1, never held a public earnings call, and has no listed float. The number is not a valuation. It is an output variable of a game with almost no players and no settlement anchor. Twenty-four-hour volume on the contract: $14.32 million. Open interest: $28.25 million. That is the entire corpus of evidence this article can work with, and it is already enough to determine that the market is doing something other than pricing Anthropic.

I have spent nine years watching crypto platforms invent price discovery for things that do not yet have a price. In 2017 I audited Bancor v1's dynamic fee formula forty hours before launch and found a rounding error that could have drained 15% of early liquidity under volatility. The developers told me it was negligible. It was not negligible. The unit economics of a thin book are always dismissed as negligible right up until they are not. ANTH is the same category of problem wearing an AI logo. Trust the hash, not the hype — and in this case, there is not even a clean hash to trust, because nobody has published the oracle spec.

Context: What HIP-3 Actually Opened

To understand why this contract exists, you have to understand what changed underneath it. Hyperliquid is a fully on-chain order book exchange with its own L1, and its HIP-3 framework allows third parties to deploy permissionless perpetual markets — anyone with the capital and the smart contract can list a perp on essentially anything, provided they meet the collateral and oracle requirements the framework specifies. Entropy, the deployer associated with this listing, used that pathway to create ANTH. The mechanical significance is larger than the ticker.

Until recently, the universe of assets a decentralized derivatives venue could list was bounded by what already existed on-chain. You could perp ETH, SOL, a basket of memecoins, and the occasional ecosystem token. You could not perp the equity of a private company, because a private company has no continuous price feed, no transferable shares, and no delivery mechanism. HIP-3 relaxes the first constraint by permitting third-party oracle design and relaxes the second and third by simply not requiring them. That is the core architectural move: HIP-3 does not solve the pre-IPO problem. It sidesteps it by removing the delivery obligation that would define a real claim.

This is not a new idea, only a newly permissionless one. FTX listed pre-IPO perpetual futures on companies like Coinbase and SpaceX before its collapse, and those contracts were cited by regulators as part of the broader failure to segregate and describe what customers actually held. Mirror Protocol minted synthetic equities (mAssets) on Terra and ran into the same wall: the synthetic tracked a price but conferred no ownership, no dividend, no vote, and no enforceable claim, which is precisely what a security is when you strip the wrapper and look at the economic substance. Each generation of this product promises price exposure without the securities-law overhead. Each generation discovers that price exposure to a security is a security, and the overhead is not optional.

What is genuinely new here is the distribution channel. The prior attempts ran through centralized venues with legal entities, KYC gates, and identifiable counterparties to subpoena. HIP-3 runs through a permissionless deployment on a decentralized exchange whose governance is opaque to most users and whose deployer is a pseudonymous-ish entity called Entropy. The question of who is legally responsible when this contract resolves badly is unanswered in the source material, and I checked — no registration status, no team disclosure, no audit reference, no oracle methodology, no custody arrangement. N/A — information not disclosed applies to essentially every diligence field a serious analyst would demand before touching this instrument.

So we have a market with no disclosed oracle, no disclosed settlement rule, no disclosed legal wrapper, no disclosed share count, and $28 million of open interest referencing a company valued, on paper, at more than the GDP of Italy. That is the object of study. Now let us take it apart.

Core: A Systematic Teardown

The Oracle Problem — Pricing Something That Does Not Trade

A perpetual contract requires a reference price. For ETH, the reference is trivial: thousands of venues, continuous two-sided quotes, arbitrage that closes any gap within milliseconds. For a pre-IPO equity, the reference price does not exist in any continuous sense. Anthropic's last known private marks sit in a range that is, depending on the round you cite and the date, somewhere between the high tens of billions and the low hundreds of billions. I am deliberately refusing to give a single precise figure because the source material does not establish one, and pretending otherwise would be the exact sin I am here to criticize.

What matters is the ratio. The ANTH-implied valuation of $2.327 trillion sits roughly an order of magnitude above any defensible private market mark. When a price feed produces a number an order of magnitude away from every observable reference, you do not have a price. You have a coordinate in a low-dimensional, thinly-populated state space that happens to be labeled "price." The oracle, whatever it is, is not tracking Anthropic. It is tracking the marginal trade on a book that a single mid-sized wallet can move.

Here is the mechanical chain. Suppose the oracle is a mark-price derived from the venue's own order book, which is the most common design on decentralized perps. Then the "price" is a function of the last trades on ANTH itself. There is no external anchor pulling it back to reality, because reality — the private equity of Anthropic — does not trade continuously and cannot be shorted by anyone outside a handful of venture funds bound by transfer restrictions. The feedback loop is closed. Price begets price. A buyer willing to pay $2,327 sets a mark that makes the position look profitable, which encourages more buying, which raises the mark. This is not a market discovering a value. It is a market manufacturing one.

The Settlement Problem — What Happens at the IPO

A perpetual has no expiry, which sounds like a solution to the settlement problem and is actually an evasion of it. Eventually one of three things happens: Anthropic IPOs, Anthropic gets acquired or stays private forever, or the market dies. The contract must specify what occurs in each case. The source material specifies none of them.

Walk through the branches. If Anthropic IPOs at a price far below the implied $2,327 — and remember, the private marks imply a price an order of magnitude lower — then every long on ANTH is catastrophically underwater and the settlement price must be defined by something. If it is defined by the IPO print, longs are wiped. If it is defined by a lagging average, longs get a grace period that only delays the wipe. If it is undefined, the market enters a governance fight that determines the outcome, which converts a derivative into a discretionary dispute. A perpetual with an undefined terminal condition is not a financial instrument. It is a governance token for a single future argument.

There is a second-order risk that the bulls systematically underweight. Pre-IPO perpetuals have no final anchor until the anchor event actually occurs. That means the contract can persist for years, accruing funding, oscillating on narrative, and never once being disciplined by the thing it claims to track. The longer it lives without settlement, the more the mark drifts from any fundamental, and the more painful the eventual convergence. This is the inverse of the Terra-UST loop I documented in early 2022, where the seigniorage model required exponential demand growth that was mathematically unobtainable in a saturated market. Here, the loop requires a terminal price that may simply never be discoverable while the contract is live. I published three papers on UST and the regulator stayed silent until $40 billion evaporated. I am not predicting the same magnitude here, because the market is tiny. I am predicting the same structure: a mechanism that only works if an assumption holds, where the assumption is not disclosed and cannot be verified from the outside.

The Liquidity Problem — $28 Million Pretending to Be $2.3 Trillion

Consider the denominators. Open interest of $28.25 million against an implied equity value of $2.327 trillion. The ratio of tradeable notional to referenced value is roughly 1.2 parts per 100,000. For comparison, a liquid equity perp on a major venue carries open interest that is a meaningful fraction of the underlying's real float. Here, the float is a rounding error against the thing it purportedly prices.

What does that ratio predict? It predicts that the price is set by the marginal participant, not by consensus, and that the marginal participant in a $14 million daily volume book can be one person. In a book that thin, a $2 million buy can move the mark several percent, and the position that benefits is whoever was already long. The classic structure — pull the price, force the shorts to pay funding or liquidate, distribute into the squeeze — is not hypothetical at this scale. It is the expected behavior of a rational, well-capitalized actor with information about order flow. I have watched this pattern in NFT floor mechanics (where a single sweep of three listings reprices an entire collection's floor), in low-cap token launches, and in the 2020 yield farms whose 80% APYs were token emissions masquerading as revenue. The mechanism is identical every time: a small circulating float plus a narrative generates a price that everyone quotes and nobody can exit at.

$14.32 million of daily volume is not price discovery. It is price suggestion. The distinction matters because readers will see the $2,327 print and treat it as information about Anthropic. It is not information about Anthropic. It is information about a very small number of people making very small, very leveraged bets on a narrative that happens to be adjacent to Anthropic.

The Funding Rate Problem — Arbitrary Numbers Wearing Mathematical Costumes

Perpetuals maintain their peg to the underlying through the funding rate, a periodic payment between longs and shorts designed to pull the mark toward the index. In a normal market, funding is bounded by arbitrage: if funding runs too positive, arbitrageurs short the perp and long the spot, collecting the spread and compressing it. The spot here is Anthropic equity, which cannot be bought, cannot be transferred, cannot be shorted by anyone except a narrow set of insiders under lockup. The arbitrage leg does not exist. Therefore the funding rate has no natural upper bound.

This is where my skepticism about interest rate models becomes directly relevant. I have argued for years that the interest rate curves on Compound and Aave are not derived from real market supply and demand — they are governance-selected parameters tuned by committee, dressed in the language of utilization curves. The same critique applies with more force to ANTH's funding. When you remove the arbitrage that would otherwise anchor a rate to economic reality, what remains is a number that governance sets and a market endures. A perpetually positive funding rate — the source material flags the "annualized >50%" scenario as a watch signal — does not mean longs are right. It means shorts are trapped and longs are paying to keep them trapped, or longs are trapped and paying to stay in. Either way, the rate is a measure of structural imbalance, not of value.

The Regulatory Problem — A Registered Security in an Unregistered Wrapper

Here is the most uncomfortable paragraph in this article, and I will not soften it. Strip away the word "perpetual" and describe the economic substance: a participant takes a position that rises and falls with the equity value of a specific, named, private company. That participant has no vote, no dividend, no information rights, and no claim on assets, but they have directional economic exposure to a security. Under the framework the SEC has applied for a decade — the Howey analysis that looks at investment of money, in a common enterprise, with expectation of profit, derived from the efforts of others — this is a security. It does not matter that it is wrapped as a derivative, and it does not matter that the venue is decentralized.

FTX learned this the hard way. The pre-IPO perpetuals were not what killed FTX, but they were part of the pattern of offering products whose legal character was never adequately disclosed, and the regulatory response afterward narrowed the aperture for every venue that followed. The CFTC has asserted jurisdiction over off-exchange derivatives in digital assets. The SEC has been aggressive on anything resembling a tokenized equity. A decentralized exchange cannot outrun the fact that the economic exposure is to a US-regulated security, and the deployer — Entropy, whoever that is — is the natural defendant. I flagged this exact gap in my Terra work: the technical reality moves faster than the regulatory apparatus, but the apparatus arrives eventually, and when it does, the question is who is holding the position. Here, nobody in the source material is named as holding it.

The Distribution Problem — HIP-3 Is a Distribution Play, Not a Technical One

I have a standing view about Layer 2 that belongs here in slightly different clothes. The real difference between the OP Stack and the ZK Stack was never the cryptography; it was who could convince more projects to deploy chains first. Distribution, not technology, decides winners in infrastructure. HIP-3 is the same game applied to derivatives venues. The technical capability to list arbitrary perps is unremarkable — builders have had the primitives for years. What is remarkable is being first to distribution: the first venue to establish the habit among traders that "if you want pre-IPO synthetic exposure, you go here."

That is the actual bet. Whoever successfully seeds the category captures the flow. If HIP-3 gets OpenAI, SpaceX, Stripe markets running and traders normalize the pattern, Hyperliquid becomes the default venue for a new asset class, and the fee revenue from that habit is durable even if any single market (like ANTH) is a distortion. The ANTH market is not the product. It is the proof-of-concept, the marketing spend, the loss leader. Read the launch as a distribution experiment, not a valuation event, and the source material's "AI + crypto narrative fusion" opportunity reads far more coherently.

Contrarian: What the Bulls Actually Got Right

The easy move here is to dismiss everything. That would be intellectually lazy, and the bulls have three points that deserve to be taken seriously rather than waved away.

First: the price is wrong, but the category is real. Pre-IPO exposure is genuinely scarce. Retail investors in most jurisdictions cannot buy Anthropic, OpenAI, or SpaceX at the private stage, and a continuous, 24/7, globally-accessible price signal for private equity is a real product that traditional finance has been unable to deliver because of transfer restrictions and accredited-investor rules. The demand is not manufactured. When I studied the NFT metadata fragility in 2021, I noted that 60% of top collections hosted images on centralized AWS servers, and I argued the ownership claims were fragile. I was right about the fragility and wrong to imply the demand was fake. The demand was real; the infrastructure was bad. Same pattern here. The appetite for pre-IPO exposure is real. The infrastructure — oracle, settlement, legal wrapper — is where the rot is.

Second: permissionless listing is a genuine architectural advance. I have spent years documenting centralized failure points in nominally decentralized systems. HIP-3 removes one of them: the gatekeeper who decides what can be listed. That is not nothing. A venue that requires no permission to list a new market is structurally more resilient to the kind of regulatory capture and selective listing that centralized exchanges practice, because there is no single compliance officer to pressure. The cost of that resilience is exactly the mess we are analyzing — no diligence, no disclosure, no accountability. But the bulls are right that the tradeoff is real and that the permissionless side has value.

Third — and this is the one the critique-leaning crowd misses — the fee revenue argument is real, and it is the same argument that saved Bitcoin's security model. I have argued that Ordinals and inscriptions injected new narrative and fee revenue into Bitcoin, and that without the inscription wave, Bitcoin's security budget would already be in visible distress. The mechanism is general: a base layer needs something to generate fee demand, and novelty assets generate fee demand. Pre-IPO perps generate trading fees, funding, and liquidation revenue on Hyperliquid. If the category sticks, that fee flow is genuine revenue, not emissions. The bulls' implicit case — that new asset classes subsidize the venue layer — is structurally sound even when the individual instrument is broken. Where they go wrong is treating the fee revenue as validation of the price. It is validation of the venue. Those are different claims, and conflating them is how people lose money.

The honest summary of the contrarian position: the bulls identified a real product and a real revenue mechanism, and then made one unfounded leap — from "this venue is doing something valuable" to "this price means something." Debug the intent, not just the code. The intent here is a profitable new asset class. The code, or the absence of it, is a contract that cannot be settled, cannot be arbitraged, and cannot be disclosed.

Takeaway: What to Watch, and What the Number Will Do

The ANTH market is a stress test for a hypothesis that will define the next two years of decentralized derivatives: can permissionless listing of non-crypto assets survive contact with settlement and regulation? If it can, Hyperliquid's HIP-3 becomes a template and the category scales. If it cannot, ANTH becomes the cautionary precedent that regulators and lawyers cite when the next venue tries.

Three signals determine which path we are on. Watch the funding rate: if it pins to the extreme positive and stays there, the contract is structurally trapped and the mark is meaningless. Watch for an official Anthropic response: a legal statement disclaiming authorization would be existential for the market, because it converts a quiet regulatory question into an active claim of unauthorized securities exposure. And watch the next HIP-3 deployment: if a second and third pre-IPO market appear with the same absence of disclosed oracle and settlement rules, the pattern is confirmed as a product design choice, not an oversight, and the accountability question becomes urgent.

I am not going to tell you the number is wrong, because "wrong" implies a correct number exists that we could agree on. There is no correct number for the equity of a private company that does not trade. What I will tell you is that $2,327 is not a claim about Anthropic. It is a claim about the people currently holding the book, and the book is small enough to fit in a single room. Trust the hash, not the hype. In this market, we do not even have the hash. We have a number, a thin order book, and an argument that has not started yet. When it starts, someone will have to be named as a defendant. That name is the only settlement price that matters, and no one has published it.