The Unanchored Mark: Anthropic's Pre-IPO Perpetuals and the Quiet Test of Crypto Price Discovery

Over the past thirty days, a synthetic instrument that exists in no regulated market anywhere on earth added roughly eighty million dollars in open interest and cleared six hundred and forty-three million dollars in notional volume across a dozen offshore venues. The instrument carries no deliverable, pays no dividend, and confers no claim on any asset a court would recognize. It is, by construction, a wager on the eventual share price of Anthropic — a private artificial intelligence laboratory that has never filed a registration statement, whose board has not confirmed a listing date, and whose most recent private-market marks sit at roughly one-tenth of the valuation the contract now implies. The market calls it a pre-IPO perpetual. I would call it something more precise: an unanchored mark. Tracing the quiet resilience beneath the market is my habit, but here the more useful discipline is to trace the quiet fragility instead. Because the number that matters is not the six hundred and forty-three million dollars. It is the two point one trillion. The vocabulary of perpetual futures has been borrowed so casually from the crypto derivatives playbook that its original engineering intent is easy to lose. A perpetual swap is a contract with no expiry date, and the absence of an expiry is precisely what makes it elegant and dangerous in equal measure. In a conventional futures contract, the price converges to the underlying by the brute force of settlement: the calendar forces the two numbers together. A perpetual has no calendar, so the mechanism that keeps the contract tethered to reality is a periodic payment called the funding rate — a small, recurring transfer between longs and shorts that punishes whichever side is crowded and rewards the side that is right. The funding rate is the heartbeat of the instrument. When it is working, it is the quiet force that keeps a synthetic price honest. When it is absent from the public record, as it is in this case, the instrument is running without a visible pulse. The second piece of machinery is the mark price. Every serious exchange distinguishes between the last traded price — which any single participant can manipulate with one aggressive order — and the mark price, a smoothed reference used to calculate unrealized profit and loss and, more critically, to trigger liquidations. On a normal crypto derivative, the mark price is assembled from an index: a basket of spot prices drawn from multiple venues, weighted and filtered to resist manipulation. It is a genuinely decentralized input, or at least a pluralistic one, and it is the reason a single exchange cannot unilaterally decide who gets liquidated. Here, that architecture inverts. Before an initial public offering, there is no spot price to index. There is no stock trading anywhere. So the mark price is derived from the contract's own recent trades — a self-referential loop in which the instrument's value is validated by the instrument's own history. In engineering terms, this is a synthetic asset with no oracle. Its correctness is determined entirely by the strength of consensus among the people already holding it. That is a profound thing to say about a market clearing hundreds of millions of dollars, and it deserves to be stated plainly: the price is whatever the participants agree it is, and nothing external can contradict them until the listing finally arrives. The third component is the convergence mechanism, and here the engineering is more thoughtful than the pricing. When SpaceX listed, the exchange that runs the dominant pre-IPO framework did not flip a switch. It rotated the mark price gradually, over a window that in that precedent stretched to roughly three hours, preserving existing positions and resting orders rather than force-closing the book. That is the right instinct. A hard switch would create a discontinuous jump, and a discontinuous jump in a leveraged market is a liquidation cascade waiting to happen. The gradual rotation is the work of people who have thought about the human beings on the other side of the margin call. But the mechanism carries an uncomfortable asterisk: the transition is not guaranteed to begin at the moment the stock starts trading. The exchange reserves the discretion to time it. That single sentence — a timing left to the operator rather than the calendar — is where the structural risk hides, because it means no trader can know the exact window in which their position will be repriced against reality. And the reality, when it comes, will be stark. Anthropic's most recent private financing placed the company in the range of roughly one hundred eighty-three billion dollars. The crypto market's implied valuation, derived from the perpetual, sits near two point one trillion. That is not a rounding difference. That is a gap of an order of magnitude — a ten-fold premium that the unanchored mark has conjured out of consensus alone. Across twelve venues, the aggregate open interest of this complex and its sibling contracts has grown from a million dollars in April to well over one hundred sixty million by late September, a six-month expansion of more than one hundred and sixty times. The exchange hosting the primary contract is also the venue that invented the framework, having first deployed it for SpaceX, and it remains the de facto anchor for pricing across the other eleven. When people describe this market as an experiment, they are being generous about the controls. I come to this with a specific bias that I should declare. I spent six months after the 2017 initial coin offering bubble auditing the smart contract infrastructure of Ripple's XRP Ledger on behalf of enterprise banking partners, and what I learned there shaped everything I have written since. I found latency issues in the consensus mechanism that quietly degraded small-scale cross-border remittances — the very use case the network advertised. Retail users in the Philippines and Nigeria were paying fees for a settlement rail that was slower than the marketing claimed, and nobody was measuring it because nobody was paid to measure it. I proposed a refined node validation protocol, and the network stabilized, but the lesson was permanent: the people who bear the cost of a fragile rail are almost never the people who design it. That is the lens I bring to every synthetic instrument now. Not whether it can clear volume. Whether the humans holding the other side can survive the moment it stops working. The self-referential mark is the first structural flaw, and it has a second-order consequence that is rarely discussed. In a normally indexed market, price discovery is distributed: thousands of independent actors, each with their own information, submit orders against a reference that no single one of them controls. Here, price discovery is closed. The only inputs are the beliefs of people who have already chosen to trade an unlisted company through an offshore derivative. There is no hedge fund with a discounted cash flow model, no sell-side analyst with a channel check, no institutional desk arbitraging against a spot market, because none of those things exist yet. The contract is a mirror reflecting the room it sits in. If the room is crowded with optimists — and the two point one trillion dollar mark suggests it is — then the mirror reflects optimism, and the reflection is mistaken for a measurement. I have seen this pattern before, though never at this scale. In the 2020 DeFi summer, I spent three weeks reverse-engineering a vulnerability in a governance interface before it was exploited, and the thing that struck me was how the protocol's internal metrics all pointed upward while the underlying security assumptions quietly rotted. The dashboard said healthy. The code said exposed. When the only feedback loop is internal, the dashboard and the reality can diverge for a very long time — right up until the moment they cannot. The turnover signature tells a similar story, and it is the kind of quiet data point that institutional readers should weigh more heavily than the headline volume. The market cleared six hundred and forty-three million dollars in notional volume against roughly eighty million in open interest. That is a ratio of about eight to one. In mature derivatives markets — Treasury futures, equity index options, even the more liquid crypto perpetuals — a monthly turnover of one to three times open interest is typical, because the people trading those instruments are hedging, financing, or taking positions they intend to hold. An eight-to-one ratio describes a market dominated by short-horizon speculation, where positions are opened and closed within hours and the open interest figure represents not conviction but the residue left behind by churn. This matters because open interest is the number that determines how much pain a convergence event can inflict. Volume is the sound of people arriving. Open interest is the sound of people staying. This market is loud with arrivals and thin on residents. And here the missing data becomes almost a story in itself. In any perpetual, the funding rate is the single most informative number an outside observer can see, because it reveals which side of the book is crowded and what that crowding costs. A persistently positive funding rate means longs are paying shorts to stay in the trade — a sign of bullish congestion. A deeply negative rate means the opposite. The funding rate is the instrument's confession. In this market, that confession has not been made publicly. Neither has the long-short distribution, nor the depth of the market-making book, nor the identity and capital base of the firms quoting prices. This is not a small omission. A market can only be evaluated for systemic risk if the structure of its positioning is legible, and here the legibility is essentially zero. Tracing the quiet resilience beneath the market requires that the market disclose where its weight is distributed. This one does not, and that absence is itself a finding. The economics of who profits from this arrangement are worth stating plainly, because they explain the incentive to keep the product alive. The exchange that runs the framework is both the referee and the house. It defines the mark price, controls the timing of the convergence, sets the liquidation rules, and collects the fees. On a typical taker fee of four to five basis points, a market clearing six hundred and forty-three million dollars a month generates fees in the low hundreds of thousands of dollars monthly for a single venue — meaningful but not transformational on its own. The larger value is strategic. A pre-IPO perpetual is a customer-acquisition instrument. It draws in traders with the highest risk appetites, and those traders tend to be heavy users of every other high-margin product the venue offers. The product does not need to be profitable in isolation; it needs to be magnetic. And a magnetic product has an institutional incentive to keep its mark elevated, because a falling mark empties the room. I do not allege misconduct. I simply observe that the entity setting the price is also the entity that benefits from the price being high, and that no third party audits the mechanism. That is not a conspiracy theory. It is a governance description. The legal architecture deserves the same unsentimental reading. American courts, following the framework established in SEC v. Howey, ask four questions to determine whether an arrangement is an investment contract and therefore a security: is there an investment of money, is there a common enterprise, is there an expectation of profit, and does that profit derive from the efforts of others. Walk through them here. Money is invested — margin posted against a leveraged position. There is a common enterprise — every holder shares exposure to the same eventual listing. There is an expectation of profit — the entire product is a spread wager. And the profit derives from the efforts of others — specifically, from Anthropic's management executing a successful listing and from the exchange's administration of the mark. All four prongs are satisfied, and the fourth is satisfied emphatically. The contracts are marketed as commodity derivatives, but the Howey arithmetic points elsewhere. That the contract settles in cash rather than delivering shares is the detail that keeps it in the gray zone. A cash-settled claim on a share price is not obviously a share; it is closer to a price-difference contract, a structure that regulators in some jurisdictions treat as gambling and in others as a financial derivative. The gray zone is not an accident. It is the design. This is where my four months working with the European Securities and Markets Authority on the Markets in Crypto-Assets regulation become relevant, because I watched the boundary get drawn in real time. The MiCA framework distinguishes between crypto-assets that function as financial instruments and those that do not, and the pre-IPO perpetual sits awkwardly across the line. It is not a transferable security, so it falls outside the traditional prospectus regime. It is not a utility token, so it falls outside the lighter crypto regime. It exists in a category the regulation did not anticipate: a synthetic exposure to a security that does not yet exist. Regulators will eventually notice, because the moment an instrument like this becomes visible to the public — and an Anthropic listing will make it visible — it invites scrutiny. The scrutiny may not arrive as enforcement. It may arrive as pressure, and pressure on an offshore venue is often sufficient to trigger a quiet delisting or a restriction on new positions. The irony is that the very event the contract is designed to anticipate — the listing — is the event most likely to end the contract's regulatory tolerance. The product's success is its own expiration. There is a deeper structural problem, and it connects to something I have argued about for years in a different context. I have been skeptical of the Layer 2 explosion on the grounds that dozens of rollups do not scale a system when they all draw from the same scarce pool of users and liquidity. They slice the pie rather than bake a larger one. The pre-IPO perpetual market exhibits the same pathology in miniature. There are twelve venues offering the same synthetic exposure, and rather than deepening a single order book, they fragment it. When the mark price is self-referential, fragmentation is not neutral. If each venue computes its mark slightly differently, then during the convergence window the venues will disagree, and the disagreements will widen precisely when coordination matters most. Arbitrageurs who would normally compress the spread may instead amplify it, because the venues are not fungible and the collateral is not portable. A convergence that should be a single, orderly repricing becomes twelve separate repricings, each with its own liquidation logic, each potentially triggering the others. The fragmentation that looks like healthy competition is, in a convergence scenario, a transmission channel for contagion. The market's diversity is its fragility. I watched exactly this dynamic during the 2022 collapse of Terra and Luna. For two months, I audited the cross-chain bridges my Central European clients depended on, and I found that three major bridge protocols lacked the liquidity reserves to honor mass withdrawals. The problem was not that any single bridge was insolvent in isolation; it was that the bridges were interdependent, and a run on one drained the collateral that backstopped the others. I negotiated quietly with operators to secure emergency liquidity pools, and we prevented some losses, but the structural lesson was that fragmented infrastructure multiplies systemic risk rather than distributing it. The pre-IPO perpetual complex is a smaller version of the same topology. It has no central clearinghouse, no shared insurance fund across venues, no coordinated liquidation logic. It has twelve independent operators each holding a piece of an unanchored mark, and when the anchor finally drops, they will all be pulling on the same rope. The custody question compounds everything, and it is where my long-standing skepticism about compliance theater becomes most pointed. These contracts require know-your-customer verification, of course, because the venues advertise their compliance. But the verification verifies almost nothing that matters. It confirms a name and a passport against a database. It does not confirm that the trader understands that they are holding a position with no external price anchor, that the mark can move against them on the operator's discretion, or that the convergence window may not coincide with the listing they are betting on. I have spent years watching honest users bear the full cost of compliance regimes while sophisticated actors route around them. The pattern here is familiar. The retail trader submits to the verification, pays the fees, accepts the leverage limits, and absorbs the liquidation when the mechanism misfires. The institutional participant, trading through a different channel or a different jurisdiction, does not. Compliance, in this architecture, is a filter that selects for the compliant rather than the protected. It sorts the honest from the connected. The honest ones are the ones holding the bag. I want to be fair to the engineering, because the mechanism does something genuinely difficult and does it reasonably well. It creates a price for an asset that has no price. Before this instrument existed, an ordinary trader in Jakarta or São Paulo had no way to express a view on a private AI laboratory; the pre-IPO market was reserved for accredited investors and sovereign wealth funds. The perpetual opened a door that had been locked. That is a real form of access, and access is not nothing. The problem is not that the door opened. The problem is that the door opened onto a room with no windows, no external clock, and a single locked exit that the host controls. Access without anchoring is not price discovery. It is collective belief with a fee schedule. And I have learned, over twenty-eight years of watching this industry, that collective belief is a wonderful thing to hold and a terrible thing to be holding when the anchor drops. The SpaceX precedent is offered as reassurance, and it deserves a closer look because it is doing a great deal of rhetorical work. Yes, the venue successfully rotated the mark for SpaceX, preserving positions and orders across a multi-hour window. Yes, the volume and open interest expanded dramatically after that listing rather than collapsing, which suggests the product line has life beyond the pre-IPO phase. But the SpaceX case is a survivorship story, and survivorship stories are dangerous precisely because they are true. SpaceX went public with a private valuation and a public pricing that were reasonably proximate; the convergence was a step, not a cliff. Anthropic's implied valuation sits an order of magnitude above its private marks. The SpaceX precedent validates the mechanism's ability to handle a gentle slope. It says nothing about the mechanism's ability to handle a precipice. Citing the gentle case as evidence that the steep case will be fine is the analytical equivalent of concluding that because one bridge held, every bridge will hold. I audited bridges in 2022 that had held for years, right up until the day they did not. Now let me turn to the part of this story that I think the consensus is getting wrong, because the contrarian case here is not the obvious one. The obvious take is that convergence will be violent, that the ten-fold gap will close in a cascade of liquidations, and that retail holders will be wiped out. I largely agree with the direction of that take. But I think it misreads the mechanism, and misreading the mechanism leads to misreading the risk. The prevailing assumption is that the convergence event is a failure mode — that the instrument is designed to track a valuation and that it will fail to do so cleanly. I want to propose a less comfortable reading. The instrument is not designed to converge. It is designed to sustain volume, and convergence is the thing that ends it. Consider the incentives. A converging mark is a dying mark; once the price is anchored to a listed stock, the exotic premium evaporates and the contract becomes an ordinary equity derivative competing on fees. The product's entire value proposition lives in the unanchored window. This means the mechanism's designers have no structural interest in a smooth convergence; they have an interest in the window staying open as long as possible. The discretion to delay the transition — to begin it 'not necessarily at the moment the stock starts trading' — is not a bug in the design. It is a feature that extends the product's life. I am not saying the delay is abused. I am saying the incentive exists, it is unmonitored, and it points in a direction that serves the operator rather than the holder. The second blind spot is the survivorship trap I mentioned, but it runs deeper than the SpaceX comparison. The industry's entire case for this product rests on the premise that the crypto market is performing genuine price discovery — that the unanchored mark is a forecast, and a good one. But price discovery requires disagreement among informed participants, and this market's participants are, by construction, a self-selected group of optimists who have chosen to buy an unlisted asset through an offshore derivative at ten times its last private mark. That is not a forecast. That is a position. A forecast is what you get when bulls and bears submit orders against a common reference. A position is what you get when only the bulls can be bothered to show up. The market cannot discover a price it is not searching for. It can only ratify a belief it already holds. The third blind spot is the legitimacy trap, and this is the one I find most instructive because it is genuinely counterintuitive. Suppose the listing prices high — say, north of a trillion dollars. The crypto market will declare victory. It will say the unanchored mark saw what Wall Street did not, that crypto priced Anthropic before the bankers did, that this is proof of the market's predictive power. And that narrative will be wrong in a specific and dangerous way. A market that guesses correctly once has not demonstrated predictive power; it has demonstrated luck, or reflexivity, or a self-fulfilling dynamic in which the crypto mark influenced the eventual pricing. Worse, the victory narrative will be used to justify more of the same — more pre-IPO contracts, more unanchored marks, more retail exposure to assets with no external reference. The one outcome that would most endanger ordinary holders is the outcome that will be most loudly celebrated. A loss teaches caution. A win teaches overconfidence. This market is structured to produce the second kind of lesson. The final blind spot concerns what happens to the humans, and it is here that my instincts as someone who believes technology must serve people rather than the reverse become most engaged. I spent part of 2026 leading a research initiative to integrate AI agents with blockchain payment rails for cross-border business-to-business settlement, and the single most important design decision we made was to keep a human in the loop. We built a micro-payment protocol that let autonomous agents settle transactions in real time, cutting friction by forty percent, and we deliberately inserted manual checkpoints not because the automation was unreliable but because the consequences of an unrecoverable automated error were unacceptable. The same principle applies here, in reverse. The pre-IPO perpetual is a fully automated pricing mechanism with no human in the loop and no external checkpoint. There is no circuit breaker that a person can pull when the mark diverges from reality. There is no auditor whose job is to ask whether the price makes sense. The mechanism is self-governing, and self-governing mechanisms with no external reference are exactly the systems that most need a human hand on the switch. The industry built the automation and forgot the switch. The world's payment rails have learned this lesson repeatedly; the world's speculative rails keep relearning it. So where does this leave a careful observer? I want to resist the temptation to end with a prediction, because predictions about convergence events are the easiest thing in this industry to make and the hardest to be right about. Instead I want to offer a framework for positioning, which is what a sideways market rewards. The core question is not whether the mark will converge; it will, because the listing is a calendar event and calendars win. The core question is whether you are positioned to survive the convergence regardless of its magnitude, and whether you understand that the entity on the other side of your position has both the power and the incentive to shape the timing. If you are holding a leveraged long in this complex, you are not making a forecast. You are trusting an operator. And trust in an unanchored system should be extended in exactly the quantity you can afford to lose, which is to say, very little. I have spent the better part of a decade building and auditing the infrastructure that lets value move across borders without breaking. I have sat with the retail users who lost funds in the 2017 bubble, and with the bridge operators who kept the lights on in 2022, and with the regulators who are now trying to write rules for instruments that did not exist when they drafted the last ones. The pattern I keep seeing is that the most dangerous instruments are the ones that feel most like access. They open a door, they let you in, they show you a price, and they never mention that the price is a mirror. The Anthropic pre-IPO perpetual is a beautiful piece of financial engineering. It is also a mirror, and the two point one trillion dollars reflected in it is not a valuation. It is a room full of people looking at themselves. Tracing the quiet resilience beneath the market is what I try to do, but sometimes the more honest work is to trace the quiet fragility instead — and to say, plainly, that the number that matters here is not the volume, not the open interest, not even the listing date. It is the gap. The gap is the whole story. The gap is where the humans live, and the gap is where the losses will be booked, and the gap is the thing that no dashboard on any of the twelve venues will show you until it is far too late to do anything about it. The world's payment rails move value because someone checks the books. This market moves belief because no one does. The question worth sitting with is not whether the mark will eventually find its anchor, but whether an industry that keeps building unanchored things will ever develop the discipline to build the checkpoint first — and whether the people holding the other side of these positions will be told the truth about the mirror before, rather than after, they are asked to look into it.