The numbers scream what the whitepaper whispers.
On Polymarket, the bettors are confident: 83% probability that Anthropic goes public by year-end 2026. The target valuation? A crisp $2 trillion. That's more than double the ~$965 billion post-money valuation from its May 2026 financing round—a round that itself was a record for private AI companies.
But here's the thing about screaming numbers: they often drown out the silence in the order book. And in this particular order book, the silence is deafening.
I've been here before. In 2017, I audited 50 ICO whitepapers in Seoul. Sixty percent had unsustainable tokenomics. The whitepapers whispered promises; the numbers screamed imminent collapse. Today, I read the same pattern in the silence around Anthropic's cost structure, its customer concentration, and the unspoken assumption that $2 trillion is a fair price for a company that has never disclosed its gross margin.
Let me break down the data—the real data, not the promo deck.
Context: The Data Behind the Hype
The core facts are from credible sources. The Wall Street Journal reports that Anthropic executives have met with investors, targeting a September or early October IPO. Reuters confirms the company's own forecast: 2028 annualized revenue of $190-200 billion, with current run-rate at $47 billion (as of May 2026). The secret IPO filing was made in June. Polymarket, a crypto-native prediction market, currently prices the October launch at 70% probability, rising to 83% by year-end.
But here's the first crack in the data: the $9.65 billion vs. $965 billion transcription error that appeared in the original reporting. Someone misread a decimal. That's not just a typo—it's a symptom of how fast the narrative is moving. When reporters can't keep up with the zeros, it means the market is pricing in a future that hasn't been fact-checked.
Core: The On-Chain Evidence Chain (or Lack Thereof)
I've spent the last decade building quantitative models for crypto and AI companies. My fundamental rule: trust the data, not the narrative. For Anthropic's $2 trillion IPO, the data is surprisingly thin.
Let's do the math.
A $2 trillion market cap on $200 billion in 2028 revenue gives a forward price-to-sales (P/S) ratio of 10x. That's in line with high-growth SaaS companies during their peak periods—Salesforce in 2018-2021 traded at 8-12x forward P/S. But those companies had proven gross margins above 70%, net revenue retention above 120%, and a clear path to profitability.
Anthropic has none of that publicly visible.
Current P/S on $47 billion run-rate and $2 trillion market cap is 42.6x. That's not a SaaS valuation; that's a unicorn-in-heat valuation. To justify 42.6x, the market is assuming that Anthropic will grow revenue at a compound annual growth rate (CAGR) of ~60% for three years—and then immediately decelerate to a more sustainable growth rate. That's a very specific, very optimistic path.
But here's the hidden variable: inference costs. Training a frontier model costs hundreds of millions. Running inference at scale for thousands of enterprise customers is a cash incinerator. If Anthropic's gross margin is 40-50% (typical for heavy AI inference), then a 10x P/S on $200 billion revenue implies a P/E ratio of 40-100x, depending on net margin. That's not just aggressive; it's speculative.
Chaos is just data waiting for a pattern. And the pattern here is clear: the market is pricing a narrative, not a balance sheet.
Contrarian: The Narrative Drag vs. The Fundamentals
Let me offer a counter-intuitive angle: correlation does not equal causation. Just because Polymarket says 83% probability doesn't mean the IPO is a sure thing. Prediction markets are liquidity-constrained and participant-skewed. The 83% reflects the optimism of crypto-native traders who have been conditioned to believe in exponential growth. It does not reflect the cautious skepticism of institutional investors who will actually buy the IPO.
Consider the investor meetings. The WSJ reports that some investors are targeting a valuation that is "more than double" the May round. But those same investors are also making the bank's job harder. If the IPO is priced at $2 trillion, the May investors get a 100% paper gain before the public even gets a chance to buy. That creates a structural conflict: the IPO has to be priced below the last private round to attract new money, but the company wants to maximize the headline number. The result is a split—a high nominal target, but a lower actual pricing.
I've seen this playbook before. In 2021, during the SPAC craze, companies like Lucid and Rivian went public at valuations that were multiples of their pre-money rounds. The early investors cashed out, and the public market was left holding the bag. The pattern is repeating: the narrative of AI inevitability is being used to justify a valuation that has no historical precedent for a company that is not yet profitable.
Trust is a variable I no longer solve for. But I do solve for incentives. And the incentives here are aligned toward a price that maximizes the bankers' fees and the early investors' exits—not necessarily the long-term value for public shareholders.
Takeaway: The Next Signal
So what do we watch for? The S-1 filing. When it drops, look for three numbers: the gross margin, the customer concentration (how much revenue comes from Amazon and Google), and the net revenue retention. If those numbers are strong, the $2 trillion narrative has legs. If they are weak, the market will correct within six months of the IPO.
Is the market buying a company, or buying a story? The answer will be written in the order book on day one. But the silence in the order book—the absence of hard data—is already telling me to be cautious.
— Root: 2022 Terra/Luna Collapse Aftermath (ESFP)