Anthropic holds its investor day tomorrow, October 14, pitching institutional investors on what would be the largest IPO in history: a $2 trillion valuation on Nasdaq, eclipsing SpaceX and Saudi Aramco. The public S-1 filing is expected October 23; trading potentially before Thanksgiving. This week, independent research firm New Constructs released a report calling the offering “the most ridiculous IPO of 2026” and “an unprecedented test of investor gullibility.” If you build on Claude, you should have an opinion on this.
The Math That Makes Bulls and Bears Argue
Anthropic’s revenue growth is legitimately extraordinary. Q2 2026 revenue hit $11.5 billion — a 14x year-over-year jump from $787 million in Q2 2025. For the first time, Anthropic beat OpenAI in quarterly revenue. By July, the annualized run rate reached $65 billion, up from $9 billion at the end of 2025. That’s not hype; that’s one of the fastest revenue ramps in enterprise software history.
But the bears have numbers too. The New Constructs reverse DCF model works backward from $2 trillion and asks: what would Anthropic need to earn to justify this price? The answer is $358 billion in annual profit by 2035 — approximately twice NVIDIA’s current annual earnings, sustained every year for a decade. At 56% compounded annual growth and a 10% profit margin from now through 2035, the model spits out $2T as the fair price. Miss either target and it’s not. New Constructs pegs fair value at $150 billion — a 93% haircut from the IPO ask.
This is the same firm that called WeWork’s 2019 offering “the most ridiculous IPO of 2019.” WeWork’s $47 billion IPO was pulled, marked down to $4 billion, and eventually ended in bankruptcy. Not predicting that here. But the track record of the comparison is worth noting.
The $518 Billion Obligation Nobody Talks About Enough
Buried in the reported S-1 is a number that deserves more airtime: $518 billion in non-cancelable compute commitments over the next decade. Amazon gets $110 billion. Google gets $111 billion. Microsoft gets $31.4 billion. Broadcom’s equipment leases account for $161.2 billion. Roughly 80% of these cannot be walked back regardless of how Anthropic’s revenue performs.
The uncomfortable subtext: Amazon, Google, and Microsoft are not just vendors — they are direct competitors in AI model development. Anthropic has signed decade-long checks to the same companies competing against its own models. This conflict is disclosed in the S-1 as a risk factor, which is the corporate equivalent of noting the problem exists without solving it.
What This Means If You Build on Claude
Anthropic’s API pricing has, until now, been tuned for developer adoption — the goal was capturing market share, not maximizing margin. That calculus changes when you answer to quarterly earnings calls. The signal is already visible: Claude Fable 5.1 launched at $10 per million input tokens and $50 per million output tokens, twice the price of the previous premium tier. The credit-billing restructuring that splits interactive and programmatic usage into separate pools looks, in retrospect, like revenue segmentation infrastructure being laid before the IPO.
None of this means prices spike on day one of trading. But the incentive that has kept aggressive API pricing — prioritize adoption — is about to share the boardroom with a second incentive: margin. Developers who have locked architecture decisions around Claude’s current price-to-capability ratio should watch the public S-1 filing on October 23 closely. The risk factor disclosures in that document will tell you more about Anthropic’s public-company obligations than any press release will.
The IPO Is a Test of a Different Kind
New Constructs’ “investor gullibility” framing is provocative, but the real question isn’t whether Anthropic is worth $150 billion or $2 trillion today. The question is whether Anthropic can become the kind of company that, over the next decade, generates $358 billion in annual profit while running the world’s most capable AI models. If it can, $2T is cheap. If open-source competition from DeepSeek and others erodes pricing power faster than Anthropic can expand margin, the math doesn’t work — and neither does the valuation.
Investor day is tomorrow. The S-1 drops in eleven days. Watch both.
Sources: New Constructs report · TechCrunch ($65B run rate) · Motley Fool ($518B compute) · IPO timeline (Financefeeds)













