Claude Wouldn’t Buy into Anthropic’s IPO at a $1.5-to-2.0 Trillion Valuation

Such is the increasing relevance and awareness of AI across the globe, that the news of Anthropic’s IPO launch extended well beyond the financial press. The first public reckoning for a pure play AI company.

The closest thing akin to it in terms of IPO hype has been SpaceX with its $1.77 Trillion launch, but that is very far away from comparison as a valuation guide for an AI company like Anthropic. SpaceX close on October 29, 2026 at $144.58 a share is above its IPO subscription price of $135 but below its market debut price of $150 on June 12. Spare a thought though for those unlucky investors how bought around its intraday high of $225.64 three days after going public.

Interestingly Anthropic’s own AI model Claude thinks that its owner’s IPO is very much hype and advocates investor caution. On October 28, 2026 we asked Claude the following question.

What are the prospects of investing in Anthropic’s forthcoming IPO, is it a good idea and what are the risks associated with it? Be specific about Anthropic’s investment risk, not a general risk analysis. Use current public information as of late October 2026. Do not invent numbers. Separate confirmed S-1 facts from press estimates. Lead with the conclusion. This is analysis, not personalized financial advice.

To which it gave the following summary response (having kindly acknowledged its own conflict of interest!). If anyone is interested in the full text of the detailed response, please click the link: Claude’s Full Reply.

Conflict of interest: I’m Claude, and Anthropic makes me. Weigh my view with that in mind, and check every figure against the S-1 on SEC EDGAR. This is analysis, not personalised financial advice.

Bottom Line

At the reported $1.5–2 trillion valuation, I would not buy at the IPO. I’d wait for the price range, audited Q3 results, and the first few insider lockup releases.

The business is real. Revenue grew more than tenfold, the company has started making an operating profit, and operating cash flow turned positive. The problem is price.

A big premium already built in: $1.5T would be about 55% above the $965B private round from just five months ago.
Slowing growth: quarter-on-quarter growth appears to be slowing sharply.
Huge fixed costs ahead: non-cancelable compute contracts are about to land on the income statement.
No say for buyers: public shareholders get almost no control.
Early insider selling: insiders can start selling within weeks of listing, not after the usual 180 days.

Why would change in my view:

Price near the May mark. An IPO valued at roughly $1.0–1.2T would make it much more interesting. That’s still about 15–25 times Q2 revenue on an annualised basis.
Q3 holds up once audited. The company told investors Q3 revenue was about $16.8B. If the audited figure confirms it and Q3 stays profitable on a GAAP basis, the slowdown looks like normalisation rather than a plateau.
Profit survives the new compute. Q4 2026 and Q1 2027 should still show an operating profit after the SpaceXAI capacity comes online.
Mostly new money. Most of the IPO should be new shares sold by Anthropic, with modest selling by existing holders. There should also be a concrete plan for spending the proceeds on the compute commitments.
Early lockup releases are small. The staggered releases should cover only a small share of insider stock.

So what we are seeing is an appropriately weighted answer, with the clearest objection to investing in the IPO being the potentially overinflated valuation. It is worth noting of course that the $1.5–to-2.0 trillion is based on analysts’ estimates not one based on official subscription pricing.

As an aside from our topic, for any readers interested in the quirks of AI, you may have noticed that there is a linguistic oddity in Claude’s response. Some of the answer is written with American-English spelling whereas other parts use UK-English spelling. Leaving one to wonder where Claude might really be from…

Investors contemplating joining the IPO party however, presumably care little for such quirks. They should though be worried at the peculiar overlap between the non-cancelable compute contracts and the early insider selling that Claude flags as two concerns.

If the Circular Capital Comes Undone

One of the striking aspects of Anthropic, is the involvement of the Silicon Valley giants as shareholders, suppliers, market-place sales/generators and competitors. To Claude’s credit in its longer answer it highlights this under the heading, “Anthropic Specific Risks”.

2. The compute contracts get paid whether or not demand holds. H1 losses weren’t paying for compute: operating cash flow was positive. The real question is whether profits survive the new compute coming online.

Q2’s 8% margin came before the big new contracts started.
SpaceXAI alone at about $3.1B a month (~$9.3B a quarter) is roughly 1.7 times Q2’s entire $5.6B compute-and-infrastructure cost.
The S-1 itself warns margins may compress as large blocks of compute arrive.
The Google, Amazon, Microsoft and Broadcom minimums are non-cancelable. If revenue growth stalls, those costs stay.

3. Amazon, Google and Microsoft play every role at once. They are investors, compute suppliers, the channel for 48% of revenue, and direct competitors with their own models.

They set marketplace fees and how prominently Claude is shown.
Their ownership caps can be waived.
They hold non-voting Class N stock that can convert and be sold.
Critics call the cash flowing both ways “circular deals.”

Point 3 here is pretty blunt, Amazon, Google and Microsoft because they can play every role at once is an investment risk. Clearly they have vested interests which, could lead ultimately to the devaluation of Anthropic. One or all could easily slip into being competitively misaligned with those unfranchised Class A public shareholders.

All three, through their vested interest, stand to benefit from an overinflated valuation, they are assured of their compute revenue, they are free to sell their shares and head for the exit on a high valuation and could then subsequently crush revenue, kill value, create a vacuum and be around to sell their own alternative AI models.

It is not that circular capital does not or cannot work. So long as it remains in everyone’s’ interests, it works. However, it is one of those things that works, until it suddenly doesn’t. In the case of Anthropic’s business model that implosion may never occur, but there is a real danger that it could, and were it to happen, the consequences for its share price would be devastating. The question for prospective IPO investors is are the potential rewards sufficiently high to make them want to shoulder that risk? Clearly Claude doesn’t think so. ■

Image Credit: Funkyart-images.