Anthropic is going public this week. I am not going to recite the prospectus. I have another take, and I stand by it.
Customers are already paying. By the end of October, the announced annual revenue run rate stood at $81 billion, alongside roughly $130 billion in cash. The shares are being offered at $70 to $78, implying a market capitalisation of $1.26 trillion to $1.41 trillion: around 16 to 17 times revenue, and 46% above May’s $965 billion valuation. If the $250 billion in revenue for 2028 materialises, Anthropic could become the most valuable listed company of all time. It has the strongest financial case of any frontier AI lab.
Two camps. Closed models, such as Anthropic’s Claude: you cannot download them; you pay to use them. And open models that you can run yourself: DeepSeek and Kimi in China, Mistral in Europe. On platforms where users seek the lowest price, they often account for more than 70% of volume. That is where their greatest strengths and risks lie. More and more volume is moving towards China and Europe as users favour models they can host under their own control.
Jensen Huang was the voice of those open models, and the real constraint on Anthropic. He left on 12 November, six days before the IPO was due to be priced. I am convinced that Anthropic is the biggest beneficiary of his downfall. The turmoil at NVIDIA, just before the listing, extends the closed model’s advantage. Rising computing costs eat into margins on both sides, including Anthropic’s. But without the voice championing DeepSeek, Kimi and Mistral, Claude holds its ground for longer. And this comes the day after an election won by the Democrats, the side Anthropic funded. The timing is too perfect to be a coincidence.
Anthropic is building intelligences capable of building better ones. The prospectus says it: the threshold at which its machines accelerate their own research has been reached. They are pressing ahead. Faced with a public already wary of artificial intelligence, preaching ethics and restraint while crossing that threshold (recursive AI, in other words, without human governance) in the very week of its IPO is indecent. Fear is the brand. Acceleration is the product.
On 10 November, Anthropic launched a suite of products within Claude similar to Microsoft Office. A few months earlier, it had taken aim at Figma by creating an AI copycat, even though Microsoft and Figma were partners. They say they do not train their models on users’ data, then release the very tools those users rely on every day through those partners. I do not call that competition. I call it an unfair method. You enter through the partner. You leave with the customer. It leads to only one thing: a software monopoly.
The market does not punish this method. It values it at more than a trillion dollars. You can champion ethics, cross a threshold you swore to fear, cannibalise your partners, and be paid for it. Capitalism here does not ask whether that is healthy. It asks whether the till is ringing. It is, and so is the opening bell. I fear this method will become the model.
Les Échos: Anthropic’s IPO and the future of AI (in French, subscriber access) →
BFM Bourse / Tradingsat: Anthropic’s stock-market debut (in French, free access) →
A first-party case study of code review alongside coding agents, using Pyratz Labs workspace data from 13 August to 5 November 2026.
Why we built shared, encrypted rooms around existing coding-agent sessions—and how Paseo’s daemon, structured questions and image sharing make collaboration possible.
An investment thesis for the intelligence age: back the technologies that create abundance, the scarce resources they depend on and the regulated institutions that enable their deployment, while preserving ownership, choice and human agency.