Anthropic Goes Public – a Bubble and a Useful Technology Can Both Be Real
Published
Anthropic, the maker of Claude, listed on Nasdaq on Thursday 19 November 2026. It was the largest IPO in history, surpassing SpaceX's record from June. The shares were priced at 83 dollars, the first trade was at 111.50 dollars and the day ended at 97.17 dollars.
Amid the celebration, it is worth looking at the numbers too. My argument is that two things are true at the same time: AI financing shows signs of a bubble, and AI is still one of the most useful technologies in decades.
What do the numbers say?
- Revenue is growing fast. Anthropic's revenue in July–September was 17 billion dollars, an annualised pace of 68 billion.
- Losses are huge. According to the prospectus, the company lost 80.7 billion dollars in the first half of 2026, and it is still losing tens of billions every quarter.
- Commitments are even bigger. Anthropic has committed around 900 billion dollars, mainly to compute and infrastructure. Some agreements can be terminated, and the company can also pay future commitments in stock.
The same money moving from pocket to pocket
Critics call this pattern in the AI boom circular financing. Roughly speaking: chipmakers and cloud companies invest in AI companies, and the AI companies spend the money on those same companies' chips and cloud services.
- Amazon has invested 18 billion dollars in Anthropic. Google has invested 12.3 billion since 2023 and promised another 30 billion if Anthropic meets agreed milestones. Both sell cloud capacity to Anthropic.
- At the IPO, Nvidia, AMD, Microsoft, Broadcom, Micron, TSMC and Google committed a total of 26 billion dollars at the IPO price. With the exception of TSMC, Anthropic has announced multibillion-dollar partnerships with these companies and buys their products.
- The same pattern repeats elsewhere. Nvidia is investing 30 billion dollars in OpenAI and has offered to guarantee up to 105 billion for an OpenAI data centre in Ohio. OpenAI is one of Nvidia's biggest customers.
The problem is not that a supplier finances its customer. That happens in many industries. The problem is that circular financing can make real demand look bigger than it is. When a chipmaker's investment comes back to it as chip sales, both companies' numbers look good even if no new money has entered the system.
The defenders' answer
Not everyone sees it so darkly. Research firm SemiAnalysis estimates that Anthropic needs roughly 280 billion dollars of revenue over four years to cover its compute costs, and considers that achievable because revenue increasingly comes from customers other than its investors. Analyst Gil Luria also points out that after the IPO the company has an exceptional amount of cash.
According to Luria, the biggest risk is not financing but competition: if Anthropic loses its place at the frontier and customers move to other models, its value could collapse. In his view Amazon and Google still benefit, because the AI runs in their data centres either way.
We have seen this before
The dot-com bubble is a good comparison. The tech-heavy Nasdaq index closed at its record on 10 March 2000 and fell about 78% by October 2002. Many companies collapsed, and at worst some turned out to be frauds, such as Enron. So much fibre-optic cable was laid that four years after the crash, 85–95% of the fibre installed in the 1990s was still unused.
Yet the internet did not go anywhere. Dark fibre was gradually lit as traffic grew. Amazon and eBay survived the bubble and became giants. The bubble was real, and so was the web.
I believe the same will happen with AI. Some of today's valuations and financing arrangements will not hold, and some data centres may end up underused. But language models that write, explain, translate and code are not going away, even if some listed company does.
What does this mean for an ordinary user?
- Use the technology for its value, not the hype. If AI saves you time today, it will save you time after a bubble too.
- Don't build everything on one service. Prices, terms and even availability can change. In the summer the US administration briefly cut off access to Anthropic's newest models worldwide.
- Be realistic about "free". When companies lose tens of billions, today's prices are not necessarily permanent.
- Investment decisions are a different matter. This post is not investment advice. A useful technology does not make every AI stock a good investment, and a bubble does not make the technology useless.
Conclusion
Anthropic's listing is a milestone. It shows that investors believe in AI to the tune of hundreds of billions, but also how much money and how many promises are moving in the industry before profits are in sight. Being critical and being enthusiastic are not mutually exclusive. You can think the financing has bubble-like features and still use AI every day because it is useful.
Sources
- CNBC: Anthropic IPO raises over $81 billion in largest offering on record (18 Nov 2026)
- Neoteo: Anthropic's Nasdaq debut (offer, open and close)
- CNBC: Anthropic's IPO plans – $80 billion in first-half losses (26 Oct 2026)
- TechCrunch: Anthropic's IPO puts AI's circular financing boom to the test (19 Nov 2026)
- The Conversation / Phys.org: The circular money machine behind the AI boom (or bubble)
- Fortune: How much of the AI boom is just Nvidia's cash being recycled? (2025)
- Fortune: AI and the dot-com bubble – parallels (dark fibre)
- Kiplinger: Lessons from the tech bubble (Nasdaq −78%)
- CNBC: Export controls on Claude Fable 5 and Mythos 5 lifted (30 Jun 2026)