What is Anthropic really worth? - Blain's Morning Porridge
Blain’s Morning Porridge October 27th, 2026 – What is Anthropic really worth?
“All that glitters is not gold…”
AI is changing the world. It is the biggest new thing ever, at least until the next big thing comes along. But there are many risks: businesses may adopt AI more slowly than expected, the debt bubble around NeoCloud companies could burst, and competition could grow. With all these risks, can we really justify valuing Anthropic at $1.5 trillion when the company reported an $81 billion loss in the first half of this year?
Key Takeaways
- AI is the future, and Anthropic looks like a major technology winner. But that does not mean it is worth any price. A $1.5 trillion valuation looks like the result of an overheated market full of excitement.
- The current valuation is fragile. The company is trying to get the highest possible value even though growth is slowing, businesses may adopt AI more slowly, and cheaper competitors are emerging.
- There are serious financial risks behind this valuation. If the deals in which chip makers, AI developers and NeoCloud companies help finance one another collapse, the problems could spread across the market.
- US financial markets seem to be ignoring foreign competition, especially from China, and AI models with openly available weights. These could make AI much cheaper and turn it into a basic product that is hard to sell at premium prices.
- Established technology companies competing in AI may be among the winners. They have enough money to survive a downturn or a sudden shortage of funding. Meta looks more attractive as its share price falls.
- At the right price, Anthropic could be worth buying. But is this the right price at the top of the market? Could its IPO be a signal that it is time to sell?
This morning's headlines are all about the 91 pages of risk warnings in Anthropic's IPO documents. Let's assume AI is not going to destroy us all. The timing of the deal may also help, as the US government is about to enter a two-year period of political deadlock. The Midterm elections could make it less likely that the government will introduce new rules that block the deal.
Even though I have a terrible record as a technology investor, I have a feeling Anthropic could be worth owning in a portfolio, but not at this price. The good news is that AI is here to stay. It has already become part of how our society thinks and works, and it will be part of all our futures. Just look at Microsoft, which quietly takes hundreds of pounds from me every year for Office, Teams and Outlook. And although I keep wondering which smartphone is best, every two years I pay for another overpriced iPhone.
Let's assume Anthropic will become one of the long-term winners in AI and grow into a major technology company. What could stop that from happening, and what is the company really worth?
These are fair questions. During the summer, an overly enthusiastic AI supporter told me Anthropic's IPO would value the company at $2.5 trillion to $3 trillion. He explained how good and careful the company was. Earlier this month, people in the market were talking about a $2 trillion valuation. This morning, the figure is $1.5 trillion. A 25% change in a month suggests that expectations for the future of this five-year-old AI company are quite fragile. To be fair, much of its loss in the first half of the year came from accounting charges.
If we look past all the legal language about risk, investors should focus on whether Anthropic can keep its high valuation. As I have said, I like the company. It offers an alternative to OpenAI, which is run by Sam Altman, whom I see as a real-life copy of the evil genius Dr. Evil. Altman seems so obviously like a bad guy! But I am now wondering whether CEO Dario Amodei is also moving to the dark side. He recently shifted from warning about AI risks to enthusiastically promoting deals. (Mwah, hah hah! Imagine deep, evil music playing loudly in the background.)
There are many questions about Anthropic's business risks. Can its valuation last in a market where everyone talks about a huge total addressable market, or TAM, but no one really knows how much money AI will make in the future? No one knows exactly how AI will develop, how the competition will change, or what future costs will be. And what happens if the funding bubble bursts?
These risks are all closely connected.
Bond markets often reveal the truth. A key moment could come if the debt bubble around NeoClouds bursts and sends a warning to the wider market. (Last week, I wrote about the coming collapse of CoreWeave.) I found it funny to read Elon Musk describing the short useful life of these companies, calling them “Guppy” businesses. That is because SpaceX is doing much the same thing: spending huge amounts of money to build chip factories and data centres that its own rather unimpressive Grok AI does not need, so that it can rent them out to other companies. The irony is hard to miss. Some investors may have thought they were buying a ticket to Mars on a space rocket, but instead they are buying into data centres that might never be used. The unused fibre-optic cables built during the dot-com boom around 2000 could be a useful lesson for them.
Some people compare building data centres and computing power to selling picks and shovels during an AI gold rush. They point out that the real winners in the Klondike gold rush were often the suppliers. But what if there are other ways to find AI gold without needing all the picks and shovels of this huge building boom? Just saying…
Aside from that, Musk has a point. Demand for computing power could fall, NeoCloud companies could face a rush of customers pulling out their money, and bond markets could become unstable. The complicated system in which chip makers help finance the infrastructure needed to buy their own chips could then grind to a halt. Suddenly, companies could face a cash shortage, leaving everyone wondering whether they drove too fast down a narrow country road. Private credit funds hold large amounts of debt backed by chips, or they have moved that risk into insurance funds they own.
Anthropic plans to spend around $400 billion building its own computing infrastructure and another $400 billion renting capacity from NeoCloud companies, Google, Microsoft and, believe it or not, SpaceX. Much of this spending is already committed to buying capacity in chips, servers and data centres needed to run Claude. But what happens if the funding bubble bursts, earnings fall short of expectations, or competitors offer AI at much lower prices?
I am not the only person who is rather shocked by how much AI costs. I suspect this is a major reason why Anthropic's very rapid earnings growth has slowed. Revenue growth fell from 140% in the second quarter of 2026 to 45% in the third quarter. This suggests that businesses are trying to control their AI costs and work out what they need and what they can do without. Companies are now limiting how much AI they use. A senior banker I know was told that he did not need the company's AI service because it was meant for analysts.
If businesses adopt AI more slowly, even Amodei admits that slower growth would be very bad news for Anthropic. At that point, the AI industry could quickly begin to consolidate, with weaker companies being bought or disappearing. Could we see the first acquisition worth $1 trillion?
According to the technology analysts that Wall Street follows so closely, the main competitors in AI are OpenAI and Anthropic, along with established companies Google, Microsoft and Meta. I will leave Grok out of the discussion, not because of Musk, but because I found it useless. (I tried Claude, but its price shocked me. OpenAI and I are getting along quite nicely.)
Meta's share price has fallen after the company announced a further $95 billion share sale to fund computing capacity, which was financed through deals involving chip makers. Even so, Meta, Google and Microsoft have the money and profits from their existing businesses to survive mistakes in AI. OpenAI and Anthropic do not have the same safety net.
However, much of what I read about AI seems to overlook one important fact: the competition does not come only from American large language models, or LLMs. Come on! DeepSeek, Alibaba and other companies offer models whose weights are openly available. These models are already close to being just as good, and they are definitely much cheaper.
Today's markets have huge expectations about how much AI will be worth. It could change the way businesses work and transform whole industries. It also brings risks for society, including jobs that may disappear and new jobs and opportunities that AI may create.
The problem is that overheated markets tend to price assets as if everything will go perfectly. In reality, even the most wonderful and amazing inventions eventually become common products that many companies can offer. AI is likely to follow the same pattern, just as many inventions have throughout history.
AI will probably go the same way as mobile phone services. As more companies offer similar services, they become ordinary products, and it becomes harder to charge premium prices. That could make the high prices of services like Claude impossible to maintain.
Even so, at the right price, Anthropic could be worth taking a chance on.
Out of time, so back to the day job…
Bill Blain
Author of the Morning Porridge
CEO Windshift Capital
Advisor – Spitfire Strategic Capital
Meanwhile, don't forget about my new book: The Battle for Hamble
You can read a review on the Society of Professional Economists' website here.