AI.bomb
AI businesses can be overpriced while the technology keeps getting more useful. I think the argument about valuations often loses sight of that.
Anthropic revenue run rate at the end of 2025
Jump in monthly tokens processed, May 2024 to October 2025
Anthropic reported annualised revenue in April 2026
There isn't an AI bubble.
Probably.
In May 1999, Barron's put Jeff Bezos on its cover as a cartoon bomb under the headline Amazon.bomb. The company went on to become much larger. That doesn't mean every criticism of its valuation was wrong, or that every AI company will follow it.
The bubble argument is easy to understand: spending and valuations have grown quickly, while returns remain uncertain. I wanted to look at the evidence that people are using the technology.
I don't think I am, but bear with me.
Useful technology can still be overpriced.
A market can overvalue companies that sell something useful. A correction can hurt shareholders even while the underlying industry grows. So evidence of demand cannot, by itself, settle the valuation argument.
People bought useful internet services in 1999 too. What interests me in 2026 is how quickly demand for AI is growing.
The something is tokens.
Google reported processing 9.7 trillion tokens a month in May 2024 and more than 1.3 quadrillion by October 2025, roughly a 130-fold increase. Its third-quarter update put the latter figure at over 20 times the level a year earlier. Token counts don't tell us how useful every request was, but that is substantial usage growth.
Anthropic reported that its annualised revenue run rate passed $30 billion in April 2026, up from about $9 billion at the end of 2025. A run rate extrapolates a recent period; it isn't $30 billion already earned over a full year.
(Sorry, I'm about to say Jevons paradox.)
When 19th-century steam engines got more efficient, Britain didn't burn less coal. It burned much more, because cheaper energy made possible uses that had been uneconomic before. Cheaper tokens probably do the same thing. They don't shrink the AI budget, they expand the list of things worth doing with AI.
The unit of work has changed, too.
From a query to an agent.
An agent task can involve many model requests and tool calls, rather than one short chatbot exchange. The amount of computation depends on the model, the task and how much work it repeats.
And agents are still getting longer time horizons. A coding agent that runs for ten minutes uses more compute than one that runs for ten seconds. An agent that runs for ten hours uses more again.
The addressable market isn't AI software.
Push the same trend out five or ten years. Most knowledge work mediated by agents. Most software interfaces replaced by a model call. A lot of stuff humans never bothered to do at all due to the impossibility, reviewing every Roblox chat message for abuse, say, now done continuously and cheaply.
It's roughly cognitive labour, plus all the things we never had cognitive labour for. Could I be wrong about that? Sure. It would require the curve to break, and so far it isn't.
Valuations look stretched on a one-year view and reasonable on a ten-year view.
The cash-burn argument still matters. A company can grow sales quickly and spend even faster. The optimistic case needs future margins to justify today's investment; demand alone won't do that.
Amazon's first full-year net profit came in 2003. It began reporting AWS separately in 2015, which made that division's profitability visible. That was a disclosure milestone, not necessarily the moment AWS first became profitable.
Uber is another reminder that a company's business model can change. It sold its self-driving business to Aurora, withdrew from some markets and added subscriptions and advertising. That history doesn't tell us which AI companies will succeed, but it makes me wary of assuming their economics will stay fixed.
OpenAI in 2030 won't look much like OpenAI today.
So yes, probably drawdowns coming. The circular deals between Nvidia, OpenAI, Oracle and CoreWeave will look ugly the moment OpenAI's revenue ramp stalls for a quarter. A blip in the grand scheme of where this is all heading.
Look ten years out. Does anyone think AI won't be in everything by then, including cars and humanoid robots?
The goalposts for what counts as intelligence will keep moving, because the genie is out of the bottle and there are trillions of dollars chasing superintelligence. Nobody is giving up after this amount of progress in such a short period of time.