Every few weeks, a new headline declares that AI-driven development is a fad or that the AI bubble is about to burst. Then, this week, Anthropic went public.
On November 19, 2026, the company behind Claude listed on Nasdaq as ANT. It priced at $83, opened at $111.50 and closed its first day at $97.17, up about 17%. It raised roughly $81 billion, the largest IPO in history, and ended the day worth about $1.76 trillion. Yes, trillion, with a T.
So is it a bubble? First, what that even means. A bubble bursting means investors suddenly decide AI companies are worth far less than they paid. Funding dries up, weaker startups go bust, and the contagion spreads through the supply chain: chipmakers, data centers and power companies. AI itself wouldn't vanish, just as the internet survived the dot-com crash.
Anthropic's IPO turns that whole debate into a single stock. Its numbers show the demand is real today. What its $1.76 trillion price really bets on is narrower: that customers keep coming, in bigger numbers, for years. Here's that bet, piece by piece.
The financial bet
The growth is real. Anthropic's revenue jumped from about $41 million in 2023 to $33 billion in just the first nine months of 2026, mostly from businesses.
The big loss is mostly on paper. Amazon invested through convertible notes, which turn into shares at an earlier, lower price. As Anthropic's valuation soared, accounting rules forced a non-cash fair-value charge, most of the headline $104 billion. Anthropic didn't actually lose that money; it's just how the accounting works.
The real bet is compute. Anthropic has signed over $540 billion in compute commitments, more than ten times this year's revenue: $100 billion-plus with Amazon, gigawatts of Google's chips, even a rented supercomputer at $1.25 billion a month. Much of it is non-cancellable, so it pays whether customers show up or not. The plan is to pour the IPO money into more of it and grow into a profit, reportedly aiming for $70 billion in revenue by 2028. In plain terms, it's building the factory before it knows how many orders will come in.
The tech side: why the bet might work
Anthropic sells tokens, small chunks of text. Every chunk you send Claude, and every chunk it writes back, gets billed. So the bet comes down to two things: will usage keep growing, and will each token keep getting cheaper to make?
Usage: coding agents are the answer to the "AI-driven dev is ending" take. Claude Code passed $2.5 billion in annual revenue by February 2026, and an agent that reads, edits and tests a whole codebase burns far more tokens than a chat reply.
Cost: in 2023, Anthropic spent about $15.60 on computing for every $1 it earned. Today it's about 57 cents. Economies of scale, more efficient models and a hardware-agnostic stack (Amazon, Google, Nvidia and AMD chips) made that possible. In Q3 2026, revenue grew 1,127% while the computing bill grew only 382%.
The risk: commoditization
The biggest threat isn't a crash; it's competition. Models keep getting cheaper and better. If a rival, or a Chinese lab, offers similar quality at half the price, customers churn. Anthropic's compute costs stay fixed, so it could lose tens of billions a year.
A bubble burst would make this worse: excess capacity would flood the market and prices would fall even further. Anthropic's moat is differentiation: frontier models, Claude Code and deep enterprise relationships.
What it means for IT jobs
Anthropic earns more when companies hand more work to AI. That's the same force squeezing routine IT jobs: entry-level hiring is slowing, while demand for AI engineering skills keeps growing.
For developers: a trillion-dollar company is betting that you'll hand AI more of your coding work every year. Your edge is shifting from writing code to system design and reviewing AI-generated code.
For founders: do your due diligence. Once AI is commoditized, "we use AI" stops being an edge, and thin AI wrappers die first, whether from a funding crunch or a price war. Build a defensible moat (proprietary data, distribution or domain expertise) and avoid vendor lock-in.
Conclusion: it's a bet on customers
Think of Patrick Jane in The Mentalist: he never saw the future, he just read the tells. Anthropic's tells all point one way. It has locked in non-cancellable compute commitments, and its unit economics improve mainly through scale. Everything hinges on customers who keep coming and keep handing Claude bigger jobs.
If they do, ANT proves the AI boom is real. If they don't, it will be one of the first places a burst shows up, and with Amazon, Google and Nvidia all holding a piece, it won't be Anthropic's problem alone.
My take
As a developer, the biggest tell for me is close to home: I hand more of my work to AI tools every month, and so do the developers around me.
That's not a bubble signal. It's a customer signal.
I'm not predicting the future, just reading the tells. Right now they say the boom is real, as long as we keep showing up.


