The AI Bubble Debate Just Went Macro — And We Build on Top of It
Every product we build runs on top of a handful of AI labs and one chipmaker. So when the conversation about whether that stack is overvalued moves from tech blogs to central banks, it’s not background noise for us — it’s a question about the ground we’re standing on. This week it made that move.
The numbers behind the optimism
In May, Anthropic closed a $65 billion Series H at a $965 billion post-money valuation, passing OpenAI (valued at $852 billion after its March raise) as the most valuable private AI company in the world. Anthropic’s run-rate revenue had crossed $47 billion, and the company was projecting a roughly 130% revenue surge to reach its first operating profit, according to reporting from TechCrunch. About 80% of that revenue comes from business customers, and Anthropic’s reported Q1 2026 growth rate was roughly 80x.
OpenAI’s numbers tell a different story. It’s still the bigger consumer name — 900+ million weekly ChatGPT users — but per BigGo Finance’s reporting, the company posted a $21 billion operating loss on $13 billion of revenue in 2025, or roughly $1.22 lost for every dollar earned. Both companies had filed confidential S-1 paperwork by June, with Forbes noting Fall 2026 as the floated IPO window for at least one of them.
Nvidia — the chip layer underneath both labs’ compute — reports its next earnings Wednesday, August 26, guiding to roughly $91 billion in revenue at a 74.9% gross margin. Its stock has reportedly fallen after each of the last four earnings prints despite beating expectations each time — a sign the bar for “good enough” has moved somewhere past what beating consensus can satisfy.
Where the skepticism is coming from now
None of that, by itself, is new — AI valuations have been climbing and the “is this a bubble” question has been asked all year. What’s different is who’s asking it. Bloomberg reported that uncertainty about the durability of the AI investment boom has become a recurring topic inside Federal Reserve policy meeting debates — not a tech-sector question anymore, but one the people who set interest rates are actively weighing.
Ray Dalio put a version of this plainly back in January, telling Fortune that AI is “in the early stages of a bubble” — a framing that doesn’t require betting against the technology itself, just against the price being paid for it right now.
A note on what we left out: the research briefing behind this post also cited a European Central Bank analysis (reported via a partisan opinion outlet) and a separate economist’s argument (reported via a fringe opinion site) making harder claims — that a correction is “likely,” and that AI capex is now propping up broader GDP figures. We’re not running those specific claims here because the only sourcing behind them in our research was low-credibility, and we’d rather cite fewer things correctly than more things loosely.
The actual tension
Set aside the sourcing question and the shape of the argument is straightforward, and it’s the one worth sitting with: revenue growth at Anthropic and Nvidia is real, audited, and larger in absolute terms than almost anything the dot-com era produced at a comparable valuation stage. That’s the strongest case that this cycle is structurally different from 2000. Set against that: a company can be a genuinely transformative technology and still be a bad price to buy into today — that’s exactly what happened to railroads, radio, and dot-com stocks, where the underlying technology mostly survived and eventually delivered real value, while investors who bought at the peak mostly didn’t get their capital back.
We don’t have a confident answer to which of those is happening right now, and we’re not going to manufacture one just to close this post out. We build products on this infrastructure, which means the outcome matters to us either way — we just don’t get to skip to the conclusion. Nvidia’s print tomorrow, and whatever gross-margin commentary comes with it, is the next real data point.