The honest answer is: probably yes, in parts — but it's messier than a clean "bubble" narrative.
The numbers are genuinely hard to square. Somewhere north of $500 billion a year is going into AI infrastructure, while actual consumer revenue from AI sits around $12 billion annually. That gap doesn't automatically mean a crash is coming, but it does mean a lot of people are betting on a future that hasn't arrived yet. 

What makes this different from a textbook bubble is that the people spending the most money aren't delusional. The hyperscalers are essentially stuck — if you pull back on AI investment and a competitor doesn't, you've potentially lost a decade of positioning. So the overspending is partly rational, even if the aggregate outcome ends up being wasteful. 

That said, the cracks are showing. MIT research found that around 95% of enterprises report zero measurable ROI from generative AI so far. And the DeepSeek moment earlier in 2025 was a preview of how fragile sentiment can be — Nvidia shed over $600 billion in market cap in a single day because one cheaper model called the hardware moat into question. 

The most likely path isn't a dramatic pop — it's a slow, uncomfortable sorting. The market is already starting to separate winners from losers rather than moving as one AI-flavored block, which is actually what a maturing cycle looks like before some names get quietly repriced into oblivion. 

For what it's worth, even OpenAI is projected to lose $17 billion in 2026 and $35 billion in 2027 by its own estimates. The revenue will come eventually — just like it did with the internet — but "eventually" is doing a lot of heavy lifting in a lot of pitch decks right now.