The AI Bubble Question Is Really an AI ROI Question
September 28, 2026 · Keith Pasley
Michael Burry thinks the AI bubble could burst sooner than he originally expected.
I’m less interested in whether his timing is right than in the reason behind the argument.
Burry points to research from Ares Management questioning how long today’s AI capital spending can continue if the revenue expected from that investment doesn’t materialize.
That’s the part enterprise leaders should pay attention to.
Billions are going into GPUs, data centers, models, cloud capacity and AI infrastructure. That investment makes sense if AI creates enough economic value to support it.
Eventually, though, the conversation has to move from capacity to outcomes.
How much revenue did AI create?
What operating costs did it reduce?
Which business processes improved?
Which decisions got better?
What did an agent accomplish that justified the cost of running it?
And which AI investments aren’t producing enough value to continue funding?
This is why I think cost per token and even model performance will become secondary measures for many enterprise AI decisions.
The important number is going to be closer to cost per successful business outcome.
That doesn’t mean the AI market is necessarily a bubble. It means the infrastructure buildout eventually has to meet the economics of the businesses paying for it.
If AI revenue or measurable productivity doesn’t develop fast enough, CFOs and boards won’t need to decide that AI failed.
They can simply decide that the next dollar of capital produces a better return somewhere else.
That would ripple through the entire AI ecosystem.
Model providers.
Cloud platforms.
GPU manufacturers.
Data centers.
Networking.
Power infrastructure.
Software vendors.
And the enterprises funding AI projects internally.
The AI infrastructure race can continue for a long time.
But ultimately someone has to demonstrate that all of that infrastructure is producing something worth paying for.
https://www.cnbc.com/2026/09/28/michael-burry-believes-the-ai-bubble-may-burst-sooner-than-he-first-believed.html
Technology doesn’t create business value. Better decisions do.