Every technology boom eventually gets compared with the dot-com bubble. Artificial intelligence is no exception.
Trillions of dollars are flowing into chips, data centers and AI models, while companies including Nvidia have reached enormous valuations. Yet calling AI a bubble simply because spending is high misses the real question:
What would actually cause the boom to break?
One increasingly important answer is the gap between AI revenue and the money required to build the infrastructure behind it.
A fresh example comes from Nvidia-backed Nscale. The AI cloud company generated $140.6 million in revenue during the first half of 2026 but posted a $1.02 billion net loss, according to its IPO filing. Revenue was nevertheless up 1,252% year over year.
AI Has a Revenue Test
The biggest companies aren't spending billions simply because AI is popular. They're betting that future AI revenue will eventually justify today's enormous capital expenditures.
That's why revenue growth matters more than hype.
OpenAI, for example, reportedly projects revenue rising from around $36 billion in 2026 to $350 billion in 2030. But its internal projections also anticipate roughly $278 billion of negative free cash flow from 2026 through 2030, alongside around $856 billion of compute and infrastructure spending.
Coinpaper's examination of AI data-center debt risk shows the problem: data centers still require interest payments, electricity and refinancing even if AI demand eventually disappoints.
What Could Actually Burst an AI Bubble?
A crash probably wouldn't require people to stop using AI.
It could happen if revenue simply grows slower than investors currently expect.
That could leave data centers underused, weaken demand for new GPUs and make heavily indebted infrastructure companies harder to refinance. Higher interest rates would amplify the pressure because capital-intensive projects become less attractive as borrowing costs rise.
There is also concentration risk. Nscale's largest customer accounted for 52% of its first-half revenue, illustrating how losing or downsizing one major contract could matter enormously for some infrastructure providers.
