The AI Bubble That Isn’t

I’ve been watching the AI infrastructure buildout pretty much every quarter for the past two years now, and I keep coming back to the same question: is this a bubble, or is this something different? The bubble fear is reasonable. Valuations got extreme. Retail piled in. Anyone calling it dot-com 2.0 had a reasonable case.

In a real bubble, the weakest fundamentals can command the highest prices. The most speculative names fly highest. SoundHound AI (NASDAQ:SOUN) is the tell. In a true bubble, SOUN would be parabolic. Instead, it is down 32% year-to-date and off 66% from its January 2025 peak. That is a market discriminating.

The single most important distinction between this cycle and a genuine bubble is whose money is being spent. Specifically, capex as a percentage of free cash flow at the hyperscalers.

Alphabet (NASDAQ:GOOGL | GOOGL Price Prediction) committed roughly $180 billion in capex. Amazon (NASDAQ:AMZN) guided to approximately $200 billion. Meta (NASDAQ:META) is spending $115 to $135 billion. Microsoft (NASDAQ:MSFT) spent almost $30 billion in capex in a single quarter, up 89% year-over-year. This is operating cash flow from the most profitable businesses humanity has ever produced, not venture dollars or SPAC proceeds.

Nvidia (NASDAQ:NVDA) generated $34.9 billion in free cash flow in a single quarter. Its data center networking revenue grew 263% year-over-year as customers locked into full-stack NVLink infrastructure. Palantir posted a Rule of 40 score of 127% while U.S. commercial revenue grew 137% year-over-year. Yet Palantir is down 21% year-to-date. In dot-com, everything went up together. Here, the market is sorting winners from losers in real time.

The risks are real: concentration is genuine, capex could produce disappointing returns, and geopolitical friction is a live variable. But the AI bubble everyone feared already partially deflated in the speculative fringe. What remains is a generational infrastructure buildout funded by cash, anchored by real demand, and judged by real results.