Is AI a bubble? Or just the largest self-funded buildout in history?

It’s a fair question, especially with how large the dollar amounts are being thrown around. But the structure of today’s AI investment looks very different from prior infrastructure booms.

Start with how this buildout is being funded.

The largest AI investments today are coming directly out of operating cash flows from some of the strongest companies in the world.

In 2025 alone, the top five hyperscalers spent an estimated $415B on CapEx against roughly $1.7T in combined revenue.

Compare that to prior infrastructure booms. U.S. railroads, telecom buildouts, and the dot-com era were largely financed with debt and speculative capital.

When demand disappointed or capital markets shut, the unwind was brutal. This cycle is different: balance sheets are strong, cash flows are real, and demand for compute continues to exceed supply.

None of this means valuations won’t overshoot. They usually do. But bubbles aren’t defined by high spending alone, they’re defined by fragile financing, weak underlying economics, and demand that vanishes when capital dries up.

AI doesn’t seem to fit that mold.

This is a capital-intensive, productivity-driven transition being funded by cash-generating incumbents, not a debt-fueled bet on eyeballs.

The long-term question isn’t whether spending slows or multiples compress.

It’s how much durable productivity AI ultimately unlocks across the economy.