The Recession That Never Came - And the AI Money Behind It

Where the Recession Went?

Everyone was waiting for a US recession. It never showed up. Earnings were expected to slow. They didn't. The answer might have been hiding in the GDP equation all along.

The US economy was supposed to be in a recession by now. It isn't. Corporate earnings were supposed to plateau. They haven't. Part of the reason may be sitting in plain sight in the GDP equation.

The "I" - investment - has been doing quiet, heavy lifting. Hyperscaler capex tied to data centers, chips, and AI infrastructure has been large enough to move the needle on headline growth. Add a wealth effect on top: buoyant markets lift household net worth, and that shows up in consumer spending too. Two channels, one narrative - growth that looks broad-based may be more concentrated than it appears.

Here's the part that doesn't get discussed enough: the accounting doesn't move at the same speed as the cash.

How AI spending became a GDP story

Follow the money, and it runs straight through one line in the national accounts:

GDP = Consumption + Investment + Government spending + Net exports

AI infrastructure enters mainly through the "Investment" line - but not uniformly. Data centers, software, R&D, engineering, and power infrastructure all count toward domestic business investment. Imported chips, on the other hand, don't add to GDP directly; imports get subtracted out through net exports. What does show up is everything built around those chips once they land - the construction, the software layer, the energy systems, the engineering hours. The silicon may cross a border, but the buildout around it doesn't.

That distinction matters, because it's exactly where the numbers start to get interesting.

Economists at the Federal Reserve Bank of St. Louis estimated that information-processing equipment, software, R&D, and data-center construction together added 1.16 percentage points to real GDP growth in the second quarter of 2025 - roughly 30% of the quarter's total expansion.

Let that sit for a second: nearly a third of a quarter's growth traced back to categories most closely associated with AI buildout.

The caveat is real, and worth stating plainly: this wasn't a clean read on AI spending alone. Those categories capture non-AI activity too, particularly within research and development, so the true AI-specific contribution is smaller than the headline number suggests. But even discounted for noise, the estimate points to something structural: the investment cycle around AI has grown large enough to move a number as broad and lagging as national GDP. That's not a niche tech story anymore. That's a macro one.