There is little doubt that artificial intelligence is changing how American companies work and where they put their money. The harder question — and one increasingly vexing economists — is how to prove it in the data.
Money in, results unclear
The spending is not subtle. Goldman Sachs has projected that AI companies could invest more than $500 billion in 2026, the bank said, and the buildout of data centers and computing gear has become a meaningful driver of economic growth on its own. Economists at the Federal Reserve Bank of St. Louis have found that spending on information-processing equipment recently made an unusually large contribution to GDP growth, the bank reported — a sign of how much capital is flowing into the technology.
Yet that investment surge has not translated into a clear, economy-wide jump in productivity. The gap has revived what economists call the "Solow paradox," after the economist Robert Solow's 1987 quip that the computer age was visible "everywhere but in the productivity statistics." The same, many argue, is now true of AI.
Why the numbers lag
Part of the problem is measurement itself. Federal Reserve researchers note that AI's rapid evolution and general-purpose nature make it hard to tell meaningful adoption from casual experimentation — a firm merely testing a chatbot can look, in the data, much like one that has rebuilt its operations around AI, according to a Fed analysis. Standard statistics also struggle to capture quality improvements and hard-to-value software and "intangible" investment, which can leave real gains uncounted.
There is also a timing issue with historical precedent. When computers spread through offices in the 1980s and 1990s, measurable productivity gains took years to arrive, showing up only after companies reorganized how they worked. Some economists expect AI to follow a similar arc, with the payoff landing later this decade.
Believers and skeptics
Not everyone is convinced the payoff is coming soon. Even as AI shows sizable gains on specific tasks in study after study, surveys of executives have found many reporting little measurable effect on their overall productivity or headcount so far. Optimists see a delayed boom as businesses learn to use the tools; skeptics warn that the enormous spending on data centers could outrun the returns, raising the risk of a bubble.
The stakes are not academic. The uncertainty complicates the Federal Reserve's job of reading the economy, and it leaves corporate boards betting billions without a clear scorecard. For now, the honest answer to how much AI is adding to the economy is that no one can yet measure it precisely — only that the money going in is real, and the proof in the aggregate numbers has not yet arrived.
