Exponential View: AI Revenue Clears the Depreciation Bar

For two years, the hundreds of billions of dollars flowing into AI infrastructure have rested on a single unresolved question: does actual commercial demand justify the spending? A new analysis from research firm Exponential View, reported by Bloomberg, offers the first systematic answer at an industry level, and for now it holds.

Global AI sales outside China reached $25 billion in the first quarter of 2026, exceeding an estimated $21 billion in quarterly depreciation tied to data-center and chip investments. That is the second consecutive quarter the industry has cleared the bar. Azeem Azhar, founder of Exponential View, told Bloomberg News that AI demand "just about clears the depreciation hurdle, and roughly speaking, it's improving over time."

The depreciation comparison matters because it is the most conservative capital-recovery test available: not profitability, not return on equity, just whether the machines being run are generating enough revenue to eventually replace themselves. Passing it once could be noise. Passing it twice is a signal. Azhar put it plainly: "For now, the economics are holding. But the margin for error is narrow."

The honest caveat is that the margin is very narrow. Depreciation charges alone still consume more than two-thirds of revenue before the industry pays for power, labor, or financing. The figures also exclude China, so the analysis reflects only part of the global AI economy. What the reporting does not give you is a clear read on how quickly the surplus is growing, or a precise definition of what counts as AI revenue versus AI-adjacent software revenue.

The near-term beneficiaries of this data point are the large cloud providers and the debt markets financing infrastructure buildout, who have been pricing risk on exactly this question. Two consecutive quarters of positive surplus, however thin, changes the actuarial assumptions for anyone deciding whether to commit to the next wave of capital spending.