Capital expenditures tied to artificial intelligence are projected to decelerate significantly over the next several years, according to Wolfe Research. The firm's baseline scenario forecasts AI-related capex growth to slow to an annualized pace of 10% to 15% by 2028, down from recent rates of around 40%.
The slowdown follows a period of exceptional expansion. Quarterly AI capex growth exceeded 80% in early 2022, then eased through 2023 before stabilizing near 40%. Consensus estimates anticipate an additional 30% increase in hyperscalers' capital outlays in 2027, following a projected 40% rise in 2026. Wolfe describes the anticipated deceleration as a "soft landing" for the AI investment cycle.
Domestic AI-related capital spending has surged from negligible levels in 2021 to roughly 2% of GDP by the second quarter of this year. Direct investment in AI has contributed between 10% and 20% of nominal GDP growth in recent quarters, excluding the drag from import accounting. Wolfe notes that the pace of AI capex expansion is among the fastest in modern history, outpacing the late-1990s dot-com boom and the mid-2000s real estate cycle.
The infrastructure build-out remains substantial. More than 700 data centers are under development, representing a 16% expansion of the existing stock, according to FracTracker. Hyperscaler facilities typically take three to six years to reach full operational status. While construction phases generate significant economic benefits—boosting local employment by about 3.5% and wages by roughly 5%—these gains are concentrated in temporary roles. Operational data centers employ only 50 to 400 permanent workers, compared with up to 10,000 during construction.
Revenue growth has not kept pace with capital deployment. Estimated AI service revenues stand at approximately $200 billion, according to The Economist, while hyperscalers' capex estimates exceed this figure by more than three times. Wolfe expects this gap to narrow in the coming quarters. The firm warns that market repricing of hyperscaler shares poses a greater risk than the underlying technology itself.
The transition from rapid expansion to measured growth underscores the maturation of the AI investment cycle, with implications for corporate earnings, labor markets, and regional economic dynamics.
