AI-related capital expenditures are rising by approximately 40% in 2026, and the consensus forecast projects a further 30% increase in capex by major cloud providers in 2027. Hyperscalers’ strong balance sheets, extensive project pipelines, and competitive dynamics are keeping spending levels high even as growth rates decelerate, according to a recent Wolfe Research note.
Stay informed with the latest news, stock impact analysis, and Wall Street estimates — get 55% off.
Quarterly growth in AI capex exceeded 80% in early 2022, slowed sharply throughout 2023, and has since stabilized at around 40%, the report states.
Wolfe’s base-case scenario projects a further slowdown to 10–15% annual growth by 2028—a trend analysts described as a “soft landing for the capex cycle itself.”
Internal AI-related capital expenditures—encompassing hardware, software, R&D, data centers, and CHIPS Act-related facilities—have surged from near-zero levels in 2021 to approximately 2% of GDP as of the second quarter, according to the report.
Wolfe characterized this as one of the fastest investment booms in recent history in terms of the capex-to-GDP ratio, outpacing the trajectories of the late-1990s dot-com boom and the mid-2000s housing market cycle.
Direct contributions from AI investments have accounted for 10–20% of nominal GDP growth in recent quarters, excluding the negative accounting impact of imports, Wolfe noted. According to FracTracker, more than 700 data centers are currently under development, representing a 16% increase in the existing fleet. A typical hyperscale data center takes three to six years to reach full operational status.
In host counties, employment rises by approximately 3.5% and wages by 5% after a facility opens; however, these benefits are concentrated primarily in the construction phase, Wolfe notes, citing academic research. Operational data centers typically provide only 50 to 400 permanent jobs—compared to up to 10,000 during construction.
Hyperscalers’ capital expenditures for 2026 are high relative to operating cash flow. Estimates cited by The Economist place AI-related service revenue at around $200 billion, whereas projected capital expenditures more than triple that figure, Wolfe points out, adding that the gap is “highly likely to narrow” in the coming quarters.
Wolfe identified a re-rating of hyperscaler stocks—which account for the largest share of capital spending and dominate the S&P 500’s market capitalization—as the scenario most likely to turn an orderly slowdown into a chaotic one.
Such a re-rating would undermine the wealth effect and accelerate cuts to capital spending, creating what Wolfe termed “cyclicality risk.”
The conditions for this negative scenario include disappointing AI revenue growth combined with multiple compression amid a flight from risk, though Wolfe emphasized that this is not his base-case scenario.
“The risk lies in the market, not the technology itself,” the broker stated.
About the Author
