The numbers are now in for Q1 2026, and the AI infrastructure buildout has entered territory that would have seemed implausible three years ago. Amazon, Google, Meta, and Microsoft are collectively on track to spend close to $700 billion on AI-related capital expenditure this year — a 45% year-over-year increase from an already elevated 2025 baseline.
Combined Q1 2026 AI-related capex from the four companies hit approximately $78 billion in the quarter alone, according to analyst estimates aggregated by Futurum Research. The full-year trajectories, extrapolated from each company’s guidance and Q1 results, tell the scale of what is being built.
The Hyperscaler Scoreboard
Amazon leads the pack with a projected $200 billion in 2026 capex — up sharply from $131 billion in 2025 — as AWS races to expand AI-ready data center capacity across North America, Europe, and Southeast Asia. Google is close behind at an estimated $175–185 billion, nearly doubling its 2025 figure of $91 billion, with a $32 billion debt raise specifically earmarked for data center expansion completed in February.
Meta, historically a distant follower in physical infrastructure, has made its most aggressive capital commitment in company history: $115–135 billion in 2026, with a stated goal of $600 billion in cumulative U.S. infrastructure investment through 2028. The bulk is directed toward AI training and inference data centers. Microsoft rounds out the group at an estimated $97.7 billion for its fiscal year 2026, having reported $37.5 billion in capex in the single quarter ending December 2025.
Beyond the hyperscalers, CoreWeave — the GPU-native cloud provider that went public in March 2026 — is targeting 5 gigawatts of expansion capacity by 2030. Nscale secured a $1.4 billion GPU-backed financing facility, described as the first major compute-collateralized debt instrument in the industry. AirTrunk closed $1.2 billion in financing for a Tokyo expansion. The capital markets have industrialized around AI infrastructure as an asset class.
Energy Becomes the Constraint
For the first quarter of 2026, the binding constraint on AI infrastructure deployment was not capital — it was energy. Global Data Center Hub’s Q1 2026 analysis describes it plainly: “Energy, not capital or demand, became the limiting factor.”
Data center power demand is now driving new procurement strategies across the industry. Meta has secured nuclear power purchase agreements. Microsoft’s deal with Constellation Energy to restart Three Mile Island Unit 1 — completed in late 2024 — has been replicated in multiple forms across the sector. Amazon signed agreements for small modular reactor output in the Pacific Northwest. Behind-the-meter natural gas generation, once a workaround, is becoming standard infrastructure at the largest campuses.
Gartner’s 2026 global IT spending forecast, released in April, puts the macro context in sharp relief. Total global IT spending is projected to reach $6.31 trillion in 2026, a 13.5% increase from 2025. The data center systems segment — the physical hardware of the AI buildout — is the fastest-growing component, rising from $505.6 billion in 2025 to $788 billion in 2026, a 55.8% single-year jump.
What Gets Built at This Scale
Campuses of 1 gigawatt or more — effectively small power grids dedicated to computation — have become normalized across the U.S., Spain, India, and Japan. EdgeMode has a 4.35 GW pipeline in Spain. Japan’s GMI Cloud launched a $12 billion sovereign AI infrastructure initiative in Q1. The physical footprint of AI is expanding faster than at any point in the history of computing.
The risk profile of this spending has not gone unnoticed. If AI application revenue growth fails to match infrastructure buildout rates, the industry faces a capital deployment problem of significant scale. Several analysts have drawn parallels to the fiber-optic overbuilding of the late 1990s. The counterargument — that the hyperscalers are building to capture markets, not waiting for demand to materialize — is supported by their revenue trajectories so far.
What is clear is that 2026 is the year the AI infrastructure wager became too large to hedge. The companies building are betting that the demand catches up. The energy grid is the last thing standing in the way.
Sources: Futurum Research AI Capex 2026 analysis; Gartner 2026 Global IT Spending Forecast; Global Data Center Hub Q1 2026 Report; company Q1 2026 earnings disclosures.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.