The $690 Billion AI Capex Sprint: Where the Hyperscaler Money Is Going

In roughly eighteen months, the aggregate annual AI infrastructure commitment from the five largest US cloud and technology companies has increased from approximately $380 billion in 2025 to a projected $660–690 billion in 2026 ^[1]. This is not an incremental increase. It is a near-doubling of capital expenditure that dwarfs anything the technology industry has attempted in a comparable period.

Futurum Group's analysis projects total hyperscaler AI capex at $690 billion for 2026 ^[1], with approximately 75% directed specifically at AI infrastructure ^[2]. The remaining quarter covers traditional cloud services, networking, and other business lines.

Key hyperscalers are expected to spend approximately 90% of their operating cash flow on capex in 2026 — up from a historical 10-year average of around 40% ^[2]. Amazon's free cash flow is projected to turn negative for the year ^[2].

Amazon: $200 Billion — The Largest Corporate Capex in History

Amazon has committed to over $200 billion in capital expenditures in 2026 ^[3] — the largest single-year capex figure ever announced by any corporation. This represents a 52% increase over the $131.8 billion spent in 2025 ^[3]. CEO Andy Jassy has framed the spending as demand-driven, pointing to AWS's cloud backlog of $244 billion — up 40% year-over-year ^[3]. AWS revenue grew 24% in Q4 2025 to $35.6 billion ^[4].

Alphabet: $175–185 Billion — Doubling Down

Alphabet has guided 2026 capex at $175 billion to $185 billion ^[5] — more than double its 2025 spend. Google Cloud's backlog reached $240 billion, up 55% year-over-year ^[5]. CEO Sundar Pichai has acknowledged that power, not capital, is the primary constraint on growth ^[6].

Microsoft: $100 Billion+ — GPUs Collecting Dust

In H1 of fiscal 2026, Microsoft spent $72.4 billion on capex — with $37.5 billion in Q2 alone, 66% higher than the prior year ^[7]. The company has disclosed an $80 billion backlog of Azure orders unfulfillable due to power constraints ^[7]. Remaining performance obligations doubled to $625 billion ^[8].

Meta: $115–135 Billion — Building Tens of Gigawatts

Meta estimates 2026 capex of $115 billion to $135 billion ^[9], up from $72.2 billion in 2025. CEO Zuckerberg plans to build "tens of gigawatts" this decade ^[10]. Key projects include the 5 GW Hyperion campus in Louisiana ($10B), a 1 GW campus in Lebanon, Indiana ($10B+), and the Prometheus site in Ohio ^[10][11].

Oracle: $50 Billion — The Stargate Gambit

Oracle's projected $50 billion in 2026 capex represents a 136% increase over 2025 ^[12]. The driving force is Stargate — the joint venture with OpenAI and SoftBank encompassing six US sites with nearly 7 GW of planned capacity and over $400 billion in total investment ^[13].

The Power Wall

Morgan Stanley Research forecasts US data center demand could reach 74 GW by 2028, with a projected shortfall of approximately 49 GW ^[14]. This gap represents the generation capacity of roughly 50 large nuclear reactors — infrastructure that takes a decade to build. The power constraint is now the binding limit on AI infrastructure growth, not chip supply, capital, or demand.