There’s a spending war happening in Big Tech, and the ammunition is measured in GPUs and concrete. Microsoft, Amazon, Alphabet, and Meta are collectively pouring money into AI data centers at a pace that would make a sovereign wealth fund blush, with combined capital expenditure projections for 2026 reaching up to $630 billion. That’s a 62% jump from the $388 billion these four companies spent in 2025.

Microsoft, though, appears to be playing the game with a slightly different hand. While it’s still spending aggressively, exceeding $120 billion in projected 2026 capex, the company is positioning itself as the more disciplined allocator in a group where free cash flow is starting to look like a casualty of war.

The numbers behind the AI arms race

Amazon is leading the charge with the biggest wallet. The company’s Q2 2026 capital expenditure hit approximately $53 billion, a 69% increase compared to the same period last year. Its full-year 2026 spending could approach $200 billion.

Alphabet isn’t far behind, with projected spending between $175 billion and $185 billion for the year. Meta sits in the $115 billion to $135 billion range. Microsoft and Meta both reported quarterly capex increases between 55% and 69%.

Revenue is growing. Microsoft’s Azure division has been a bright spot, demonstrating that demand for AI cloud services is real and accelerating. But free cash flow across these companies has taken a meaningful hit, and for Amazon and Google in particular, the gap between capital deployed and capital returned is widening.

Wall Street is now forecasting a collective $1.5 trillion in data center and AI chip spending by these hyperscalers through 2027.

Why Microsoft looks different from the pack

Microsoft’s approach to this spending spree has a few distinguishing features. The company is reportedly limiting the pace of its data center buildout rather than writing blank checks. While $120 billion-plus is hardly frugal, it represents a more measured posture compared to Amazon’s sprint toward $200 billion.

Microsoft already has Azure, the second-largest cloud platform globally, generating substantial and growing revenue from enterprise AI workloads. It doesn’t need to build from scratch. It’s upgrading and expanding existing infrastructure rather than racing to establish new capacity from zero.

What this means for crypto and broader markets

The energy angle is significant. Data centers consume enormous amounts of power, and the buildout of AI infrastructure at this scale is intensifying competition for energy resources. Bitcoin miners, who already compete for cheap electricity, are watching their power costs get influenced by tech giants willing to pay premium rates. Some mining operations have already pivoted to selling their energy contracts and hosting capacity to AI companies, a trend that accelerates as hyperscaler demand grows.

The AI infrastructure buildout is concentrating computational power in the hands of a very small number of corporations. Decentralized compute networks, the kind that various crypto projects have been trying to build, become a more compelling narrative when four companies control the majority of the world’s AI processing capacity.

For investors watching the hyperscaler spending war, the key metric isn’t total capex. It’s the ratio of AI revenue growth to capital deployed. Microsoft currently leads on that metric. The $1.5 trillion question isn’t whether this spending will happen. It’s whether the returns will justify it before shareholders lose patience.