2026 Hyperscaler CapEx: $710B, +89% YoY — AI Infrastructure Is Becoming the Moat

Meta, Microsoft, Amazon, and Google are collectively pushing 2026 CapEx toward $710B, up 89% YoY. The scale is massive.

Microsoft Cloud grew 29% YoY, Azure grew 40%, and Microsoft 365 Copilot passed 20M paid seats. AWS revenue accelerated to +28.4% YoY at a $37.6B quarterly scale, supported by AWS AI revenue above a $15B run rate. Google Cloud grew +63.4% YoY, with Cloud AI revenue up nearly 800%. Meta is seeing AI usage gains after Muse Spark, with AI glasses daily users tripling YoY from a small base.

Margins are depressed.

Microsoft still delivered operating margin of 46% despite heavy AI infrastructure spending. Google Cloud margin expanded to 32.9%, up from 17.8% last year. AWS operating margin declined to 37.7%, still a very high level for a business absorbing major AI buildout costs. Meta’s ad engine continues to fund one of the largest infrastructure expansions in the sector.

The key question: what happens if CapEx keeps growing faster than monetization?

Amazon spent $43.2B in Q1 cash CapEx. Google spent $35.7B. Microsoft reported $31.9B in CapEx. Meta spent $19.8B, including finance lease principal payments.

Full-year 2026 guidance is even more aggressive.

Microsoft expects about $190B in calendar 2026 CapEx. Google raised guidance to $180B–$190B. Meta raised guidance to $125B–$145B. Amazon is also spending heavily across AWS, generative AI, data centers, power, chips, servers, and networking.

Microsoft says Azure demand still exceeds available capacity. Google said Cloud revenue would have been higher with more capacity. AWS often spends 6 to 24 months before revenue recognition. Data centers can monetize for 30+ years, while chips and servers typically monetize over five to six years.

Custom silicon. Microsoft’s Maia 200 offers more than 30% better tokens per dollar. Cobalt is deployed in nearly half of Microsoft’s data center regions. Amazon expects Trainium to save tens of billions of dollars in annual CapEx and add several hundred basis points of margin advantage. Meta is diversifying with custom silicon developed with Broadcom, plus AMD and NVIDIA systems.

AI infrastructure is now tied directly to product velocity.

Microsoft is using CapEx to scale Copilot, agents, GitHub Copilot, and Azure. Amazon is using it to support AWS AI, Bedrock, Amazon Q, Trainium, and agentic advertising. Google is using it to scale Gemini, Search AI, Cloud AI, generative media, and Waymo. Meta is using it to improve AI feeds, ads, assistants, glasses, and model quality.

Higher memory pricing, data center costs, finance leases, and longer payback periods are pressuring free cash flow. Amazon’s free cash flow margin fell to -2.4%. Google’s free cash flow margin fell to 9.2%, down 11.8 percentage points YoY. Microsoft Cloud gross margin is expected to move toward 64% in Q4. Meta recorded a $107B increase in contractual commitments.

For hyperscalers, CAPEX is no longer just cost - it's moat, pricing power, and strategic flexibility.