Microsoft, Meta and Amazon boost 2026 AI capex to mega-scale levels

This sell-side snapshot underlines just how dramatically hyperscalers have pivoted their balance sheets toward AI. Microsoft’s quarterly capex touching the high‑$30 billions, Google doubling its annual spend guidance, and Amazon targeting roughly $200 billion in 2026 are historically unprecedented numbers for digital infrastructure. Importantly, Guosen’s breakdown shows not just brute-force GPU buys, but also rapid growth in AI-native products: Azure AI Foundry hosting GPT‑5 and Claude 4.5 for 1,500+ shared customers, Gemini Enterprise with millions of paid seats, and AWS’s Trainium/Graviton chip revenues crossing $10 billion annually.

For the AGI race, this level of committed spend is a structural accelerant. It locks in multi‑year roads for bigger training runs, denser inference fleets and bespoke silicon—all while ROIC remains robust enough that boards are willing to keep signing the checks. It also reinforces the gravitational pull of the Big Four clouds as the primary distribution channels for Anthropic, OpenAI and the next tier of frontier labs. The risk is that such concentration makes the overall system more brittle to regulatory shocks or hardware bottlenecks, but in pure timeline terms, it pushes toward earlier arrival of increasingly general models.