Microsoft Increases 2026 AI Capital Expenditure by $25 Billion Amid Rising Component Costs

Microsoft has increased its 2026 AI capital expenditure by $25 billion to a total of $190 billion, driven largely by surging component costs. Memory and storage prices have more than tripled since last autumn due to AI infrastructure demand. In the last quarter, Microsoft spent roughly $32 billion on compute capacity and expects to spend an additional $158 billion by year-end, with CFO Amy Hood stating the company will remain constrained through 2026 despite massive outlays. Wall Street has expressed concern over the ROI gap: over the past four quarters, Microsoft spent $97 billion on infrastructure to generate only $37 billion in annual recurring revenue from AI services, though that ARR grew 123% year-over-year. Hood reassured investors citing strong demand signals and platform efficiencies.

While the AI business is not yet profitable, Microsoft's cloud division continues to perform strongly. Q3 earnings saw profits rise 23% year-over-year to $31.8 billion on revenues of $82.9 billion, with Azure accounting for $54.5 billion (up 29%). To manage costs, Microsoft recently pivoted GitHub Copilot to a pay-per-token model and opened its partnership with OpenAI to other models and clouds. However, personal computing revenues slipped 1% to $13.2 billion, with Windows OEM sales dropping and Xbox content/services down 5%. Hood forecast Windows OEM revenue to decline in the mid-teens next quarter, but overall Q4 revenue is expected to grow 13-15% to $86.7-$87.8 billion.