Microsoft Commits to Doubling AI Infrastructure as Azure Capital Expenditure Reaches $190 Billion for 2026

Microsoft has committed to doubling its AI infrastructure capacity within two years, with CEO Satya Nadella confirming the company added 1 gigawatt of additional capacity in the March quarter of fiscal year 2026 and is on track to double its overall capacity measured in gigawatts. The company raised its calendar 2026 capital expenditure spending plan to $190 billion, with $25 billion covering cost increases for components including DRAM, flash, wafers, and substrates.

Microsoft's AI business achieved an annual run rate of $37 billion in Q3 FY2026, representing a 123% year-on-year increase. Azure cloud now boasts over 80 regions, more than 500 datacenters, and over 190 points of presence globally. The company is deploying its homegrown Cobalt Arm server CPUs in half of its regions, while its "Braga" Maia 200 homegrown XPUs are live in datacenters in Iowa and Arizona.

The infrastructure expansion coincides with major enterprise AI platform developments. At SAP Sapphire 2026, Microsoft and SAP announced expanded collaboration on SAP Business Data Cloud Connect for Microsoft Fabric, enabling zero-copy data sharing that provides instant access to trusted, business-ready insights for advanced analytics. Microsoft Azure is expanding global availability for SAP Business Data Cloud to 13 Azure regions by end of 2026, including new deployments in Japan and Germany.

Microsoft's "Frontier Transformation" platform positions Azure as a foundation designed for agents, intelligence platforms, and continuous learning. Microsoft IQ, a shared intelligence layer, powers enterprise AI by connecting workforce operations, business processes, and organizational knowledge. Agent-to-agent communication between SAP Joule and Microsoft 365 Copilot frameworks enables autonomous cross-platform workflow execution.

The broader enterprise AI infrastructure market is seeing parallel investments from competitors. Google Cloud unveiled its eighth-generation TPUs at Cloud Next 2026, with the TPU 8t delivering 121 exaflops of compute in a single superpod and the Virgo Network fabric connecting up to one million TPUs across multiple sites. Anthropic committed over $200 billion largely in collaboration with Google Cloud for compute infrastructure, signaling that control over AI compute resources has become a strategic imperative.

NTT DATA research released in May 2026 highlights a critical gap in enterprise AI readiness: while over 95% of respondents consider private and sovereign AI important, only 29% are prioritizing sovereign AI in a concrete near-term way. Approximately 82% of organizations view their current infrastructure as not fully ready to support on-premises AI workloads, creating significant opportunity for managed infrastructure providers.