Microsoft has declined to renew an arrangement to rent artificial intelligence computing resources from Oracle and has begun shutting down the capacity, as it reviews cloud commitments and places greater emphasis on its own data centers. The resources, used under a multiyear agreement announced in November 2023 to support AI-powered Bing searches, were based on older Nvidia chips that no longer met Microsoft’s technical requirements, people familiar with the matter said on November 6. The decision highlights the risk that AI hardware can remain operational but lose commercial appeal as newer models demand more computing power. Oracle said demand remains strong, including for previous-generation GPUs, and said capacity coming up for renewal in its fiscal first quarter had been renewed or resold at prices about 20% above previous long-term contracts. Microsoft also rejected a potentially multibillion-dollar Nscale expansion while keeping existing commitments intact, and IREN delayed portions of a Microsoft deployment after customer-requested networking changes. Microsoft’s capital expenditures and finance leases reached $45.2 billion last quarter, while CFO Amy Hood said some GPU-related cloud contracts had slipped to the first half of 2027. Oracle’s cloud infrastructure revenue rose 121% year over year to $7.4 billion in September earnings, and remaining performance obligations reached $664 billion. Its fiscal 2026 free cash flow was negative $23.7 billion, underscoring the financing burden of expanding AI infrastructure ahead of revenue realization. Oracle’s cloud business has also expanded through AI contracts with customers including OpenAI and Meta, reaching $18 billion in revenue in the fiscal year ending May 2026. Oracle shares fell about 3% intraday after the report before recovering in a deep V-shaped session, and have since declined by more than half from their June 1 peak of $248.15.