Microsoft is stepping back from an Oracle AI cloud deal as older Nvidia chips fall short of its needs — news with wider implications for AI infrastructure buyers and crypto miners racing into the market. What happened - Bloomberg reported on Nov. 6 that Microsoft has begun turning off capacity it rented from Oracle after choosing not to renew a 2023 agreement. Oracle had supplied servers to support Microsoft’s AI-powered Bing, but Microsoft determined the older-generation Nvidia GPUs no longer met its evolving specs. - Oracle pushed back publicly, saying demand for its AI infrastructure — including earlier-generation GPUs — remains strong. The company said GPU capacity that came up for renewal in its fiscal Q1 was either renewed or resold on long-term contracts at roughly a 20% premium to prior pricing. Most of those GPUs, Oracle added, are at least four years old. Why it matters - The move illustrates how rapidly shifting performance needs in generative AI are reshaping cloud sourcing: hardware that was acceptable a year ago can quickly become inadequate. - Bloomberg also reported Microsoft has committed more than $80 billion to renting compute capacity from third-party providers (including CoreWeave, Nscale and Nebius). Executives have grown frustrated with high costs and supplier delays, triggering tighter oversight of outside projects and a reassessment of future agreements. - Reuters says Microsoft’s infrastructure strategy remains mixed — a combination of company-owned data centers, leased facilities and third-party partnerships to keep capacity flexible and efficient. Still, Microsoft has pulled back on some expansions, including declining an option to grow a Texas deal with Nscale. Why crypto audiences should care - The review of outside AI capacity comes at the same time Bitcoin miners and other crypto-focused infrastructure players are aggressively pivoting into AI cloud builds for large tech customers. Miners see this as a diversification route, supplying compute and data-center capacity. - IREN — an example often cited in reporting — has faced delays after customer-requested network changes and pushed its target for more than $4 billion in annual recurring revenue from December into early 2027. IREN reported $200 million in AI cloud revenue in its September quarter, while Bitcoin mining revenue tumbled 58% quarter-over-quarter to $28 million, and its shares dropped nearly 17% on Nov. 5. - Financing for miner-led AI projects has varied. BlocksBridge Consulting noted different IREN financing tranches carrying interest rates around 6% and 9%, but warned those packages fund different projects and aren’t directly comparable. Broader context and partnerships - Oracle and Microsoft continue to collaborate in other ways: in June 2024 Oracle announced a pact with Microsoft and OpenAI to provide additional capacity via Oracle Cloud Infrastructure, naming OpenAI and other AI firms (Nvidia, Together AI, xAI) as OCI customers. - Oracle has not disclosed the size of Microsoft’s departing contract or which customers acquired the released capacity. Bottom line Microsoft’s pullback from older Oracle-supplied GPUs is a reminder that AI infrastructure demand is both enormous and highly performance-sensitive. That creates both opportunity and risk for third-party suppliers — including crypto miners — who are investing heavily to become reliable providers of AI compute. Read more AI-generated news on: https://app.chaingpt.org/news