Microsoft Winds Down Oracle AI Capacity Deal

Bloomberg reported that Microsoft is winding down an arrangement to rent AI computing from Oracle.[1] People familiar with the matter said Microsoft recently turned off Oracle capacity after declining to renew the contract.[1] The deal was announced in 2023 and gave Microsoft server access for the then-new AI version of Bing.[2] Oracle separately announced a June 2024 agreement with Microsoft and OpenAI to provide additional capacity through Oracle Cloud Infrastructure.[3]

The contract’s full size was never determined, and its value was never disclosed.[2] Oracle has also not disclosed the size of Microsoft’s departing contract or the identities of customers that acquired the released capacity.[3] Microsoft declined to comment on the reported decision.[2] Those gaps limit what outsiders can conclude about the amount of computing being removed, its financial effect on Oracle, or where that capacity may go next.

A report says Microsoft, frustrated with costs and delays, plans to cut its provider list to a few key suppliers.[2] Microsoft spent $45.2 billion on capital expenditures in its September quarter, and Nvidia GB300 systems make up most of its GPU fleet.[2] Oracle said demand for its AI infrastructure remains strong, including for earlier-generation GPUs, and said most of those GPUs were at least four years old.[2] GPU fleet utilization was 97.9% that quarter.[2]

Bitcoin miners and other crypto-focused infrastructure players are pivoting toward AI cloud builds for large technology customers.[3] IREN moved its target for more than $4 billion in annual recurring revenue from December to early 2027 after customer-requested network changes delayed its plans.[3] IREN reported $200 million in AI cloud revenue in its September quarter.[3] Its Bitcoin mining revenue fell 58% quarter over quarter to $28 million, and its shares dropped nearly 17% on November 5.[3]