Microsoft Reportedly Winds Down Oracle AI Computing Deal

Microsoft is winding down a deal to rent artificial intelligence computing capacity from Oracle, according to a Bloomberg report published on 6 November. The reported change is a reminder that the AI infrastructure boom is not simply a race to secure more data centre space. The age and usefulness of the chips inside that space matter too.

Citing people familiar with the matter, Bloomberg said Microsoft had declined to renew the arrangement and had been turning off the rented capacity over recent months. The hardware reportedly uses an older generation of Nvidia processors that no longer fits Microsoft’s changing needs. Neither company has publicly confirmed the contract’s end or disclosed its value.

Oracle told Reuters that demand for its AI infrastructure remains strong, including for earlier-generation GPUs. Microsoft did not immediately respond to Reuters’ request for comment. The distinction matters. A decision by one customer to stop renting a particular pool of hardware is not evidence that Oracle has lost all Microsoft business, nor that demand for older chips has vanished.

The companies announced a multiyear arrangement in 2023 under which Microsoft would use Oracle Cloud Infrastructure to support its AI-powered Bing service. The financial terms were not made public. That deal belonged to an earlier stage of the generative AI surge, when companies were scrambling for available accelerators wherever they could find them.

Computing contracts often run for years, while AI workloads and chip generations change quickly. A cluster that made economic sense for one model or product may be less attractive when a company can use newer hardware, a different network design or its own infrastructure more efficiently. The reported Oracle wind-down illustrates that buying capacity and getting lasting value from it are separate questions.

Microsoft continues to invest heavily in AI and cloud infrastructure. Reuters noted that IREN had delayed some infrastructure deliveries after Microsoft requested changes to network design. That is a separate project, but it points to the same pressure on suppliers to meet increasingly specific technical requirements.

For Oracle, the immediate question is whether it can replace the capacity Microsoft is reportedly releasing with other customers. For Microsoft, it is how to balance owned facilities, leased data centres and cloud partners without being locked into computing that ages faster than its contracts. Neither answer can be read from the Bloomberg report alone.

TechBooky has previously covered the gap between Microsoft’s AI build-out plans and usable computing power, as well as a separate Texas data centre arrangement involving Microsoft and Oracle. These are distinct deals. The latest report is best understood as a change in one rental relationship, not a wholesale retreat from AI infrastructure.

The larger lesson is that AI spending has entered a more selective phase. Investors will watch not only how much computing capacity technology companies sign up for, but how much of it they keep using as hardware and model requirements evolve.