Microsoft Ends Oracle AI Cloud Deal Over Outdated Nvidia GPUs
Microsoft is winding down its AI‑computing contract with Oracle after older Nvidia chips failed to meet its evolving requirements, part of a broader pullback from external cloud providers.
Key takeaways
- Microsoft stopped renewing its Oracle AI‑cloud contract because the older Nvidia GPUs no longer satisfied its specifications.
- The move reflects Microsoft’s broader effort to curb $80 billion‑plus spending on external cloud capacity amid cost and delivery concerns.
- Oracle says demand for its legacy AI infrastructure remains strong, pricing renewals at a 20% premium over prior terms.
Microsoft has begun winding down an agreement to rent AI‑computing capacity from Oracle, according to Bloomberg reporting on Nov. 6. The decision follows the tech giant’s refusal to renew a contract that relied on older‑generation Nvidia graphics processors, which Microsoft says no longer meet its evolving AI workload requirements. The arrangement, launched in 2023 to support the AI‑powered Bing search engine, is being phased out as Microsoft shifts to newer hardware and tighter control over external cloud spending.
Why the Oracle contract fell short
Sources familiar with the matter told Bloomberg that the rented capacity became less useful as Microsoft’s specifications changed, prompting the company to switch off Oracle‑provided servers in recent months. Oracle defended its legacy hardware, noting that most of the GPUs involved were at least four years old but still attracted customers at a 20% premium over previous agreements. The cloud provider highlighted strong demand for its OCI AI infrastructure, even for earlier‑generation GPUs, without disclosing the contract’s total value.
Microsoft’s broader external‑cloud strategy
The Oracle pullback is part of a larger reevaluation of Microsoft’s $80 billion‑plus commitment to external cloud capacity, which includes providers such as CoreWeave, Nscale and Nebius. Bloomberg reported that several Microsoft executives have expressed frustration with high costs and delivery delays from these suppliers, leading to tighter oversight of data‑center construction and a more cautious approach to new agreements. The company told Reuters its infrastructure model combines owned data centers, leased facilities and partner capacity to maintain flexibility and reliability.
Implications for the AI‑cloud market
Microsoft’s move underscores the pressure on cloud providers to keep pace with rapid AI hardware advancements. While Oracle claims continued demand for older GPUs, the premium pricing it cites suggests a niche market that may shrink as hyperscalers adopt newer chips. The shift also highlights the growing intersection between AI and cryptocurrency mining, as miners like IREN adjust AI‑cloud deployments amid volatile revenue streams.
What to watch next
Analysts will monitor whether Microsoft expands its use of newer‑generation GPUs from other partners or brings more capacity in‑house, and how Oracle’s legacy‑GPU business evolves amid competing demand for cutting‑edge AI hardware.

Discussion 0