AMD gets another vote of confidence in the AI chip wars — this time from Oracle (www.businessinsider.com)

AMD announced it will deploy 50,000 of its Instinct MI450 AI GPUs inside Oracle Cloud data centers beginning in the second half of 2026. Financial terms were not disclosed. The MI450s—the same class of accelerators AMD recently agreed to supply to OpenAI—are marketed for both model training and inference workloads. Oracle executives signaled particular confidence in AMD’s inferencing performance, positioning this deal as part of Oracle’s strategy to broaden its cloud AI hardware options beyond Nvidia. The pact is another high-profile vote of confidence in AMD amid a flurry of gigawatt‑scale infrastructure deals reshaping the AI hardware market. Coming shortly after AMD’s multi‑billion arrangement with OpenAI, the Oracle agreement strengthens competition with Nvidia and contributes to a more diversified supply chain for cloud AI compute. For the AI/ML community this means more vendor choice, potential price and performance competition, and an expanding pool of cloud-hosted GPUs for large-scale training and deployment—although the 2026 start date underscores that capacity expansion will continue to roll out over the next few years.