Oracle Cloud Infrastructure announced an initiative to deploy 50,000 of AMD’s Instinct MI450 GPUs beginning in the second half of 2026, with plans to expand further in 2027 and beyond.

The move strengthens AMD’s competitive positioning in a market long dominated by Nvidia and underscores Oracle’s ambitions to become a central player in the AI infrastructure market.

The MI450 chips are AMD’s most advanced AI processors to date, designed to be clustered into full rack systems where 72 chips can operate as a single unit. This architecture is essential for handling the massive workloads required by today’s generative AI models.

Oracle’s deployment is a direct response to the explosive demand for compute in the AI sector, where capacity now defines competitive advantage.

Why It Matters: Cloud providers are rethinking their approach to AI infrastructure, placing greater emphasis on reducing reliance on a single chip supplier. With compute demand accelerating, companies such as Oracle are building new partnerships to ensure access to high-performance hardware from multiple sources. Its growing ties with AMD and OpenAI point to a more distributed and deliberately diversified ecosystem, moving beyond the longstanding dependence on Nvidia.

• Oracle to Deploy 50,000 AMD AI Chips in 2026 Rollout: Oracle Cloud will integrate 50,000 MI450 chips into its infrastructure starting in Q3 2026. These chips are optimized for high-scale inferencing and training tasks and are AMD’s first AI chips that can scale across large rack-based clusters. The deployment signals Oracle’s confidence in AMD’s ability to deliver high-performance alternatives to Nvidia’s GPUs.

• AMD’s MI450 Positioned as Scalable Nvidia Rival: The MI450 chips represent a technical leap for AMD. They support a rack-scale configuration in which up to 72 chips function as a single, unified system. This architecture is essential for training and deploying the most advanced AI models. AMD’s “Helios” rack design is positioned as a direct competitor to Nvidia’s fully integrated GPU and CPU systems.

• OpenAI-AMD Collaboration Anchors Deal: OpenAI played a key role in helping AMD shape the MI450 chip for AI workloads. The AI lab has already signed a separate multi-year agreement to purchase AMD chips, with plans to build a one-gigawatt compute facility beginning in 2026. If the rollout is successful, OpenAI could acquire up to 160 million shares of AMD, which would represent approximately 10% of the company.

• Nvidia Still Leads, but Pressure Builds: Nvidia continues to dominate the AI data center GPU market with more than 90% market share and maintains a close relationship with OpenAI. However, Oracle’s move toward AMD, along with AMD’s increasing momentum among major AI clients, signals a broader push for supply chain diversification. Following the announcement, AMD shares rose by more than 3%, while Nvidia and Oracle stocks declined amid overall market volatility and shifting competitive expectations.

• Industry Leaders Weigh In: Oracle Cloud VP Karan Batta emphasized AMD’s rising appeal, especially for inferencing workloads. At Oracle AI World, Chairman Larry Ellison is expected to discuss how the company is preparing to lead in the next phase of AI infrastructure. Analyst Daniel Newman noted that Oracle is positioning itself to “go all in to meet the AI moment,” but success will depend on turning raw capacity into enterprise value.

Go Deeper -> Oracle Cloud to deploy 50,000 AMD AI chips, signaling new Nvidia competition – CNBC

Oracle to offer cloud services using AMD’s upcoming AI chips – Reuters