Oracle to Deploy 50,000 AMD MI450 Chips!

On October 14, Oracle Cloud Infrastructure (OCI) announced plans to deploy 50,000 units of AMD's next-generation MI450 AI accelerators across its global data centers starting in Q3 2026.

Karan Batta, Senior Vice President at Oracle Cloud, told CNBC that the company expects strong customer adoption of AMD chips, especially for AI inferencing workloads. He emphasized that AMD's software stack will play a "crucial role" in driving this adoption.

Following the announcement, AMD shares briefly surged nearly 4% before closing up 0.77%, while Oracle's stock fell 2.93% after an intraday drop of over 5%.

AMD's upcoming MI400 series, revealed in June 2025, will be built on a 2nm process node and feature 432GB of HBM4 memory with an impressive 19.6TB/s bandwidth. The GPU will support 300GB/s interconnect bandwidth, delivering up to 40 PFLOPS in FP4 precision and 20 PFLOPS in FP8, representing a twofold leap in AI compute performance compared to the MI355X.

AMD CEO Dr. Lisa Su highlighted that the MI400 series "brings together everything we've learned across silicon, software, and systems to deliver a fully integrated AI platform built from the ground up." The upgrade offers 2.5× higher memory bandwidth and up to 10× better inferencing performance.

Earlier this month, AMD and OpenAI jointly announced a 6-gigawatt (6GW) partnership to power OpenAI's next-generation AI infrastructure, which will utilize multiple generations of AMD Instinct GPUs. The first phase, involving 1GW of MI450 GPU deployments, is expected to begin in late 2026.

In addition, OpenAI signed a five-year cloud deal with Oracle in September, reportedly worth up to $300 billion, to support large-scale AI workloads. Oracle's new order for AMD MI450 chips may thus be linked to the broader AMD–OpenAI–Oracle collaboration.

While NVIDIA still dominates over 90% of the data center GPU market—shipping around 1.5 million AI chips in Q2 2025 compared to AMD's 100,000—these recent partnerships show a growing trend among enterprises to diversify their AI compute ecosystems and secure more balanced supply chains.