Advanced Micro Devices has secured a substantial artificial intelligence (AI) chip order from Oracle, indicating progress in the company's efforts to catch up with NVIDIA in the booming AI chip market. Boosted by this news, AMD shares rose nearly 4% during Tuesday's trading session.
According to a statement released Tuesday, Oracle Cloud Infrastructure (OCI) will deploy 50,000 units of AMD's upcoming MI450 AI chips in its data centers starting from the third quarter of 2026. The MI450 represents AMD's most advanced GPU (graphics processing unit) to date and will be integrated into the company's proprietary Helios server rack systems, combined with AMD's self-developed central processing units (CPUs), directly targeting NVIDIA's next-generation "Vera Rubin" series AI chips.
As AI computing demand experiences explosive growth, major technology companies and AI chip manufacturers are engaging in a race to build computing infrastructure. AMD is striving to become a formidable competitor to AI chip leader NVIDIA. Similar to its larger-scale competitor, AMD offers GPU-based AI accelerators, general-purpose processors, and some networking chips. The company is further enhancing its capabilities, aiming to provide complete computing system solutions for data center operators, just like NVIDIA.
According to data from research firm IDC, AMD shipped approximately 100,000 AI chips in the second quarter, while NVIDIA shipped 1.5 million AI chips during the same period. This latest order again demonstrates market recognition of AMD's technology.
Oracle and AMD have not disclosed the specific timeline for complete deployment, but both companies expect the project to expand further in 2027 and beyond. Neither company has specified what proportion of AMD's total production Oracle's order will represent.
OCI Senior Vice President Karan Batta stated: "We believe customers will be very aggressive in adopting AMD's chips, especially in the area of artificial intelligence (AI) inference."
OCI Executive Vice President Mahesh Thiagarajan commented: "We will continue to work with AMD to provide the best value, open, secure, and scalable cloud foundation to meet customer needs for the new era of AI."
Industry analysts believe this collaboration marks the first large-scale application of AMD's MI450 in public cloud scenarios, meaning more customers can lease its AI computing resources, potentially helping AMD further narrow the gap with NVIDIA in the data center market.
Previously, AMD had just signed a chip supply agreement with AI startup OpenAI, which triggered a strong rally in AMD's stock price. Under this long-term cooperation, OpenAI will purchase computing systems with a total capacity of 6 gigawatts (GW) using AMD AI chips over the coming years. While Oracle's data centers also provide some computing resources for OpenAI, the two collaborations do not overlap.
Meanwhile, OpenAI also announced Monday that it reached an agreement with chip manufacturer Broadcom to add 10 gigawatts of computing capacity to its AI computing infrastructure.
On another front, former chip giant Intel is also attempting to re-enter the AI chip market. The company is developing a new AI accelerator called "Crescent Island," aimed at directly competing with NVIDIA and AMD products. Intel stated in Monday's announcement that the "Crescent Island" chip is optimized specifically for the AI inference stage (the operational phase after model training is complete), featuring high energy efficiency and large memory capacity. The company plans to provide samples to potential customers in the second half of 2026.