Advanced Micro Devices has introduced its MI400 family of AI accelerators, setting up a fresh challenge to Nvidia in the fast-growing data center market. The company says the new chips are designed to deliver substantial gains in both AI training and inference, with performance improvements that could reach as high as 10 times compared with the current MI300X.

The announcement underscores how aggressively chipmakers are competing to supply the hardware behind generative AI and large-scale cloud computing. AMD is pitching the MI400 line as a next-generation platform for customers that need more speed and efficiency for demanding workloads, especially in data centers where power and compute density are key concerns.

While AMD’s claims point to a significant leap, the real test will come as customers benchmark the chips against rival products in real-world deployments. Nvidia remains the market leader, but AMD has been steadily expanding its presence and is now trying to close the gap with a sharper focus on performance and system-level capabilities.

The MI400 launch adds more heat to an already intense race in AI infrastructure, where faster chips can shape spending decisions for cloud providers, enterprise buyers, and hyperscale operators. If AMD’s estimates hold up, the company could strengthen its position in a market that has become central to the future of artificial intelligence.