AMD unveils new AI chips at CES

AMD showcased new AI processors at CES, unveiling the MI455 for data‑center racks and an enterprise‑oriented MI440X tailored for on‑premise infrastructure, company executives said. CEO Lisa Su displayed the chips amid growing demand from cloud and AI firms; OpenAI’s president Greg Brockman joined her on stage, underscoring the importance of advanced processors for large‑scale model training and inference. AMD said the MI455 boosts compute, memory bandwidth and scalability over prior generations and emphasized the MI440X’s compatibility with existing enterprise hardware so customers can deploy AI workloads without wholesale infrastructure redesign.

The MI455 will be offered to hyperscalers and cloud providers and is already a component in server configurations AMD is selling to major AI customers, while the MI440X targets businesses that need AI acceleration but lack bespoke AI clusters. AMD noted expected shipments later in the year and signaled continued expansion of its AI software stack to help developers optimize performance. The company framed the chips as delivering improved performance-per-watt and a lower total cost of ownership to attract organizations looking to diversify away from a single dominant supplier.

Analysts welcomed AMD’s technical gains but said the company still faces an uphill battle against market leader Nvidia, which continues to exhaust production with strong demand. AMD’s October deal with OpenAI, viewed as an important validation of its silicon and software, bolsters its credibility but is unlikely by itself to dislodge Nvidia’s market share in the near term. Observers stressed that real‑world adoption, software optimization and supply scale will determine how far AMD can close the gap.

Benchmark claims released by AMD highlighted significant improvements for large‑scale AI tasks such as training large language models and high‑density inference, but independent testing and customer deployments will be pivotal to substantiate those results. Industry sources said the MI440X’s on‑premise focus could open new enterprise opportunities where regulatory, latency or data‑sovereignty concerns limit cloud use.

AMD framed the announcements as part of a broader push to supply both hyperscale cloud customers and enterprises with alternatives to incumbent vendors, stressing partnerships, software tooling and an expanding customer pipeline. The CES reveal underscores intensifying competition in AI hardware as firms race to deliver faster, more energy‑efficient chips to meet surging demand for generative AI and other advanced workloads.