AMD unveils new AI chips at CES 2026 and previews MI500 leap

AMD’s CES 2026 announcements signal that the race for AI compute isn’t slowing down just because Nvidia is ahead. By pushing the Instinct MI455 series for data centers, a MI440X variant for more traditional enterprise workloads, and Ryzen AI 400 and AI Max+ for PCs, AMD is trying to own both sides of the equation: the cloud-scale training and inference market and the fast-growing AI PC category. The explicit on‑stage linkage to OpenAI underscores that hyperscalers and frontier labs are hungry for a credible second source of high‑end accelerators, if only to gain pricing leverage against Nvidia. ([reuters.com](https://www.reuters.com/business/amd-unveils-new-chips-ces-event-las-vegas-2026-01-06/))

Strategically, the teaser for MI500 with a claimed 1,000x performance jump over older generations is as much about narrative as it is about silicon. AMD is telling investors and model builders that its roadmap can underpin the next wave of larger, more agentic models without everyone being locked into one GPU vendor. The cameo from Generative Bionics’ GENE.01 humanoid robot at the same event hints at where this compute will go: embodied systems that blend large models with real‑world actuation, which will be compute‑hungry in their own right. ([reuters.com](https://www.reuters.com/business/amd-unveils-new-chips-ces-event-las-vegas-2026-01-06/))

If AMD can actually ship and scale this portfolio, it will put pressure on Nvidia’s margins and give cloud providers and labs more bargaining power. That kind of competition tends to accelerate hardware innovation cycles, which, in turn, can shorten the time between ambitious model ideas and feasible training runs.