Taking the stage in Las Vegas, CEO and chair Dr Lisa Su positioned the announcements as part of a single strategy: scaling AI capability seamlessly from consumer hardware to hyperscale and sovereign deployments.

On the client side, AMD introduced the Ryzen AI 400 and Ryzen AI PRO 400 Series, built on its “Zen 5” architecture and incorporating XDNA 2 neural processing units. The processors deliver up to 60 TOPS of AI compute, clearing the performance bar for Microsoft’s Copilot+ PCs and reinforcing AMD’s push into the AI PC market for both consumers and enterprises.

The company also expanded its Ryzen AI Max+ portfolio, bringing high-performance AI compute, integrated desktop-class graphics and a unified memory architecture to ultra-thin laptops, workstations and compact systems. These chips are designed to support demanding creative workloads, local AI inference and high-end gaming in form factors traditionally associated with lower power budgets.

AMD further highlighted the Ryzen AI Halo mini-PC, a compact system aimed at developers looking to run large language models locally. Built on the Ryzen AI Max+ platform, the device is designed to make desktop-grade AI development more accessible outside the data centre.

Gaming remains a core pillar, with the launch of the Ryzen 9850X3D, which AMD described as the fastest gaming processor currently available. Based on Zen 5 CPUs and second-generation 3D V-Cache technology, the chip targets performance-hungry gamers while maintaining efficiency.

In the data centre, AMD revealed the full Instinct MI400 Series, positioning it as a flexible portfolio covering enterprise AI, high-performance computing and large-scale training. At the top end sits Helios, a reference architecture for what AMD calls yotta-scale compute, capable of delivering up to three AI exaflops in a single rack for trillion-parameter model training.

Alongside this, AMD outlined the MI440X, an eight-GPU on-premise system for enterprise training and inference, and the MI430X, aimed at sovereign AI and HPC environments. Looking further ahead, the company disclosed early details of the MI500 Series, planned for 2027, which it expects to deliver up to a 1,000-fold AI performance increase over the MI300X, based on CDNA 6 architecture, a 2nm process and HBM4E memory.

The embedded segment also received a boost, with the launch of Ryzen AI Embedded P100 and X100 Series processors. Combining Zen 5 CPUs, RDNA 3.5 graphics and XDNA 2 NPUs, the chips deliver up to 50 AI TOPS in a compact footprint, targeting automotive, industrial and autonomous systems.

Rounding out the keynote, AMD confirmed a collaboration with Generative Bionics to develop a next-generation humanoid robot platform, underlining its ambition to push AI beyond traditional compute environments.

RELATED STORIES

AMD unveils AI roadmap through 2027 with new advancements

AMD doubles cash flow as data centre demand drives record revenues