At CES 2026, AMD presented a broad set of AI-related hardware and software updates covering data center infrastructure, consumer and commercial PCs, gaming processors, and embedded systems. The announcements focused on expanding compute capacity for large-scale AI workloads while increasing on-device AI capabilities across PCs and edge devices.
Data center and large-scale AI infrastructure
AMD previewed a rack-scale system design called Helios, described as a modular platform intended for future large AI clusters. The system combines Instinct MI455X GPUs, EPYC “Venice” CPUs, and Pensando networking, tied together through the ROCm software stack. AMD said a single rack is designed to deliver up to 3 exaflops of AI performance, though systems are not yet shipping.
Alongside Helios, AMD detailed additions to its Instinct accelerator lineup. The new MI440X GPU targets on-premises enterprise deployments in an eight-GPU configuration, positioned for training and inference workloads inside existing data center infrastructure. AMD also reiterated plans for the MI500 Series GPUs, expected in 2027, which the company says will substantially increase performance compared with the MI300X generation. These figures are based on internal projections rather than shipping products.
AI PCs and client processors
On the client side, AMD announced the Ryzen AI 400 Series and Ryzen AI PRO 400 Series mobile processors. These chips integrate “Zen 5” CPU cores with XDNA 2 NPUs capable of up to 60 TOPS, meeting Microsoft Copilot+ PC requirements. Systems using these processors are expected to ship starting in early 2026 from major PC vendors.
AMD also expanded its higher-end mobile lineup with Ryzen AI Max+ 392 and 388 processors. These chips pair integrated Radeon graphics with large unified memory pools, targeting thin laptops, compact desktops, and local AI workloads that typically rely on discrete GPUs. AMD says configurations can support models well over 100 billion parameters when equipped with 128 GB of memory, though performance will vary by workload.
For developers, AMD introduced the Ryzen AI Halo platform, a small-form-factor desktop built around Ryzen AI Max+ processors. The system is designed to run Windows or Linux and is optimized for ROCm-based AI development. Availability is planned for the second half of 2026.
Desktop gaming processors
AMD also announced the Ryzen 7 9850X3D, a desktop CPU using second-generation 3D V-Cache. The processor is positioned as a successor to the 9800X3D and is scheduled to ship in the first quarter of 2026. Performance comparisons cited by AMD are based on internal testing against competing CPUs and specific game selections.
Embedded AI for automotive and industrial use
Separately, AMD introduced its Ryzen AI Embedded P100 and X100 processor families, aimed at automotive, industrial, and autonomous systems. These chips combine Zen 5 CPU cores, RDNA 3.5 graphics, and XDNA 2 NPUs in a single package, with AI performance ranging up to 50 TOPS depending on configuration.
The P100 series targets in-vehicle infotainment and industrial control systems, with support for extended temperature ranges and long product lifecycles. Sampling has begun for some configurations, with broader availability expected later in 2026. Higher-end X100 models, designed for more demanding autonomous and robotics workloads, are scheduled to follow.
Software and ecosystem updates
AMD also announced updates to its ROCm software platform, including broader support for Ryzen AI processors on Windows and Linux, and tighter integration with tools such as ComfyUI. The company said the goal is to reduce friction for developers running AI workloads across PCs and servers, though ROCm adoption on Windows remains relatively new.
Education and public-sector efforts
During the CES keynote, AMD said it plans to commit $150 million toward AI education initiatives, including classroom programs and community access. The company also referenced its participation in U.S. government-backed AI infrastructure projects at Oak Ridge National Laboratory, where AMD-powered systems are under development.
Overall, AMD’s CES 2026 announcements emphasized scale and breadth rather than a single flagship product. Most of the hardware discussed will roll out over the next one to two years, with several performance claims based on projections rather than shipping systems.