CES 2026: AMD unveils ‘AI Everywhere’ strategy
AMD unveiled GPUs, Helios racks, Ryzen AI 400 PCs, ROCm 7.2, and new embedded chips to bring AI across devices and industries.
On 5 January 2026 at CES in Las Vegas, AMD set out an ‘AI everywhere, for everyone’ agenda that spans cloud infrastructure, AI PCs, graphics software, and edge systems.
Chair and CEO Dr Lisa Su and senior leaders detailed new processors, a rack-scale reference design, and developer tools that, according to the company, are intended to make on-device and hybrid AI more accessible to consumers, enterprises, and embedded customers.
Data centre: ‘Helios’ rack design and new Instinct GPU
AMD provided an early look at Helios, a rack-scale blueprint that targets what it describes as yotta-scale AI infrastructure. Built around Instinct MI455X accelerators, EPYC ‘Venice’ CPUs and Pensando ‘Vulcano’ networking, Helios is pitched to deliver up to 3 AI exaflops per rack with tight integration through the ROCm software stack.
The company also introduced the Instinct MI440X GPU for on‑premise enterprise deployments and previewed next‑generation Instinct MI500 GPUs, planned for 2027, that are claimed to significantly exceed the performance of the MI300X generation.
AMD added a community investment note, announcing a commitment of roughly ₹1,250 crore, about $150 million, to expand AI access in classrooms and communities, alongside partner showcases from organisations including OpenAI, Luma AI and AstraZeneca.
Client computing: Ryzen AI 400, Max+, Halo and a gaming CPU refresh
For PCs, AMD launched the Ryzen AI 400 Series for Copilot+ PCs and the Ryzen AI PRO 400 Series for commercial laptops. Built on Zen 5 cores and second‑generation XDNA 2 NPUs, the line‑up offers up to 60 TOPS of NPU compute, integrated Radeon 800M graphics, and up to 12 CPU cores.
Systems from major OEMs are expected to ship in the first quarter of 2026, with desktops using Ryzen AI 400 slated for the second quarter. Two Ryzen AI Max+ SKUs were unveiled for premium thin‑and‑light notebooks and small form factor desktops, with the company stating support for local inference on large models and up to 128 GB unified memory.
AMD also announced Ryzen AI Halo, a compact developer‑focused mini‑PC that brings Max+‑class performance for AI experimentation at the edge, expected to be available in the second quarter of 2026.
On the desktop CPU side, AMD introduced the Ryzen 7 9850X3D, positioned as its fastest gaming processor with higher boost clocks over the prior 9800X3D, while on the graphics software front the firm highlighted a new FSR ‘Redstone’ update that adds machine‑learning based upscaling and frame generation for recent titles.
How ROCm 7.2 will change developer workflows
AMD said its open ROCm platform now extends to Ryzen AI 400 processors, with the upcoming ROCm 7.2 release adding broader Windows and Linux support and direct integration into ComfyUI. The company is also rolling out an AI Bundle in AMD Software: Adrenalin Edition to simplify local image generation and LLM set‑ups, including PyTorch builds for Windows, which aims to reduce configuration time for creators and developers.
Edge and embedded: Ryzen AI Embedded P100 and X100
Expanding into industrial and automotive deployments, AMD introduced the Ryzen AI Embedded portfolio, combining Zen 5 CPUs, an RDNA 3.5 GPU, and an XDNA 2 NPU on a single chip. The P100 Series, launching with 4 to 6 cores, is rated for harsh environments from minus 40°C to 105°C, supports up to four 4K or two 8K displays at 120 fps, and delivers up to 50 AI TOPS depending on the model. AMD positions these parts for digital cockpits, human–machine interfaces and factory automation.
The company said P100 Series shipments for early access customers begin now with wider production in the second quarter of 2026. Higher‑core P100 variants, 8 to 12 cores, are expected to start sampling in the first quarter, while X100 Series processors with up to 16 cores are slated to sample in the first half of the year.


