At CES 2026, AMD Chair and CEO Dr. Lisa Su outlined a vision of “AI for everyone,” highlighting that global compute capacity is on track to grow from 100 zettaflops today to over 10 yottaflops within five years. Su introduced AMD’s full-stack strategy—from data center GPUs to edge devices—and underscored a $150 million commitment to expand AI education in classrooms and communities. She cited partnerships with OpenAI, AstraZeneca and Blue Origin as proof points for AMD’s role in training trillion-parameter models and driving real-world AI deployments across biotech, space exploration and generative media.

AMD provided an early look at its Helios rack-scale blueprint, capable of delivering up to 3 AI exaflops per rack by combining Instinct MI455X accelerators, EPYC “Venice” CPUs and Pensando Vulcano NICs under the open ROCm software stack. The company also launched the Instinct MI440X GPU for enterprise training and inference in eight-GPU nodes, and previewed next-generation MI500 Series accelerators—built on CDNA 6 and 2 nm processes—promised to deliver up to 1 000× more AI performance versus the MI300X when they ship in 2027.

To capture surging AI PC demand, AMD introduced its Ryzen AI 400 Series processors for notebooks and desktops, delivering up to 60 TOPS of NPU performance and 1.3× faster multitasking versus competitors. The new chips feature 12 CPU cores and 24 threads, offer 1.7× faster content creation and will power over 250 AI-optimized PC platforms by Q1 2026—double last year’s count. Gamer and enthusiast segments gained the Ryzen 7 9850X3D, which at 5.6 GHz peak clock speeds yields roughly 7 percent higher gaming performance over its predecessor by leveraging higher-quality dies and a 100 MHz boost.

AMD rolled out its Ryzen AI Embedded family—P100 Series for digital cockpits and industrial automation, and X100 Series for autonomous systems. P100 variants, with 4–6 Zen 5 cores, RDNA 3.5 GPU and XDNA 2 NPU, deliver up to 2.2× CPU performance gains and 35 percent faster graphics versus prior generations, supporting up to 50 TOPS in a 15–54 W envelope. These BGA processors operate from –40 °C to 105 °C, drive four 4K120 displays, and feature 10 GbE TSN ports—all available for sampling now, with production shipments slated for Q2 2026.