AMD announced its most ambitious AI infrastructure portfolio yet at Advancing AI 2026 in San Francisco on July 22-23, led by the new AMD Helios rack-scale solution, 6th Gen EPYC "Venice" CPUs, and Instinct MI400 Series GPUs.
AMD Helios: rack-scale performance
Helios integrates 72 AMD Instinct MI455X GPUs and 18 6th Gen EPYC CPUs per rack, connected by AMD Pensando networking and accelerated by ROCm open software. AMD claims up to 30% more inference tokens per dollar than the leading competitive solution.
Leading AI labs are already adopting Helios. Anthropic and AMD announced a strategic partnership to deploy up to 2 gigawatts of AMD Instinct MI455X GPUs in Helios racks, alongside a multiyear engineering collaboration using Claude to accelerate AMD software development. OpenAI is partnering to optimize GPT-class workloads on Helios, with deployments starting in Q4 2026. Meta is co-designing for gigawatt-scale deployments.
Next-gen silicon
6th Gen EPYC processors deliver the broadest server CPU portfolio for agentic AI, with up to 256 cores and 512 threads. The Instinct MI455X GPU delivers 34× higher token throughput compared to the previous MI355X. For scientific computing, the MI430X accelerator offers up to 288 TFLOPS of FP64 performance.
AMD also introduced ROCm.ai, an AI-driven development platform enabling coding agents like Claude, Codex, and Cursor to understand AMD platforms natively.
Physical AI and roadmaps
AMD introduced the Kria AI portfolio for robotics — system-on-modules with Ryzen AI Embedded X100 processors and a robotics developer platform combining CPU, GPU, NPU, and FPGA compute. The company extended its roadmap through 2030, with MI500 Series GPUs in 2027, MI600 Series in 2028, and Zen 7 EPYC CPUs in 2028.
AMD's total addressable market in AI is projected to reach ~$2 trillion by 2030, as CEO Lisa Su declared "the next phase of AI will span frontier models, agents and physical AI."
