AMD launches Helios AI rack system to rival Nvidia's data‑center dominance

Advanced Micro Devices unveiled its Helios rack‑scale AI system at the Advancing AI 2026 conference in San Francisco. The server rack integrates 72 Instinct MI455X GPUs with EPYC CPUs, delivering up to 2.9 exaflops of peak compute, 31 TB of HBM4 memory and 1.7 PB/s bandwidth. AMD claims the platform provides roughly 15 % higher compute performance, 50 % more memory capacity and 30 % more token output per dollar compared with Nvidia’s Vera Rubin system.

Helios entered full production, with the first shipments scheduled for the third quarter of 2026. Major AI players such as Microsoft Azure, OpenAI, Meta, Oracle and Anthropic were listed as early customers. AMD also announced a strategic partnership with Anthropic that will supply up to two gigawatts of GPU capacity – the first gigawatt to be operational in the first half of 2027 – and includes a potential investment of up to $5 billion.

In parallel, AMD presented its broader AI roadmap, including the Ryzen AI Embedded X100 processors and Kria AI system‑on‑module for physical‑AI robotics, and detailed future CPU (Zen 7/8) and GPU (MI500/MI600) releases that will underpin next‑generation Helios‑500 and Helios‑600 rack solutions.