AMD Shows Helios Rack-Scale AI Platform at Interop Tokyo 2026
AMD is presenting Helios as an open-standard alternative to proprietary rack-scale AI systems, aiming to challenge NVIDIA's dominance in the AI infrastructure market.
Reporting from 1 sources: ASCII.jp.
AMD virtually exhibited its Helios rack-scale AI platform at Interop Tokyo 2026, positioning it as a symbol of its AI strategy. The platform integrates 72 GPUs and 18 CPUs per rack, using open standards. AMD also won a Best of Show Award for its Ryzen AI Embedded P100 Series edge processors.
AMD used Interop Tokyo 2026 to showcase its Helios rack-scale AI platform through a virtual exhibition on an 86-inch monitor, as the actual unit weighs about 3.6 tons. Helios is built on the Open Rack Wide standard from the Open Compute Project and uses open networking standards like UALink and Ultra Ethernet, contrasting with competitors' proprietary systems. The platform delivers up to 2.9 AI Exaflops (FP4 precision) per rack, with 18 compute trays each holding four AMD Instinct MI455X GPUs and one 6th-gen EPYC Venice CPU. AMD Japan field application engineer Fumiyasu Ishibashi called Helios a product that embodies AMD's AI strategy and its shift to a full-stack AI infrastructure vendor. Separately, the AMD Ryzen AI Embedded P100 Series won a Best of Show Award in the edge AI category.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.