Science & Tech · AI · 2 hrs ago
AMD unveils Helion rack designs for AI agents
AMD has introduced Helion, a rack-scale system design for running AI agents and related workloads.
It was unveiled at the Open Compute Project summit in San Jose, California.
Companies can build CPU-only racks or systems that combine CPUs and GPUs, with either air or liquid cooling.
AMD says the design can bring agent execution and, in some configurations, AI model inference together in one rack.
Its earlier Helios system was designed for large-scale AI training and inference, while Helion is aimed at enterprise uses and varied workloads.
AMD plans to share Helion specifications with the Open Compute Project community, and companies including Cisco, Dell, HPE, Foxconn, Lenovo and Supermicro are signed on.
Partner-built systems are scheduled for November 2026.
AMD unveiled Helion, a rack-scale AI reference design for running AI agents, at the Open Compute Project Summit 2026 in San Jose, California.
The design offers CPU-only or mixed CPU/GPU configurations, with options for liquid or air cooling.
Its enterprise configuration combines agent execution, GPU inference and supporting workloads in one rack.
Depending on configuration, Helion can provide up to 32,768 Zen 6 CPU cores, 16 Instinct MI350P GPUs and 102 Tbps of networking bandwidth.
AMD plans to contribute the specifications to the Open Compute Project, and partner-built systems are scheduled for November 2026.
- Who
- AMD unveiled the Helion rack design. Partners including Cisco, Dell, HPE, Foxconn, Lenovo and Supermicro are listed as on board.
- What
- A rack-scale AI reference design for agent workloads, available in CPU-only or CPU/GPU configurations.
- When
- It was revealed at the OCP Summit 2026. Partner-built configurations are scheduled for November 2026.
- Where
- The OCP Summit 2026 in San Jose, California.
- Why
- AMD says agentic AI workloads have different compute requirements and need infrastructure that can adapt to them.
This story does not have two clearly opposing sides.
To do that takes optimized compute and infrastructure that can adapt to different workloads.
With AMD Helion, we are bringing Epyc performance, open software and flexible rack designs together so our partners can build systems around customer needs and make efficient use of available power,
AMD launched Helios for frontier-scale AI training and large-scale inference.
AMD revealed the Helion rack-scale AI reference design in San Jose, California.
Partner-built Helion configurations are scheduled to be available.
- Configurations
- CPU-only hyperscale or mixed CPU/GPU enterprise
- Maximum CPU capacity
- Up to 32,768 Zen 6 CPU cores
- GPU capacity
- Up to 16 Instinct MI350P accelerators
- Maximum networking bandwidth
- Up to 102 Tbps
- Partner systems
- Scheduled for November 2026
Sources
Related news
Microsoft president calls for global AI safety rules
Nadella calls for human-controlled emergency brakes on advanced AI
Meta releases its first frontier AI models, Muse Wonder and Wonder Full
CNN asks readers to share questions about AI safety risks
Meta launches paid Muse Wonder AI model for its personal agent app
AI agent accessed Australian government systems during research task
Anthropic AI submits false murder tip to Philadelphia police
Meta Unveils Muse Wonder AI Models to Compete With Leading Systems
Meta Unveils Two Advanced AI Models
Zuckerberg says Meta has reached AI frontier and rejects slowdown calls
Meta launches Muse Wonder and Wonder Full AI models
