Choose Rich Live
Tech

AMD unveils Helion open rack design for agentic AI workloads

AMD has introduced Helion, an open server rack reference design scaling up to 32,768 CPU cores to handle the compute demands of autonomous AI agents.

Rendering of the AMD Helion server rack platform against a dark graphic background.
Image: @analyticsinme

AMD has introduced Helion, a server rack reference design specifically engineered to support agentic AI workloads. Based on an Open Compute Project specification that AMD intends to contribute to the open-source hardware initiative, the platform targets autonomous workloads such as personal assistants, code execution, and simultaneous complex enterprise operations.[1][3][4][5][6]

The platform scales up to 32,768 CPU cores—identified by Tom's Hardware as Zen 6 'Venice' cores—to address rising demand for CPU compute alongside graphics processors. Helion comes in both air-cooled and liquid-cooled configurations across two key models: a CPU-only setup for hyperscalers and an all-in-one design pairing EPYC processors with Instinct MI350P GPUs for enterprise environments. Market analyst Shay Boloor reported backing from Dell, HPE, Cisco, and Super Micro, while HotHardware noted the systems can be deployed independently or combined with Helios GPU racks.[2][4][5][6][7]

Helion also integrates a built-in token router designed to dispatch requests across frontier and open-source models to optimize operating costs. Release schedules differ between reports: Quartz and social media account @StockMKTNewz reported availability in November 2026, while Tom's Hardware reported the rack design will ship next year.[2][3][4][6]

Key facts

  • AMD introduced Helion, an open server rack blueprint based on an Open Compute Project specification and tailored for agentic AI tasks.
  • The Helion rack scales up to 32,768 CPU cores, reported by Tom's Hardware as Zen 6 'Venice' cores.
  • Helion comes in liquid-cooled and air-cooled formats, featuring a CPU-only setup for hyperscalers and an all-in-one option pairing EPYC CPUs with Instinct MI350P GPUs.
  • The platform includes a built-in token router that directs queries across frontier and open-source models to maximize cost-efficiency.
  • Dell, HPE, Cisco, and Super Micro are backing the Helion architecture, according to analyst Shay Boloor.
  • HotHardware reported that Helion racks can be deployed as standalone units or paired with Helios GPU racks.
  • Reports conflict on delivery timelines, with Quartz and @StockMKTNewz citing a November 2026 launch and Tom's Hardware reporting availability next year.

Sources · 6 sources

  1. QU

    Quartz@qzPost on X ·

    AMD's open AI rack blueprint scales to 32,768 CPU cores: AMD just unveiled Helion, an open-blueprint rack designed specifically for agentic AI workloads — the kind that juggle personal assistants, code execution, and complex enterprise tasks all at… dlvr.​it/TVtXV5 #AMD #AI https://t.co/2Aa53wsqp2

    Open source
  2. SB

    Shay Boloor@StockSavvyShayPost on X ·

    $AMD just launched Helion which is an AI agent rack with up to 33K EPYC cores backed by $DELL, $HPE, $CSCO and $SMCI as agents drive more demand for CPU compute beyond GPUs. AMD is also building in a token router that sends each request to the most cost effective AI model. https://t.co/tUdVpmrwI5

    Open source
  3. QU

    Quartz@qzPost on X ·

    AMD is introducing an open rack design for agentic AI with up to 32,768 CPU cores: The Helion platform, based on an Open Compute Project spec, pairs EPYC processors with optional Instinct GPUs and is expected to be available in November dlvr.​it/TVtZ3b https://t.co/yYt4dRpRmb

    Open source
  4. EV

    Evan@StockMKTNewzPost on X ·

    AMD $AMD JUST UNVEILED HELION, A SERVER RACK BUILT FOR AI AGENTS It's an open design that AMD plans to give to the Open Compute Project, - CPU-only version with up to 32,768 cores for hyperscalers - All-in-one version with EPYC CPUs and Instinct MI350P GPUs for businesses - Comes in liquid-cooled and air-cooled versions - A built-in router sends agents' requests across frontier and open-source models - Available in November 2026

    Open source
  5. AI

    Analytics Insight@analyticsinmePost on X ·

    𝗔𝗠𝗗 𝗛𝗲𝗹𝗶𝗼𝗻 𝗔𝗜 𝗥𝗮𝗰𝗸 𝗧𝗮𝗿𝗴𝗲𝘁𝘀 𝗔𝘂𝘁𝗼𝗻𝗼𝗺𝗼𝘂𝘀 𝗔𝗜 𝗪𝗼𝗿𝗸𝗹𝗼𝗮𝗱𝘀 𝗪𝗶𝘁𝗵 𝗨𝗽 𝘁𝗼 𝟯𝟮,𝟳𝟲𝟴 𝗖𝗣𝗨 𝗖𝗼𝗿𝗲𝘀 AMD is stepping up its AI infrastructure game with Helion, a new rack-scale platform designed for agentic AI workloads. Featuring EPYC processors, Instinct GPUs and flexible configurations, the platform aims to support the next generation of autonomous AI applications. #AMD #AgenticAI #ArtificialIntelligence #AgenticAI #ArtificialIntelligence #AIInfrastructure #EPYC #InstinctGPU #AnalyticsInsight #AnalyticsInsightMagazine Read More 👇 https://t.co/iuSaxQltJj

    Open source
  6. TH

    Tom's Hardware@tomshardwarePost on X ·

    AMD debuts Helion reference design for agentic AI servers with up to 32,768 Zen 6 ‘Venice’ cores — rack design ships next year in CPU-only and all-in-one configurations https://t.co/FtmbpLLwtm

    Open source
  7. HO

    HotHardware@HotHardwarePost on X ·

    AMD's new Helion racks are squarely targeted at Agentic AI workloads, and can be deployed separately or in combination with Helios GPU racks. https://t.co/kTYFM6gErj

    Open source