AMD Unveils “Helios” – The Next-Gen AI System Set for 2026 Deployment

At the 2025 OCP Global Summit in California, AMD unveiled a game-changing innovation: the Helios AI system. This next-generation rack-level solution is tailored for demanding AI workloads, marking the first public display of its static design. Let’s explore the impressive capabilities and future potential of Helios.

Helios AI system

Key Highlights of the Helios AI System

  • Innovative Design: Helios adopts the ORW “dual-width” rack specification pioneered by Meta. This design optimizes power, cooling, and maintenance needs to support next-gen AI systems at scale.

  • Unprecedented Performance: Powered by 72 AMD Instinct MI450 series GPU accelerators, Helios delivers an astounding 2.9 exaFLOPS of FP4 performance. With a total HBM memory capacity of 31 TB, the system is built to handle the most intense AI workloads.

  • Massive Bandwidth: Helios is engineered for extreme data throughput, offering 260 TB/s of UALoE vertical scaling bandwidth and 43 TB/s of UEC horizontal scaling bandwidth. This ensures high-speed communication across GPUs, nodes, and racks—essential for large-scale AI models.

What’s Next for Helios?

Currently available as a reference design for OEM and ODM partners, AMD plans to kick off mass deployment of Helios in 2026. This move is set to bolster AI infrastructure and meet the escalating demands of AI applications, positioning AMD as a key player in the future of AI systems.

Helios AI system

Conclusion

The launch of Helios represents a pivotal moment in the evolution of AI infrastructure. With its unparalleled performance, cutting-edge design, and forward-thinking architecture, Helios is poised to redefine what’s possible in the realm of AI. Keep an eye on AMD as they move closer to the full-scale deployment of this groundbreaking system in 2026.

Back to blog

Related Posts

  • _{region} ( _{language} )

    • _{area}

    _{region}
    _{language}