Hewlett Packard Enterprise unveiled its first AMD‑based “Helios” rack‑scale AI architecture on December 2, 2025, announcing a purpose‑built platform that blends AMD’s EPYC “Venice” CPUs and Instinct MI455X GPUs with HPE’s Juniper‑derived networking and Broadcom’s Tomahawk 6 silicon. The solution is built around the Open Compute Project’s Ultra Accelerator Link over Ethernet (UALoE) standard, delivering the high‑bandwidth, low‑latency connectivity required for trillion‑parameter model training and high‑throughput inference.
Helios delivers 2.9 AI exaflops of FP4 performance and 2.5 exaflops of FP16, positioning it as a direct competitor to NVIDIA’s Blackwell‑based platforms. The architecture’s use of UALoE over standard Ethernet eliminates the need for proprietary interconnects, reducing deployment complexity and cost. HPE’s Juniper‑derived scale‑up switch and Broadcom’s Tomahawk 6 chip provide lossless, 200‑Gbps data paths that support the platform’s demanding bandwidth requirements.
The launch signals HPE’s strategic pivot toward an open‑standard, AI‑native infrastructure that can be integrated into its GreenLake hybrid‑cloud service. By offering a turnkey, Ethernet‑based solution, HPE can attract cloud service providers and enterprise customers seeking flexibility and cost efficiency, while also creating new recurring revenue streams through GreenLake’s pay‑per‑use model. The Helios platform’s compatibility with existing data‑center fabrics also eases migration for customers already invested in HPE’s networking stack.
Management emphasized the significance of the partnership. HPE President and CEO Antonio Neri said the collaboration “brings together the full stack of AMD compute technologies and HPE system innovation to deliver an open, rack‑scale AI platform that drives new levels of efficiency, scalability, and breakthrough performance.” AMD Chair and CEO Dr. Lisa Su added that Helios “provides a compelling alternative to NVIDIA‑centric solutions, reinforcing AMD’s position in the AI infrastructure market.” Broadcom President and CEO Hock E. Tan highlighted the role of Tomahawk 6, noting it “delivers industry‑leading ultra‑low latency and massive performance, enabling customers to build powerful AI data centers with standard Ethernet.”
The announcement follows HPE’s July 2025 acquisition of Juniper Networks, which added a robust networking portfolio and accelerated the company’s AI‑focused strategy. The Helios platform leverages the synergies from that acquisition, combining Juniper’s scale‑up switch technology with AMD’s compute stack to create a differentiated offering that challenges NVIDIA’s dominance in the AI hardware market. The move also aligns with broader industry trends toward open standards and multi‑vendor ecosystems, positioning HPE as a comprehensive AI‑ready platform for both cloud service providers and enterprise customers.
The content on BeyondSPX is for informational purposes only and should not be construed as financial or investment advice. We are not financial advisors. Consult with a qualified professional before making any investment decisions. Any actions you take based on information from this site are solely at your own risk.