BARCELONA – Hewlett Packard Enterprise (HPE) has collaborated with Broadcom on the first commercial deployment of an AMD Helios-based AI rack-scale system.

As announced at HPE Discover Barcelona 2025, the architecture is built on purpose-designed HPE Juniper Networking hardware and software, Broadcom’s Tomahawk 6 networking chip, and AMD’s upcoming flagship GPU, the MI450 series.

Under the hood, the system connects 72 AMD Instinct MI455X GPUs in a single rack, delivering 260 terabytes per second of aggregate scale‑up bandwidth and up to 2.9 AI exaflops of FP4 performance to accelerate workloads and handle very large AI models.

It also provides 31 terabytes of fourth‑generation high‑bandwidth memory (HBM4) and 1.4 petabytes per second of memory bandwidth, enabling, HPE said, high capacity and data movement from trillion-parameter model training, high inference throughput, and extremely large AI model sizes.

Helios also follows specifications from the Open Compute Project (OCP), as well as the open Ultra Accelerator Link over Ethernet (UALoE) standard.

With the release, HPE will be one of the first companies to offer an AMD Helios AI rack-scale solution with integrated scale-up Ethernet networking.

In a press briefing prior to HPE's flagship European event, Rami Rahim, EVP, president and GM of HPE Networking, said: “This is an industry first scale up solution using standard Ethernet, so that means it's 100% open standard, avoids proprietary vendor lock in, leverages proven HPE Juniper networking technology to deliver scale – and optimize performance for AI workloads. I could not be more excited about this groundbreaking partnership, it's a true first of its kind. We're bringing Ethernet to a new layer of the AI data center network.”

The release follows a preview of the Helios architecture at the OCP Global Summit in October, with AMD initially promising deployment in 2026 with its OEM and ODM partners.

Helios is designed on Meta’s Open Rack Wide (ORW) specification and enables operators to integrate AMD’s family of hardware, including its Instinct GPUs and Pensando GPUs, with their own differentiated solutions.

Besides OCP standards, the system also complies with DC-MHS, UALink, and Ultra Ethernet Consortium (UEC) to support both open scale-up and scale-out fabrics.

Helios also features a double-wide layout for improved serviceability and standards-based Ethernet for multipath resiliency, and quick-disconnect liquid cooling on the backside for efficient thermal performance at high density.

“With the introduction of the new AMD ‘Helios’ and our purpose-built HPE scale-up networking solution, we are providing our cloud service provider customers with faster deployments, greater flexibility, and reduced risk in how they scale AI computing in their businesses,” said Antonio Neri, president and CEO at HPE.

“HPE has been an exceptional long-term partner to AMD, working with us to redefine what is possible in high-performance computing,” said Dr. Lisa Su, chair and CEO at AMD. “With ‘Helios’, we’re taking that collaboration further, bringing together the full stack of AMD compute technologies and HPE’s system innovation to deliver an open, rack -scale AI platform that drives new levels of efficiency, scalability, and breakthrough performance for our customers in the AI era.”