HPE Lands First Helios Order: Vultr Commits $1.2 Billion for AMD-Powered AI Racks

Hewlett Packard Enterprise said Vultr, described as the world's largest privately held cloud infrastructure company, has placed a $1.2 billion order to deploy AMD Helios AI Rack by HPE systems across Vultr's U.S. cloud data centers. HPE called it the company's first order for the new AMD Helios system, a rack-scale platform built around AMD's next-generation Instinct GPUs and purpose-built HPE networking gear.

What HPE and Vultr Announced

HPE made the announcement via Business Wire from Sunnyvale, California, on September 30, 2026. According to HPE's statement, the order covers deployment of AMD Helios AI Rack by HPE systems at Vultr's U.S. data center locations, and HPE says it is the company's first order for the new Helios system, which the company says features purpose-built HPE Networking scale-up switching and software.

HPE CEO Antonio Neri framed the deal in broad terms, saying, "AI is driving the largest infrastructure buildout in history and together with AMD and Vultr, we are building an open foundation to accelerate the next phase of AI... AMD Helios by HPE demonstrates HPE's unique ability to engineer the complete AI architecture with rack-scale compute, scale-up networking, and direct liquid cooling, all delivered by our global services," according to HPE's announcement. Vultr CEO J.J. Kardwell said "demand for high-performance AI infrastructure continues to outpace available capacity, and our work with AMD and HPE enables us to bring Helios rack-scale compute and scale-up networking online for customers faster," per the same release.

HPE also said the order builds on nearly three years of collaboration between Vultr and Juniper Networks, the networking vendor HPE acquired, extending that relationship now that Juniper operates inside HPE.

Inside the Helios Rack

Each Helios rack integrates 72 AMD Instinct MI455X GPUs paired with AMD EPYC "Venice" CPUs, AMD Pensando Vulcano AI networking interface cards, and AMD's ROCm software stack, per HPE's release. On the networking side, HPE says its content includes six HPE Juniper Networking QFX5252 scale-up Ethernet switch trays per rack, which connect all 72 GPUs using what the company describes as high-bandwidth, low-latency, standards-based Ethernet, while supporting open interconnect standards such as UALink over Ethernet.

AMD has said the Helios architecture, built on the Open Compute Project's Open Rack Wide design originated by Meta, can deliver up to 2.9 exaFLOPS of FP4 performance per rack along with 31TB of HBM4 memory, according to reporting from W.Media and TechRadar. Those figures come from AMD, not from independent lab testing.

Ethernet Versus NVLink

The scale-up fabric behind Helios, developed with Broadcom, is built on Ethernet rather than Nvidia's proprietary NVLink interconnect. Reporting from Yahoo Tech notes this is a deliberate contrast: Nvidia's GB200 NVL72 rack keeps 36 Grace CPUs and 72 Blackwell GPUs inside a single NVLink domain and relies on InfiniBand for system-to-system traffic, whereas Helios uses an Ethernet-based approach intended to offer an open alternative.

This is where reported specifications give way to interpretation. Comparable GPU counts on paper do not guarantee comparable real-world throughput, and TechRadar reported that performance under sustained multi-node workloads remains untested, with actual results depending on cooling effectiveness, network traffic handling and software optimization. In other words, the architectural bet on open Ethernet fabrics versus Nvidia's closed NVLink stack is a competitive positioning claim at this stage — clearly labeled here as interpretation — not a proven performance outcome.

Context: HPE's Networking Push

The Vultr order arrives around the time of an HPE investor day. According to Investor's Business Daily, HPE shares had climbed 160% in 2026 heading into the late-September event, which was expected to focus on HPE's networking business following the roughly $14 billion, 2025 acquisition of Juniper Networks. Interpretation: the Vultr deal, which includes HPE Juniper Networking QFX5252 switch trays inside each Helios rack, gives HPE a concrete data point for how that acquisition fits its AI infrastructure pitch.

Bottom Line

HPE has secured a large, named customer for the first deployment of AMD's Helios rack architecture, with Vultr's $1.2 billion order serving as an early test — an interpretation, not a source conclusion — of whether an Ethernet-based, open rack-scale design can compete with Nvidia's more established NVLink ecosystem. The specifications AMD and HPE have published are notable on paper, but independent reporting has noted that performance under sustained real-world workloads remains untested. For now, the announcement confirms a first order and a technology direction; how the racks perform once running in Vultr's data centers is the open question ahead.