Vultr has placed a $1.2 billion order with Hewlett Packard Enterprise for AMD Helios racks at its U.S. data centers. HPE announced the deal on September 30, marking its first order for the new system. The cloud provider plans to support customers running AI model training and inference workloads. The purchase brings AMD computing hardware and HPE networking into a substantial commitment to expand American AI infrastructure.

The agreement extends nearly three years of collaboration between Vultr and Juniper Networks, now part of HPE. Vultr CEO J.J. Kardwell said demand continues to “outpace available capacity,” describing the commercial pressure behind the investment. HPE CEO Antonio Neri described the partners as “building an open foundation to accelerate the next phase of AI.” Those ambitions now carry a disclosed order value, giving HPE a customer commitment for its integrated system.

AMD Helios Racks Combine Compute and Networking

Each rack combines 72 AMD Instinct MI455X GPUs with AMD EPYC “Venice” processors and AMD ROCm software. AMD Pensando Vulcano AI network interface cards form another part of the hardware configuration. Six HPE Juniper Networking QFX5252 Ethernet switch trays connect the GPUs within each rack. The announcement makes networking a defined component of the purchase, alongside the processors that perform AI calculations.

HPE positions Helios for training trillion-parameter models and serving high volumes of inference requests. Its architecture supports open interconnect technologies, including UALink over Ethernet, within a design that addresses power, cooling and serviceability. HPE contributes liquid-cooling expertise and deployment support to the system. The product sits within a broader portfolio that combines accelerated computing, networking, software and services for demanding AI workloads.

The strategic significance lies in how those components work together, an inference supported by HPE’s integrated product design. A rack purchase creates requirements for installation, validation and ongoing support across the equipment that delivers the service. HPE markets its role across those stages, including performance tuning and lifecycle assistance. For Vultr, the approach offers a way to source an integrated platform for the cloud capacity it intends to sell.

Customer Reservations Precede the Hardware Announcement

Vultr outlined support for Helios and MI455X GPUs in a July 23 announcement, ahead of the HPE order disclosure. Its published offering targets training, inference, fine-tuning and agentic AI workloads through its cloud platform. The company described deployment choices spanning bare-metal and virtualized infrastructure, supported by cloud orchestration tools. That earlier announcement established the customer offering around the hardware platform now underpinning the procurement news.

However, customers should distinguish the equipment commitment from a confirmed date for accessing production capacity. Vultr’s July announcement invited reservations for deployments during 2027 and 2028. That window provides context for its broader Helios plans, rather than a delivery schedule for this specific HPE order. Buyers planning workloads still need confirmation of their own allocation, location and service commencement date.

HPE Raises Its Networking Growth Expectations

Meanwhile, HPE announced the Vultr order alongside higher growth expectations at its Networking Investor Day. The company now forecasts fiscal 2027 networking revenue growth in the high teens to low-20s percentage range. It projects a high-teens compound annual growth rate for the segment from fiscal 2026 through fiscal 2029. Those targets place the Helios win within a larger effort to turn AI infrastructure demand into networking sales.

HPE expects its data center networking business to deliver compound annual revenue growth in the 50s percentage range through fiscal 2029. The company raised its Juniper-related annual savings target to $800 million by the end of fiscal 2028. Its previous target stood at at least $600 million, making integration savings another part of the financial story. The Vultr contract provides a concrete example of how HPE can combine Juniper networking with its computing and deployment capabilities.

Deployment Details Remain the Next Milestone

HPE’s order announcement does not disclose the number of racks, individual deployment sites or a delivery timetable. Those omissions limit what customers can infer about when the investment will translate into usable cloud capacity. The disclosed value establishes the scale of the procurement commitment, without establishing the volume available to any particular buyer. Delivery milestones and customer availability will determine how that commitment translates into working AI infrastructure.