Vultr’s Next-Gen AI Data Centers to Use HPE Systems Integrated With NVIDIA Networking and Accelerators

Vultr, described by Hewlett Packard Enterprise as the world’s largest privately held hyperscaler, has chosen HPE and NVIDIA to build its next-generation AI data center infrastructure for cloud-scale demand, HPE said. The announcement was made during HPE Discover Las Vegas 2026 on June 17 and centers on deploying systems aimed at private cloud use cases, model training, and large-scale AI inference.

Under the plan, HPE will supply and integrate NVIDIA GB300 NVL72 systems, then connect those servers with NVIDIA Spectrum-X Ethernet networking, HPE said. The pairing is presented as an end-to-end approach combining accelerated computing and networking to support AI workloads that require both high compute density and high bandwidth interconnects.

HPE framed the effort as part of NVIDIA’s “AI Computing by HPE” portfolio, a joint offering intended to bundle NVIDIA’s computing and networking stack with HPE’s systems integration and deployment capabilities. HPE said this architecture is designed to address rapidly growing enterprise demand for AI services delivered through cloud infrastructure.

Beyond the hardware selection, HPE said it would provide what it calls its full “AI factory architecture.” In HPE’s description, that includes rack-scale system design, liquid cooling technology for high-density GPU clusters, specialized AI services, deployment expertise, and long-term lifecycle support for the installed systems.

Hewlett Packard Enterprise President and Chief Executive Officer Antonio Neri said the partnership is intended to build AI infrastructure designed for the next phase of AI adoption, including what he called the agentic enterprise. He also said Vultr represents a newer generation of AI cloud providers and that its selection of HPE underscores the importance of data center architectures built for the next wave of global AI growth.

For NVIDIA, the update reinforces its push to be more than a chip supplier, positioning its accelerators and networking as part of broader infrastructure platforms delivered and supported by partners like HPE. For HPE, the deal highlights a continuing bet that its value in AI infrastructure will come from integrating compute and networking, and from packaging deployment and lifecycle services that operators can scale across fleets.

The company did not disclose timelines, the number of data center sites, expected capacity, or whether the deployments are incremental or tied to a multi-year procurement agreement. It also did not provide performance benchmarks, pricing, or details on software stacks beyond describing the architecture and components at a systems-and-network level.

Investors and customers will likely watch for further specificity in later announcements, such as deployment schedules, any expansion beyond the initially selected NVIDIA GB300 NVL72 and Spectrum-X networking components, and how HPE describes scaling the liquid-cooled “AI factory” approach across additional Vultr regions.