Vultr Selects HPE and NVIDIA GB300 for AI Cloud Expansion

Vultr has selected HPE and NVIDIA technology for a new generation of AI-focused data center deployments aimed at supporting growing enterprise demand for private cloud and AI workloads. The initiative will utilize NVIDIA GB300 NVL72 systems delivered through the NVIDIA AI Computing by HPE portfolio and connected via NVIDIA Spectrum-X Ethernet networking. The infrastructure is designed to support large-scale AI model training and inference while serving as the foundation for Vultr’s expanding global AI cloud platform.

The deployment extends Vultr’s strategy of building globally distributed AI infrastructure optimized for enterprise AI applications. Future facilities will incorporate rack-scale GPU systems, 400GbE and 800GbE Spectrum-X networking, optical interconnects, NVIDIA SuperNICs, and HPE liquid-cooling technology. HPE will provide system integration, deployment services, and operational support, drawing on experience from large-scale HPC and supercomputing projects.

Vultr said the investment supports increasing demand for AI compute resources delivered closer to users and enterprise locations. The company is positioning its AI cloud platform around decentralized, latency-sensitive AI applications while expanding GPU-as-a-service offerings for enterprise and private cloud deployments. The announcement reflects continued investment by independent cloud providers seeking to build alternatives to hyperscaler-operated AI infrastructure.

• Vultr selected NVIDIA GB300 NVL72 systems supplied through HPE.
• AI clusters will use NVIDIA Spectrum-X Ethernet networking with 400GbE and 800GbE interconnects.
• Future deployments include optical transceivers, Spectrum-X switches, and NVIDIA SuperNICs.
• HPE will provide AI factory architecture, liquid cooling, deployment services, and lifecycle support.
• Vultr is expanding globally to support enterprise AI training, inference, and private cloud workloads.
• The deployment targets decentralized and latency-sensitive AI applications.

“As a powerful extension of our global cloud infrastructure platform, Vultr is deploying dedicated AI infrastructure focused on GPU architecture and AI inference to accelerate customer innovation while maintaining cost efficiency,” said J.J. Kardwell, CEO of Vultr.

🌐 Analysis

This win highlights HPE’s growing position as a systems integrator for large-scale AI factories built around NVIDIA technology. Over the past year, HPE has expanded its AI portfolio through the NVIDIA AI Computing by HPE initiative, combining accelerated compute, networking, liquid cooling, software, and deployment services into turnkey AI infrastructure offerings. The company’s strategy increasingly resembles its historic role in supercomputing, where customers rely on HPE to integrate complex compute, networking, cooling, and operational systems rather than sourcing individual components.

The announcement is also notable because it comes from a rapidly growing class of independent AI cloud providers. Companies such as Vultr, CoreWeave, Crusoe, Lambda, and others are investing heavily in AI infrastructure to provide alternatives to traditional hyperscale cloud platforms. These providers often focus on specialized GPU services, sovereign AI deployments, lower-cost inference environments, and geographic coverage that places AI resources closer to enterprise users.

From a networking perspective, Vultr’s adoption of NVIDIA Spectrum-X reinforces the momentum behind Ethernet-based AI fabrics. NVIDIA has aggressively positioned Spectrum-X as an alternative to InfiniBand for large-scale AI deployments, while the broader industry continues to debate the roles of Ethernet, Ultra Ethernet, InfiniBand, and emerging scale-up interconnects. The use of 400GbE and 800GbE networking alongside GB300 NVL72 systems underscores how network architecture is becoming a critical differentiator in AI infrastructure deployments, particularly as inference workloads scale globally and require efficient east-west traffic movement across GPU clusters.