Nutanix storage wins NVIDIA AI enterprise certification

Nutanix's Unified Storage product has achieved NVIDIA enterprise certification for storage supporting production artificial intelligence workloads.

The certification gives customers a validated configuration for use with NVIDIA AI infrastructure. It strengthens Nutanix's position in a market where businesses and cloud providers are building larger AI systems around scarce and costly graphics processors.

Storage has become a key part of that build-out because AI systems depend on a steady flow of data to keep GPUs busy. Across the sector, vendors are trying to show their systems can avoid bottlenecks as organisations move from pilot projects to broader production deployments.

The certified setup is based on a 10-node all-NVMe cluster. It uses parallel NFS and GPUDirect Storage over NFS with RDMA to create a direct path between GPUs and storage, reducing latency and keeping data moving during demanding workloads.

According to Nutanix, the configuration can scale from 10 GB/s read and 5 GB/s write for 32 GPUs to 160 GB/s read and 80 GB/s write for 1,024 GPUs. The architecture is intended to support training, fine-tuning, inference and retrieval-augmented generation workloads across several NVIDIA-based compute platforms.

AI infrastructure

The certified design uses NVIDIA Spectrum-X Ethernet, including Spectrum-4 switches and BlueField-3 DPUs. The reference architecture is now available, while support for NVIDIA Vera BlueField-4 STX is planned for later this year.

The move comes as infrastructure suppliers seek to tie their products more closely to NVIDIA's ecosystem, which has become central to commercial AI deployment. Certification programmes have taken on added importance as customers look for tested combinations of compute, networking and storage that can be rolled out with lower integration risk.

Thomas Cornely, Executive Vice President of Product Management at Nutanix, said fragmented systems remain a problem for companies trying to build large AI environments.

"To build and run AI factories successfully, enterprises must move past fragmented infrastructure and data silos that limit GPU infrastructure efficiency," Cornely said.

"This NVIDIA certification validates that Nutanix Unified Storage delivers the full-stack interoperability, linear scalability, and reliable data velocity that modern AI workloads demand. By collaborating closely with NVIDIA, we are giving customers a unified, high-performance foundation to scale their production AI operations with confidence."

Jason Hardy, Vice President of Storage Technology at NVIDIA, said storage has become central as AI deployments expand.

"As enterprises scale their AI factory deployments to meet demanding agentic AI workloads, storage is foundational to unlocking full-stack performance, efficiency, and accuracy," Hardy said.

"Nutanix Unified Storage achieving NVIDIA certification gives customers a trusted, interoperable foundation to eliminate data bottlenecks, maximise GPU utilisation, and scale production AI workloads with confidence."

Regional push

Nutanix also highlighted the announcement in Asia-Pacific and Japan, where customers are looking for simpler ways to organise data and workloads for AI use. The region has become a priority market for infrastructure vendors as governments, telecommunications groups, financial institutions and cloud operators increase spending on AI-related systems.

Helder Queiros, Senior Director of OEM & Alliances for APJ at Nutanix, described the certification as part of that effort.

"With Nutanix Unified Storage achieving NVIDIA enterprise certification, we're empowering our customers across Asia-Pacific and Japan to build a more reliable and high-performing foundation for AI from day one. We are focused on helping APJ organisations create a unified and simplified environment for AI, where data and workloads can flow cohesively to drive maximum value," Queiros said.

Nutanix, which focuses on hybrid multicloud software and infrastructure, has been expanding its message around enterprise AI as customers look for ways to run newer workloads alongside existing applications and data estates. The company says it serves more than 30,000 customers worldwide.

The announcement underlines how competition in AI infrastructure is extending beyond chips and servers to the storage systems that feed them, with suppliers increasingly using third-party certification to show their products can operate within larger NVIDIA-centred deployments.