Progress You Can Measure: Nutanix Unified Storage Pushes AI Performance Further

AI is reshaping the enterprise, fueling predictive analysis, automation and GenAI models. As these workloads surge across industries, the ability to train, fine-tune and deploy models at scale is no longer optional – it's a competitive necessity.

Behind the curtain of every successful AI pipeline is a silent hero: Storage. Fast, consistent and scalable high-performance storage keeps GPUs fed, CPUs busy and the data scientists happy.

The Nutanix Unified Storage solution delivered benchmark results for the 2025 MLPerf Storage benchmark that reflect its continued efforts to support the performance needs of AI and ML workloads. The MLPerf benchmark is an industry leading measurement of storage performance for machine learning. In this latest round, Nutanix showed significant advancements that reflect our product vision and commitment to deliver on our strategic roadmap.

Purpose-built for the demands of modern AI

AI workloads generate unique pressure on infrastructure. They require:

• High throughput to feed massive model training jobs.
• Low latency for efficient data pre-processing.
• Linear scalability as datasets grow from terabytes to petabytes.
  

Unlike traditional storage solutions, Nutanix Unified Storage doesn’t force compromises. It blends performance with simplicity by offering file, object and block storage from a single, unified platform.

This architecture gives IT teams the ability to support AI workloads without managing silos. Nutanix Unified Storage is designed to meet the relentless demands of AI head-on and we do it by leveraging our software-defined architecture that’s designed for performance at scale.

MLPerf Storage: Why It matters

The MLPerf Storage benchmark is developed by MLCommons. It’s designed specifically to measure how well a storage system can support end-to-end AI and machine learning pipelines, from data ingesting and preprocessing to model training.

Unlike traditional benchmarks that evaluate IOPS or raw throughput in isolation, MLPerf Storage focuses on real-world machine learning scenarios, such as:

• Reading large training datasets (e.g., ResNet50, Unet3D).
• Performing randomized access patterns common to deep learning frameworks.
• Supporting multihost, multithreaded AI pipelines.

That’s why MLPerf Storage has become a key benchmark for enterprises evaluating AI- and machine learning-ready storage.

The results: Nutanix delivers exceptional performance

In this year's MLPerf Storage benchmark, Nutanix Unified Storage posted strong results:

• We demonstrated that a single cluster can serve up to 2,312 accelerators when running the resnet50 benchmark, a test that evaluates the ability to recognize image content (i.e., does a photo contain a dog, a cat, people, cars, etc). 
• Combining improvements in software and newer hardware, we produced exceptional results on a per node basis from the prior year.
• A sample of some of our results over the last two years (per node).

Note: This reflects the growth from our first submission back in 2023 to now with just a single test example.

Nutanix delivered the above results without implementation of proprietary hardware or special client-side drivers. Rather, the results were achieved using the NFS protocol and a file server architecture engineered for scalability. We showed the ability to deliver a level of performance that AI workloads demand. 

The results show that Nutanix can sustain heavy AI loads critical for organizations running these jobs across large GPU clusters. Why is this important? It highlights that Nutanix is a platform that can easily manage today’s demands with performance to spare to accommodate tomorrow’s challenges.

What’s behind the numbers?

Our strong showing is a result of our efforts to integrate game-changing performance enhancements into our product. A few new product highlights include:

• GPU direct storage – Increase performance ability by leveraging a direct path between storage and GPU memory. This bypasses the CPU and the associated overhead. While GPU direct storage isn’t specifically evaluated during an MLPerf run, it is a key driver to supporting the performance needed for AI workloads.
• NFS over RDMA support – Reduces latency and CPU overhead for environments leveraging GPU-heavy training. Last year we delivered back-end enhancements via iSER (iSCSI over RDMA). This year we completed the story for the client side.
• SR-IOV – Single root IO virtualization (SR-IOV) allows a physical device, typically a physical network adapter, to be passed through to a VM. This allows the IO to bypass the vSwitch in our AHV hypervisor, which reduces overhead and provides near physical machine network performance.
• High-performance networking – The benchmark results combined Nutanix Unified Storage software advancements with modern high-performance networking components to reduce latency and increase throughput for data-intensive applications.

Our product team is highly observant of the needs of modern workloads and is honing in on the improvements needed to meet the demands of the ever-evolving landscape of modern workloads.

From datacenter to AI at the edge

Whether you’re building a centralized AI cluster or deploying inference at the edge, Nutanix Unified Storage gives you the flexibility to:

• Run on your choice of hardware and clouds.
• Start small and scale incrementally.
• Keep your storage operations simple, even in complex hybrid and multicloud environments.

In today’s dynamic IT landscape, flexibility is essential, but it also needs to be matched with robust capabilities that drive business outcomes. Nutanix Unified Storage supports AI workloads on-prem in your data center, in the cloud on AWS and Azure, providing the resilience and ease of management that enterprises demand. We adapt to evolving global needs with a secure platform built for scale, simplicity and control.

A platform that delivers more

Performance is critical, but it’s only part of the story. Nutanix Unified Storage also delivers:

• Resiliency – Enterprise-grade data protection, high availability and snapshot support.
• Security – Role-based access control, data encryption and ransomware threat behavior monitoring.
• Simplicity – One platform, one interface, one support experience.

Whether you’re an AI researcher, data engineer or IT administrator, the goal is the same: Get results faster without complexity getting in the way.

The latest MLPerf Storage results are more than just numbers, they’re proof of the progress we’ve made in delivering a high-performance, enterprise-ready storage platform built to handle  the demands of AI. But performance is only part of the equation.

Nutanix Unified Storage also brings the operational simplicity, security and scalability that IT leaders need to support AI initiatives with confidence. Whether you're scaling GPU clusters, managing hybrid environments or preparing for the next wave of innovation, Nutanix provides a unified foundation that helps to accelerate outcomes and reduce complexity.

Ready to experience it?

As AI grows more demanding, we’re ready for what’s next. If you're building or scaling your AI initiatives and need a storage platform that performs under pressure, we invite you to reach out to your Nutanix account team or partner.

As we said last year, this is just the beginning of remarkable things to come. We are still working hard to further improve Nutanix Unified Storage, and there’s more greatness coming in the near future.

_{©2025 Nutanix, Inc. All rights reserved. Nutanix, the Nutanix logo and all Nutanix product and service names mentioned are registered trademarks or trademarks of Nutanix, Inc. in the United States and other countries. All other brand names mentioned are for identification purposes only and may be the trademarks of their respective holder(s).Certain information contained in this content may link or refer to, or be based on, studies, publications, surveys, and other data obtained from third-party sources and our own internal estimates and research. While we believe these third-party studies, publications, surveys, and other data are reliable as of the date of publication, they have not been independently verified unless specifically stated, and we make no representation as to the adequacy, fairness, accuracy, or completeness of any information obtained from a third-party. Our decision to publish, link to or reference third-party data should not be considered an endorsement of any such content.This content reflects an experiment in a test environment. Results, benefits, savings, or other outcomes described depend on a variety of factors including use case, individual requirements, and operating environments, and this publication should not be construed as a promise or obligation to deliver specific outcomes.This content may contain express and implied forward-looking statements, including but not limited to statements regarding our plans and expectations relating to new product features and technology under development with respect to Nutanix Unified Storage and other Nutanix products, the capabilities of such product features and technology, and our plans to release product features and technology. Such statements are not historical facts and are instead based on our current expectations, estimates and beliefs, including statements about The accuracy of such statements involves risks and uncertainties and depends upon future events, including those that may be beyond our control, and actual results may differ materially and adversely from those anticipated or implied by such statements, including, among others: failure to develop, or unexpected difficulties, delays or disruptions in developing, releasing or distributing, new products, services, product features or technology on a timely or cost-effective basis. Any forward-looking statements included speak only as of the date hereof and, except as required by law, we assume no obligation to update or otherwise revise any such forward-looking statements to reflect subsequent events or circumstances. Certain products and features or functionalities described herein remain in varying stages of development and will be offered on a when-and-if-available basis. The development, release, and timing of any such products, features or functionalities are subject to change. Nutanix will not have any liability for any failure to deliver or delay in the delivery of any such products, features or functionalities. Any future product or product feature information is intended to outline general product directions, and is not a commitment, promise, or legal obligation for Nutanix to deliver any functionality. This information should not be used when making a purchasing decision.}