NetApp says it intends to acquire Manchester-based PEAK:AIO, a software-defined storage company, to add technology designed to help customers manage the huge numbers of files generated by large AI systems. Financial terms were not disclosed, and the proposed deal remains subject to customary closing conditions and regulatory approvals.
The announcement, made on 25 September, is about a less visible part of AI infrastructure. Graphics processors do the intensive calculations, but training and running AI models also requires systems that can find, share and move vast quantities of data without leaving expensive computing equipment waiting for files.
Why file management is becoming an AI constraint
AI projects can involve millions or billions of small and large files, from training data and model checkpoints to logs and outputs. A storage system needs more than raw capacity. It must keep track of where files are, who can access them and how many systems are asking for them at once.
That recordkeeping layer is called metadata. A file’s name, location, ownership and permissions are all examples. If the metadata service becomes a bottleneck, GPUs can be left idle while workers or software wait for data.
NetApp says PEAK:AIO’s architecture separates metadata from data so that metadata services can be expanded independently. The company also said the combined technology is intended to support a shared file namespace, meaning users and applications can see files through one logical view even when capacity sits across multiple systems.
What the planned deal would add
PEAK:AIO provides storage software designed for high-performance computing and AI environments. NetApp plans to combine that technology with ONTAP, its established data-management platform.
The companies say the result is intended to give customers parallel access to shared storage through NFS, a widely used network file-sharing standard. Parallel access lets many computing processes read or write data at the same time. That matters when a cluster of GPUs is training a model or processing large datasets together.
NetApp describes the target architecture as capable of supporting trillions of files and multi-exabyte deployments. Those are future product-direction statements, rather than capabilities customers can assume are available today. An exabyte is one billion gigabytes.
Storage is part of a wider infrastructure race
The transaction shows how suppliers are expanding beyond individual server components. An AI installation needs processors, networking, power, cooling and storage that can operate together at high utilization.
Our recent report on HPE’s AI servers and networking sales explains why faster connections between computing systems have become commercially important. NetApp’s proposed acquisition addresses a related issue: making the data available to those systems quickly enough.
PEAK:AIO says it is already used at organizations including Los Alamos National Laboratory, the Oxford Robotics Institute and several UK universities. NetApp’s announcement does not disclose customer revenue, a purchase price or the timing for a completed integration.
What customers and investors should watch
For existing NetApp customers, the practical question is whether the proposed technology can be integrated without making storage management more difficult. NetApp says it intends to provide an evolution path from ONTAP to the new architecture, but it has not committed to specific product releases or dates.
For NetApp, the deal is a bet that AI infrastructure buyers will pay for a storage layer designed around much higher concurrency and larger file counts than conventional enterprise workloads. It also places the company in a competitive market that includes specialist file-system vendors, cloud providers and companies selling integrated AI systems.
The key distinction is that NetApp has announced an intention to acquire PEAK:AIO, not a completed transaction or a finished product. The value of the deal will depend on regulatory clearance, integration and whether customers adopt the combined technology at scale.
NetApp announced the planned acquisition in its 25 September release.
