NetApp has agreed to acquire PEAK:AIO, the Manchester software company behind a scale-out metadata and parallel NFS stack for GPU clusters. It is a sign that standard NFS, not proprietary parallel file systems, is where mainstream AI storage is heading, but PEAK:AIO customers now face a roadmap question.
What was announced
- NetApp announced its intent to acquire PEAK:AIO on 25 September 2026. Terms were not disclosed and the deal is subject to customary closing conditions and regulatory approvals.
- NetApp says it will combine PEAK:AIO's metadata services, global namespace and parallel NFS access with ONTAP, in an architecture that separates metadata from data so each can scale independently, aimed at trillions of files and multi-exabyte deployments.
- NetApp's own release states that its description of future integration should not be relied on for purchasing decisions and that it has no obligation to deliver the described capabilities.
- PEAK:AIO was founded in 2021 and is based in Manchester. Blocks & Files reports it raised around 7 million dollars in seed funding in October 2025.
- PEAK:AIO initiated pNFS Lattice, an open-source scale-out metadata server for NFSv4.2 and pNFS Flex Files, with Los Alamos National Laboratory and Carnegie Mellon University. It is MIT licensed and labelled an early release, not yet recommended for production.
- In StorageReview testing PEAK:AIO's software ran on standard servers: a 2U Dell-based AI Data Server with 1.5 PB delivered 120 GB/s, and on a Dell PowerEdge R7725xd it served 160 GB/s over NVMe-oF RDMA to two clients across four 200Gb links.
- This is NetApp's second AI-focused acquisition in about two months, following DataPelago in July 2026.
The ETON view
The interesting part of this deal is not NetApp getting bigger in AI, it is the technology it chose. PEAK:AIO's pitch was high-throughput shared storage built from ordinary servers, NVMe drives and fast NICs, served over standard NFS rather than a proprietary client. NetApp buying that approach tells you where mainstream AI storage is going: parallel NFS on standard Linux clients, with metadata scaled out separately. That lowers the barrier for hosting providers and enterprises who do not want to run a specialist parallel file system just to keep GPUs fed.
There is a caveat for anyone already running or evaluating PEAK:AIO. NetApp's release is explicit that its integration plans are not a commitment, and its language is aimed at AI clouds at hyperscale, which is a different customer from the university, hospital or single GPU server deployments where PEAK:AIO built its following. If you bought into it for independence and low entry cost, ask for the support and licensing position in writing before your next expansion. The open-source Lattice project is worth watching, but it is labelled early release and is not a production answer today.
For storage refreshes generally, this reinforces a vendor-neutral route. Whatever software layer you choose, the hardware underneath is the same class of dense NVMe server, Gen5 drives and 200Gb or 400Gb networking, and that is where price and lead time are decided. We can quote UK availability on NVMe-heavy storage servers, enterprise SSDs and the NICs to match, and keep the software decision separate so a change of ownership upstream does not strand the hardware.
Related infrastructure
Category: Storage · Vendor: NetApp · Technology: Parallel NFS (pNFS), NFSv4.2, metadata disaggregation, NVMe-oF · Last verified: 26 Sep 2026 · ~2 min read
Sourcing this kind of infrastructure? Talk to ETON about availability, lead time and pricing.
