NetApp Completes Acquisition of UK AI Storage Firm PEAK:AIO

NetApp said it will add PEAK:AIO's metadata services and parallel namespace technology to its ONTAP-based data architecture. ONTAP is the data management software that runs on NetApp's storage systems. The company said the technology is meant to support file systems at zettabyte scale for AI clouds and GPU-as-a-service providers, and to back NetApp Novus, its high-performance storage offering. It also said the deal gives existing ONTAP customers a way to upgrade.

According to the release, PEAK:AIO's platform delivers high-performance storage for AI on industry-standard hardware and can scale from a single server to exabytes. Its customers include Los Alamos National Laboratory, Carnegie Mellon University and the NHS AIDE programme.

NetApp chief executive George Kurian said that "AI infrastructure is reaching unprecedented scale, demanding a new approach to storage built from the ground up." Chief product officer Syam Nair said that "the AI factory is only as productive as its data pipeline". He said combining the two companies' technology would let performance scale with the number of concurrent users while relying on open standards.

PEAK:AIO founder and chief technology officer Mark Klarzynski said the deal would let the company combine its technology with "the reach, expertise, and trusted customer relationships" of a global data infrastructure provider.

NetApp's forward-looking statements refer to parallel NFS-based access as a planned capability for massively parallel workloads. The company also said its plans for the PEAK:AIO technology could change, and that it is not committing to deliver any specific product, feature or timing. Industry analysts have described the purchase as a technology tuck-in that follows NetApp's acquisition of DataPelago in July.

The deal shows storage vendors adjusting their products for AI training and inference clusters, where thousands of GPUs read data at the same time. Data pipelines that cannot keep pace leave expensive accelerators sitting idle. For operators building GPU capacity, including the sovereign and commercial AI clouds now going up in the Gulf, storage throughput and metadata handling are becoming part of the hardware decision alongside chips and power.