Nvidia’s stock recently took a hit, but it remains a powerhouse in AI hardware. Recently, they introduced terms like “agentic AI” and “reasoning” at their GTC 2025 event in San Jose. Their new AI Data Platform is making waves, drawing interest from storage suppliers eager to get on board.
At the heart of their announcements was the Blackwell Ultra graphics processing unit (GPU), built for AI datacenter processing. This GPU is tailored for reasoning models, like DeepSeek R1, and significantly improves memory and inference performance. With Blackwell Ultra leading the charge, Nvidia is also rolling out various rack-scale platforms in its GB/NVL line, alongside new DGX family SuperPod clusters, workstations, network interface cards, and laptop GPUs.
This comes as a response to findings that DeepSeek is more efficient and less demanding on GPUs than models like ChatGPT. Nvidia is emphasizing the need for faster AI processing to maximize these advancements. Storage suppliers are also keenly aware of the demand for high I/O capabilities, akin to how pharmaceutical firms rely on data for drug development. Handling massive amounts of data for AI training requires extensive storage solutions capable of rapid data access.
Nvidia’s AI Data Platform reference architecture is central to this. It enables third-party suppliers, especially those in storage, to design products that meet Nvidia’s specifications, specifically for workloads involving agentic and reasoning techniques. Companies working alongside Nvidia include DDN, Dell, HPE, Hitachi Vantara, IBM, NetApp, Pure Storage, Vast Data, and Weka.
At GTC, these storage players made some notable announcements. DDN introduced its Inferno fast object appliance, integrating Nvidia’s Spectrum-X switch with its Infinia storage, which currently supports S3 object storage. Dell unveiled 20-petaflop-scale PCs designed for AI, and its PowerScale scale-out file system has now been validated for Nvidia’s Cloud Partner Program.
HPE highlighted its “unified data layer,” capable of handling both structured and unstructured data across the enterprise. They also announced upgrades for unified block and file access in their MP B10000 array. Hitachi Vantara launched the iQ M Series, merging its Virtual Storage Platform One with Nvidia’s AI Enterprise software, focusing on integrating the Nvidia AI Data Platform for agentic AI.
IBM’s collaboration with Nvidia included planned integrations based on the Nvidia AI Data Platform. They’re set to enhance their hybrid cloud offering, IBM Fusion, with content-aware storage capabilities and expand Watsonx integrations. NetApp received validation for SuperPOD and confirmed that its AFF A90 product meets DGX SuperPOD requirements. Their AIPod also earned the Nvidia-Certified Storage status to support Nvidia Enterprise Reference Architectures.
Lastly, Pure Storage announced compatibility with the Nvidia AI Data Platform shortly after its FlashBlade//Exa reveal. Vast Data introduced a new AI Stack combining its InsightEngine with Nvidia DGX products, BlueField-3 DPUs, and Spectrum-X networking. Weka achieved data store certification for Nvidia GB200 deployments and its WEKApod Nitro Data Platform Appliances are now certified for Nvidia Cloud Partner deployments with HGX H200, B200, and GB200 NVL72 products.