Hitachi Vantara expanded its Hitachi iQ portfolio to let enterprises run agentic AI on premises using nvidia Blackwell GPUs and new orchestration software, the company said.
The update bundles storage, networking and accelerated computing into a pre validated stack built on the Virtual Storage Platform One data platform, and adds air cooled and liquid cooled support for Blackwell Ultra GPUs. Hitachi said it also supports a 2U NVIDIA MGX based system with up to four NVIDIA RTX PRO6000 Blackwell Server Edition GPUs, and plans support for the NVIDIA RTX PRO 4500 Blackwell Server Edition GPU.
Hitachi Vantara said the stack keeps data close to compute to reduce bottlenecks for data intensive AI workloads and cited rising oversight and data readiness gaps as drivers for the release. Octavian Tanase, Chief Product Officer at Hitachi Vantara, said "AI is moving into production faster than many organizations' data foundations are ready to support," conveying why the firm focused on regulated and virtualised environments.
On the software side, Hitachi iQ Studio adds expanded AI blueprints and multi agent coordination that define supervisor and worker roles, and it supports NVIDIA Nemotron models aimed at agentic tasks and tool use. The studio also includes what Hitachi described as "time machine capabilities" to let AI systems work with historical datasets and added integration with Hammerspace for distributed data access through the Model Context Protocol.
Market Context And Nvidia Revenue Signals
Nvidia reported strong revenue and signalled broad enterprise momentum for agentic AI, which aligns with Hitachi's platform focus on on premise deployments. The company posted Q4 FY2026 revenue of $68.13 billion, up 73.2% year on year, with data center revenue of $62.31 billion representing 91.5% of the quarter's total, the company reported.
Jensen Huang, Nvidia's founder and chief executive, said "We're going to see agents in every single part of every single company, and it's going to help us become a lot more productive," arguing that digital agents are already deployed internally for chip design and software engineering. Nvidia said Blackwell Ultra hardware delivers up to 50 times better performance and 35 times lower cost for agentic AI workloads versus Hopper, a claim Nvidia used to frame broader enterprise adoption.
For the full fiscal year Nvidia reported $215.94 billion in revenue and guided Q1 FY2027 revenue to about $78 billion. Analysts and market services reacted, with Morningstar raising a fair value to $260 from $240 following GTC 2026, and the report noting 58 analysts rate Nvidia a buy or strong buy with an average target of $267.54 against a stated current price of $181.93. Hitachi positioned its stack to work with NVIDIA STX reference architecture elements including Vera Rubin, BlueField 4 and Spectrum X networking to align with Nvidia's infrastructure roadmap.
