NVIDIA's new storage processor enforces security policies in silicon at 800Gb/s, targeting the emerging threat surface of autonomous AI agents.
NVIDIA's new storage processor enforces security policies in silicon at 800Gb/s, targeting the emerging threat surface of autonomous AI agents.
As enterprises shift from AI chatbots to autonomous agents that read, write and share data without direct human oversight, storage infrastructure has become a real-time security control point — and NVIDIA is embedding defenses directly into the data path. The company on Sunday unveiled the Vera BlueField-4 STX, an accelerated storage processor that enforces file access, network isolation and runtime threat detection in silicon at speeds of up to 800Gb/s, extending its DOCA security stack to the AI storage layer.
"Agentic AI turns enterprise data into a living, real-time system — and that system must be protected where data moves, where context is stored and where agents act," Jensen Huang, founder and chief executive officer of NVIDIA, said at GTC Taipei.
The Vera BlueField-4 STX platform introduces three new DOCA security capabilities. DOCA Vault microservices ensure only authorized AI workloads access specific files with correct permissions. DOCA Argus provides visibility into agent behavior and AI workload activity. DOCA Flow isolates network traffic across multi-tenant AI environments. Together, they deliver runtime threat detection up to 1,000x faster than existing agentless solutions, with policy enforcement that does not tax host CPU resources, according to NVIDIA.
The announcement comes as NVIDIA's Vera Rubin platform — the third-generation MGX rack-scale system that integrates the Vera BlueField-4 STX — ramps into full production across 350-plus factories in 30 countries, with 150 supply chain partners in Taiwan alone. Vera Rubin delivers 10x agent throughput at scale compared with the prior-generation Grace Blackwell platform, the company said. Production shipments of Vera Rubin begin this fall, while STX-based platforms from partners are expected in the second half of 2026.
The Security Gap in Agentic AI
Autonomous agents create new exposures because they continuously access proprietary data and context memory without human supervision. Traditional perimeter-based security models cannot inspect agent-to-data interactions at AI factory speeds. NVIDIA's approach embeds policy enforcement directly in the BlueField-4 silicon, enabling zero-trust file access and multi-tenant network isolation without slowing data throughput.
Storage providers building on the STX platform include Dell Technologies, Hewlett Packard Enterprise, IBM, NetApp, VAST Data and WEKA, alongside Cloudian, DDN, Hitachi Vantara, MinIO and Nutanix. Manufacturing partners AIC, ASUS, Foxconn, Gigabyte, Quanta Cloud Technology, Supermicro, Wistron and Wiwynn are developing STX-based systems. Global systems integrators Accenture, Deloitte and Worldwide Technology are bringing the solutions to enterprise customers.
What It Means for Investors
NVIDIA's expansion beyond GPUs into storage security strengthens its moat in the AI infrastructure stack at a time when its Data Center revenue surged 92% year over year to a record $81.6 billion in the fiscal first quarter. The Vera BlueField-4 STX positions NVIDIA to capture a share of enterprise storage security spending as agentic AI workloads scale, while the broader Vera Rubin platform — with seven purpose-built chips across five racks — threatens to displace incumbent networking and storage vendors. Cloud providers CoreWeave, Lambda, Microsoft Azure and Oracle Cloud Infrastructure are among the early adopters of NVIDIA's confidential computing and networking technologies.
Nvidia shares trade at roughly 35x forward earnings. The company's deepening vertical integration from compute to networking to storage security gives it pricing power and customer lock-in that competitors including AMD and Intel have struggled to match.
This article is for informational purposes only and does not constitute investment advice.