Nvidia Safe Superintelligence Partnership Expands Compute

Nvidia Corp. (NVDA) said July 27, 2026, that its Safe Superintelligence partnership will give Ilya Sutskever’s lab access to Nvidia’s Vera Rubin compute systems, expanding its capacity by an order of magnitude. The company also announced the launch of the Open Secure AI Alliance, an industry coalition focused on security for open AI models.

SSI Partnership Expands Compute Access

Nvidia announced in a GlobeNewswire press release that it has formed a long-term strategic partnership with Safe Superintelligence Inc., the lab founded by former OpenAI co-founder and chief scientist Ilya Sutskever. The partnership aims to scale the lab’s AI research after two years in stealth.

The release said Safe Superintelligence will gain access to Nvidia’s Vera Rubin systems, increasing its compute capacity by tenfold. It did not specify FLOPS, GPU counts, or rack numbers. People familiar with the deal described Nvidia’s commitment as substantial, though financial terms were not disclosed.

Separate reports indicate that the Vera Rubin platform is in full production and shipping to major AI customers including OpenAI, CoreWeave, Google Cloud, Microsoft Azure, Meta, and Dell. Nvidia’s NVL72 Vera Rubin cabinets deliver token-processing throughput up to 10 times that of the previous generation and faster Python execution than some third-party processors.

Open Secure AI Alliance Launches With Industry Partners

In an official blog post, Nvidia announced the Open Secure AI Alliance to focus on security and safety for open AI models—systems whose core components are accessible and inspectable. The alliance aims to remediate and disclose vulnerabilities using open technologies.

The inaugural members include cloud providers, cybersecurity firms, enterprise software vendors, and open-source foundations such as Adobe, Cisco, CrowdStrike, Databricks, Dell Technologies, HPE, Hugging Face, IBM, the Linux Foundation, Microsoft, Palantir, Red Hat, Salesforce, SAP, Siemens, and Snowflake.

Nvidia will contribute open models, model weights, data, and “new agent harness research.” It published the NVIDIA Labs Object-Oriented Agent (NOOA) framework as an open-source research preview on GitHub, providing code and an evaluation write-up to help developers test, track, audit, and govern agent behavior. The alliance will build on the Linux Foundation’s Akrites initiative and the Open Source Security Foundation (OpenSSF) community for coordinated vulnerability disclosure. Near-term activities include developing reference architectures and security toolkits tailored to government and critical-infrastructure use cases.

Strategic Positioning and Policy Context

Nvidia framed the alliance launch as a response to the recent Hugging Face security incident, which highlighted the need for defenders to have inspectable, frontier agentic systems for self-defense. The launch follows a July 24, 2026, letter titled “Open Weights and American AI Leadership,” in which a coalition of companies urged U.S. policymakers to avoid premature restrictions on open-weight AI models that could stifle innovation or push it overseas.

Together, the Safe Superintelligence compute deal and the Open Secure AI Alliance deepen Nvidia’s role as an infrastructure provider to OpenAI-linked talent and other prominent labs. They also position the company at the center of emerging standards for AI cybersecurity, vulnerability disclosure, and open-model governance. Industry observers note these moves align Nvidia commercially with elite research customers while advancing a policy argument that openness combined with safeguards, rigorous evaluation, and rapid remediation is the appropriate approach for defensive AI tooling.

Quote

“The Open Secure AI Alliance will work to remediate and disclose vulnerabilities using open technologies.” — Nvidia blog