Safe Superintelligence Суцкевера заключила долгосрочное партнёрство с Nvidia

SafeSuperintelligence (SSI) — the startup founded by former OpenAI chief scientist Ilya Sutskever (Ilya Sutskever) — has announced a long-term partnership with Nvidia to scale its research after two years of operating in stealth mode. TechCrunch reported on the deal on July 27, 2026.

What happened

SafeSuperintelligence has entered into a long-term partnership with Nvidia — the largest maker of accelerators for training AI models. The company has emerged from two years of stealth mode and, according to the announcement's wording, is preparing for the "next phase" of its development. SSI did not disclose the specific terms of the deal — the amount of compute, the sum, or the timeline.

SafeSuperintelligence has made almost no public statements since its founding in 2024, so the announcement of a partnership with an infrastructure provider marks the lab's first major step into the public eye.

• Company — SafeSuperintelligence (SSI)
• Founder — Ilya Sutskever, former chief scientist at OpenAI
• Partner — Nvidia, long-term collaboration
• Announcement date — July 27, 2026, according to TechCrunch
• Before the announcement, the company spent two years operating in stealth mode

"A long-term partnership with

Nvidia as we prepare for the next phase," — from SafeSuperintelligence's announcement, as reported by TechCrunch.

Why SSI needs a partnership with Nvidia

The partnership with Nvidia gives SSI guaranteed access to compute power — the main scarce resource in the race for frontier AI models. Training frontier systems requires tens of thousands of GPUs, and Nvidia accounts for the bulk of that market. The long-term nature of the deal means SSI is reserving resources for scaling in advance, rather than buying them piecemeal.

Sutskever is building the company around a single goal — creating safe superintelligence — without intermediate commercial products. This approach requires massive investment in compute without quick revenue, which makes reliable access to hardware critical for the lab.

What is known about Sutskever's company

SafeSuperintelligence was founded in 2024 by Ilya Sutskever after he left OpenAI, where he had been one of the key scientific architects and took part in an attempt to remove the company's leadership in late 2023. According to Reuters, SSI raised around $1 billion at a $5 billion valuation as early as 2024, and by 2025, according to media reports, its valuation had grown to roughly $32 billion — despite the startup having no public product at all.

SSI (SafeSuperintelligence Inc.) has stated that its sole mission is research into safe superintelligence. Unlike OpenAI or Anthropic, the company has deliberately opted out of releasing intermediate services and has focused on a single long-term goal.

What this means

The partnership cements an alliance between two centers of power: Sutskever's scientific bet on "safe superintelligence" and Nvidia's computing might, which accounts for the bulk of the AI accelerator market. For the market, it's a signal: even a startup valued at tens of billions of dollars with no product to its name cannot scale without long-term access to Nvidia GPUs.

Frequently asked questions

Who founded SafeSuperintelligence?

The company was founded in 2024 by Ilya Sutskever — former chief scientist at OpenAI and one of the authors of key papers in deep learning.

What does SSI do?

According to the company itself, SSI's sole goal is creating safe superintelligence. Unlike OpenAI or Anthropic, the startup does not release intermediate commercial products.

What does the partnership with Nvidia provide?

The partnership gives SSI long-term access to Nvidia GPUs for training large models — a key resource whose scarcity is constraining the entire industry.