Safe Superintelligence

Safe Superintelligence (Safe Superintelligence Inc., or SSI) is a private research laboratory dedicated to a single stated mission: building safe superintelligence as fast as possible, with no product, no revenue and no near-term commercial timetable. It was founded in June 2024 by Ilya Sutskever, former chief scientist and co-founder of OpenAI, together with Daniel Gross and Daniel Levy.^[1]

History

Sutskever incorporated SSI on 19 June 2024, shortly after his departure from OpenAI. In September 2024 the company raised about $1 billion at a $5 billion valuation from Sequoia Capital, Andreessen Horowitz, DST Global and SV Angel — at the time one of the largest seed rounds ever recorded for a pre-product company.^[2]

In April 2025 SSI raised a further $2 billion at a $32 billion valuation, led by Greenoaks Capital with Alphabet and NVIDIA joining as strategic investors, and signed an agreement to use Google Cloud for research compute. Daniel Gross left in July 2025 to join Meta's superintelligence effort, leaving Sutskever as chief executive and Levy as president.^[1][3]

On 27 July 2026, NVIDIA announced a $5 billion strategic investment bundled with a long-term compute partnership that SSI said would expand its computing capacity roughly tenfold over the following twelve months via NVIDIA's Vera Rubin platform. Despite that funding, the laboratory had still not publicly released a model or a research paper as of late 2026, and employs roughly 50 people.^[3][4]

Approach

SSI is organised around what Sutskever has described as an alignment-first strategy: instead of shipping incremental products and retrofitting safety work, the laboratory keeps a single team working on capability and safety in one process, without a separate commercial division. The company maintains no product roadmap and no published research, arguing that revealing intermediate results would surrender competitive advantage without improving safety. This makes SSI an extreme test case of the proposition that a frontier AI effort can be valued almost entirely on the credibility of its founders and the scale of its compute commitments.^[2][4]

Why It Matters for AI

The laboratory embodies two debates central to AI policy. First, it demonstrates that compute and talent, not revenue, are the scarce inputs in frontier AI: NVIDIA's investment converted chip supply into an equity stake, tightening the link between hardware vendors and labs. Second, its valuation — roughly $640 million of paper value per employee with zero revenue — is the most cited example of the question of whether private markets are pricing superintelligence or signalling. For safety researchers, SSI's structure is a live experiment in whether concentrating a small elite team on alignment ahead of deployment produces safer systems.^[2][4]

>See Also

• AI Safety
• AI Alignment
• Superintelligence
• Artificial General Intelligence
• Malaysia AI Safety Institute