Safe Superintelligence: Sutskever's Secret Bet & Nvidia's $5B Compute Deal

Overview

On July 27, 2026, Nvidia announced a long-term partnership with Safe Superintelligence (SSI), Ilya Sutskever's secretive AI lab, reportedly investing about $5 billion and granting access to its next-generation Vera Rubin platform — enough to increase SSI's compute by roughly 10 times. SSI has no product and keeps its research direction secret; the bet rests on Sutskever's thesis that the 'age of scaling' is over and that SSI has found a new idea 'worthy of scaling up.'

In this video

• On July 27, 2026, Safe Superintelligence and Nvidia announced a long-term strategic partnership to expand SSI's computing power.
• Nvidia's investment in SSI is reported at around $5 billion — one of the chipmaker's largest deals of the AI boom — though the official announcement did not state a figure.
• SSI gains access to Nvidia's next-generation Vera Rubin platform, increasing its compute by an order of magnitude — roughly 10 times what it had.
• Vera Rubin systems begin reaching partners in the second half of 2026, making SSI an early recipient of frontier hardware.
• Safe Superintelligence was founded in June 2024 by Ilya Sutskever and calls itself the world's first 'straight-shot' lab, with one goal and one product: a safe superintelligence.
• SSI has no commercial product and no revenue, yet is valued at about $32 billion and has raised on the order of $7 billion.
• SSI's specific technology is deliberately secret — it publishes no papers and ships no demos — so its value rests on Sutskever's reputation and an unpublished research direction.
• Sutskever argues the 'age of scaling' of 2020 to 2025 is over, that adding more compute and data has hit a wall, and that the real bottleneck is now ideas, not compute.
• The apparent paradox of a scaling skeptic taking $5 billion of compute resolves in Sutskever's words: SSI has 'research that is worthy of scaling up' and wants a big Nvidia computer to do so.
• The two companies will co-develop Nvidia's current and future compute platforms, and Jensen Huang said he is excited to see what new breakthroughs SSI discovers on the Vera Rubin platform.

Frequently asked questions

How much is Nvidia investing in Safe Superintelligence?

Nvidia's investment is reported at about $5 billion (per Bloomberg), one of its largest deals of the AI boom. The official July 27, 2026 announcement confirmed a major investment but did not state an exact figure.

What is Safe Superintelligence's secret technology?

It is genuinely secret. SSI publishes no papers, ships no demos, and shows no product, so its specific technical approach is unknown outside the lab. What is public is Ilya Sutskever's thesis that pure scaling has stalled and that SSI has a new idea 'worthy of scaling up.'

Why is Nvidia giving SSI more compute if Sutskever says scaling is over?

Sutskever says the bottleneck is ideas, not brute-force scaling — but he also says SSI has 'research that is worthy of scaling up.' Nvidia's roughly 10x compute via Vera Rubin is meant to test that specific new idea at scale, not simply to run the old scaling recipe.

What is the Vera Rubin platform and how much compute does SSI get?

Vera Rubin is Nvidia's next-generation compute platform, reaching partners in the second half of 2026. Access to it, together with the investment, increases SSI's compute by an order of magnitude — about 10 times what it had.

What is Safe Superintelligence and who runs it?

SSI is an AI lab founded in June 2024 by Ilya Sutskever, formerly OpenAI's chief scientist and a creator of AlexNet and the sequence-to-sequence method. It calls itself the world's first 'straight-shot' lab, pursuing one product — a safe superintelligence — with no commercial releases along the way, and is valued at about $32 billion.