Nvidia Invests in Ilya Sutskever's SSI, Computing Power to Scale by an Order of Magnitude
Miles Bennett
Nvidia announced a major investment in Safe Superintelligence (SSI), co-founded by Ilya Sutskever, alongside a long-term partnership that will boost SSI's access to flagship Nvidia GPUs by an order of magnitude — a move that locks another marquee AI lab into Nvidia's chip ecosystem.
What does this deal actually buy?
The financial terms were not disclosed. The core commitment: SSI gets access to large quantities of Nvidia's flagship GPUs, scaling its compute resources up by "an order of magnitude."
SSI previously ran most of its training on Google's TPUs — tensor processing units, Google's in-house AI training chips. This means → the deal shifts SSI's compute base from the Google camp squarely into Nvidia's.
Per the Wall Street Journal, Nvidia received a rare inside look at SSI's research progress before writing the check. In plain terms = Nvidia saw SSI's cards first, judged the hand worth backing, then invested.
Who is SSI?
SSI was co-founded in 2024 by Ilya Sutskever, positioned as a single-mission research lab focused on "safe superintelligence."
Sutskever is a defining figure in AI: he studied under Nobel laureate Geoffrey Hinton, co-authored the 2012 computer-vision paper widely seen as the starting gun of the modern AI race, then joined OpenAI and led the work behind ChatGPT.
The company has raised roughly $2 billion from Andreessen Horowitz, Sequoia Capital, and other top-tier VCs, reaching a valuation of about $30 billion last year.
Sutskever says the research "focuses on overlooked aspects of how the human brain works" — but SSI has yet to publish any externally verifiable results. This reflects an investment that is betting on the person, not on a product.
Why is Nvidia doing this?
This is not a one-off. In March, Nvidia made a similar investment in Thinking Machines Lab, co-founded by former OpenAI CTO Mira Murati. That lab shipped its first AI model trained on Nvidia hardware this month.
This means → Nvidia is running a repeatable playbook: invest + long-term compute agreement, locking the most influential AI labs into deep dependency on its GPUs.
In plain terms = rather than wait for customers to pick a chip, Nvidia takes a stake first and supplies the hardware second — investor and supplier at once, binding the relationship from both ends.
Where is the risk?
SSI's core asset is Sutskever's reputation and his team, not a verified product or revenue stream. Whether the company can convert that reputation into demonstrable research results is the key milestone to watch.
Sutskever left OpenAI after a falling-out with CEO Sam Altman. This reflects the governance risk inherent in top-tier AI talent moves.
For Nvidia, even if SSI's research ultimately falls short, locking in a major customer on its GPUs long-term while pulling compute share away from Google's TPU camp means the strategic math on this deal already works.
Content is for reference only, not financial advice.