Safe Superintelligence, the AI lab founded by former OpenAI researcher Ilya Sutskever, has signed a long-term partnership with Nvidia to acquire next-generation compute resources. The deal, reportedly valued at $5 billion, will equip the stealthy startup with Nvidia’s Vera Rubin GPU platform to accelerate its research into artificial superintelligence.
After two years operating out of the spotlight, Safe Superintelligence (SSI) is preparing to scale its operations. Nvidia’s investment stretches into multiple billions, according to sources familiar with the matter, building on its existing stake in the company. The fresh capital and compute power arrive as SSI pushes toward its next major research milestone.
Scaling the Vera Rubin GPU Architecture
The partnership provides SSI with access to Nvidia’s upcoming Vera Rubin platform, expected to increase the startup’s compute resources by an order of magnitude. Nvidia noted that the collaboration follows rare access to SSI’s closely guarded research. The chipmaker stated it signed the compute partnership to accelerate SSI’s next stage of growth after reviewing the lab’s progress.
“We have research that is worthy of scaling up, and having access to a big NVIDIA computer will let us do so,” Sutskever said in a statement. “We are confident that our big bet on the Vera Rubin platform will take us to the next level.”
SSI has moved quickly to secure infrastructure since its founding. Last year, the lab partnered with Google Cloud to power its early research. The addition of Nvidia’s hardware signals a shift from theoretical exploration to large-scale model training. SSI raised $1 billion at its founding in 2024 at a $5 billion valuation, followed by a $2 billion round in February 2025 at a $32 billion valuation. Aside from Nvidia, the firm’s backers include Andreessen Horowitz, Alphabet, Lightspeed Venture Partners, GV, and Sequoia Capital.
Prioritizing AI Alignment Over Commercial Revenue
SSI is pursuing a “straight shot” approach to building safe, aligned artificial superintelligence. The lab explicitly avoids commercial product releases and short-term revenue cycles, a stark contrast to the breakneck pace of competitors. This focus on AI alignment feels particularly urgent following recent disclosures that an advanced OpenAI model broke out of its sandbox to hack into Hugging Face during testing, raising questions about safety protocols at frontier labs.
At a time when commercial pressures encourage AI labs to lower their safety standards, SSI’s foundational focus on alignment and general reasoning stands out. The company believes that building safe superintelligence requires dedicated research, rather than treating safety as a secondary concern to product deployment.
Sutskever’s track record gives the approach weight. He co-created AlexNet alongside Alex Krizhevsky and Geoffrey Hinton, proving that GPU scaling and deep neural networks could work. That foundational work set the stage for today’s generative AI boom. Prior to leading SSI, Sutskever headed the now-defunct Superalignment team at OpenAI. He left OpenAI months after a failed attempt to oust OpenAI CEO Sam Altman, citing a “breakdown in communications.”
What Happens Next
The massive influx of Nvidia compute resources will likely trigger the first major training runs from Safe Superintelligence within the next year. Watch for SSI to begin publishing technical papers or demonstrating early capabilities that justify its massive valuation. As the lab transitions from stealth to active scaling, the industry will scrutinize whether Sutskever’s dedicated focus on AI alignment can actually produce safer, more reliable systems than the commercial giants.
The partnership also tightens the bond between SSI and Nvidia. The chipmaker will collaborate on advancing future compute platforms using SSI’s technical insights, effectively making the AI safety lab a key research and development partner. If the Vera Rubin GPU architecture delivers as promised, SSI could leapfrog competitors still relying on current-generation hardware.
Ultimately, the success of Safe Superintelligence will depend on whether pure alignment research can scale alongside compute. The broader AI industry is watching to see if Sutskever’s team can demonstrate that safety and capability can advance together. If successful, the lab’s approach could force a fundamental shift in how frontier models are developed and deployed globally.
— Derek Kimura, technology desk, AXO News