By Kenrick Cai, Krystal Hu and Anna Tong
(Reuters) – Safe Superintelligence, an artificial intelligence startup co-founded by OpenAI’s former chief scientist Ilya Sutskever last year, is in talks to raise funding at a valuation of at least $20 billion, four sources told Reuters.
That would quadruple the company’s $5 billion valuation from its last funding round in September, when it raised $1 billion from five investors including Sequoia Capital, Andreessen Horowitz, and DST Global.
SSI’s fundraising tests the ability of high-profile AI ventures to continue to command premium valuations following an industry-wide reappraisal prompted by Chinese startup DeepSeek’s unveiling of its low-cost AI last month.
SSI, which has not generated any revenue, has said its mission is to develop “safe superintelligence” that is smarter than humans while aligned with human interests.
The company’s conversations with existing and new investors are still in the early stages and terms could still change, the sources said this week, who requested anonymity to discuss private matters. It was not clear how much money SSI was seeking to raise.
SSI, which was founded in June with offices in Palo Alto and Tel Aviv, did not respond to requests for comment. Sutskever’s co-founders are Daniel Gross, who previously led AI initiatives at Apple, and Daniel Levy, a former OpenAI researcher.
SECRETIVE STARTUP
Beyond the cursory explanation of the company’s goals for safe AI, not much is known about the secretive startup or its work. What has fueled interest among investors is Sutskever’s reputation and the novel approach he has said his team is working on.
In AI circles, he is a legend for his contributions to breakthroughs that underpin the investment frenzy in generative AI. He was an early advocate of scaling, which means dedicating vast amounts of computing power and data to refining AI models.
That concept was the foundation that led to generative AI advances like OpenAI’s ChatGPT, setting the course for a wave of tens of billions of dollars in investment in chips, data centers and energy.
Sutskever was also early in seeing the potential ceiling of such an approach due to the dwindling pool of available data to train models. Recognizing the importance of putting in resources in the inference stage, or the stage of AI when a trained model draws conclusions, he founded the team that worked on what would become OpenAI’s latest series of reasoning models, setting a new research direction that has been widely followed.