By Kenrick Cai and Krystal Hu
SAN FRANCISCO (Reuters) – Alphabet and Nvidia have joined prominent venture capital investors to back Safe Superintelligence (SSI), a startup co-founded by OpenAI’s former chief scientist Ilya Sutskever that has quickly risen to become one of the most valuable artificial intelligence startups months after its launch, a source familiar with the matter said.
The funding illustrates renewed interest from the big tech and infrastructure providers in making strategic investments in the startups developing cutting-edge AI that requires massive amounts of computing power. Alphabet, which has its own AI models, earlier in the week announced a deal by its cloud computing arm to sell SSI access to tensor processing units (TPUs), its in-house AI chips.
SSI, which sources say was recently valued at $32 billion in a round led by Greenoaks, is one of the highest-profile startups working on AI model research, thanks to Sutskever’s stellar track record in predicting the next big thing in AI development.
Like many of its competitors, it has a huge demand for chips.
Reuters could not determine the exact terms of Alphabet’s and Nvidia’s investment in SSI. Spokespeople for all three companies declined to comment.
The twin moves by Alphabet’s corporate and cloud division with high-profile AI labs including SSI and Anthropic show the tech giant’s evolving AI hardware strategy.
Google originally reserved TPUs for in-house use. The deal to sell SSI chips in significant quantities to support its frontier AI research exemplifies the company’s ongoing strategy to expand sales to external customers, Darren Mowry, a managing director in charge of Google’s partnerships with startups, said in an interview with Reuters this week.
“With these foundational model builders, the gravity is increasing dramatically over to us,” he said.
AI developers have historically preferred Nvidia’s graphics processing units, which hold more than 80% of the AI chips market.
But SSI is so far primarily using TPUs rather than GPUs for its AI research and development, two sources said.
Google offers both Nvidia GPUs and its own TPUs through its cloud service. Its own chips are intended to excel at specific AI tasks and are more efficient than general-purpose GPUs. These chips have been used to build large-scale AI models, such as Apple and Anthropic, an OpenAI competitor that has received billions of dollars of funding from Google and Amazon.
Google and Nvidia also face a challenger in Amazon, which is building its own competing processors called Trainium and Inferentia. Amazon has said as far back as 2023 that Anthropic would develop its technology on those chips. The tech giant announced in December that Anthropic would be the first customer to use a massive supercomputer powered by hundreds of thousands of its own chips.
