Nvidia is throwing more than $150 million at Baseten, a San Francisco startup that runs big AI models for customers.
Baseten has raised $300 million at a $5 billion valuation, more than doubling its last mark, according to people familiar with the deal.
Venture firm IVP and CapitalG, Alphabet’s independent growth fund, led the round, with Nvidia putting in $150 million. The bet underlines the industry’s pivot from training ever-larger models to the messy business of running them at scale, or “the process of AI models generating output in response to prompts”.
Baseten, founded in 2019, sells plumbing that helps outfits such as AI code editor Cursor and note-taking platform Notion deploy and run large AI models. The company has now raised $585 million in total, including this new cash.
Baseten co-founder and chief executive Tuhin Srivastava said he aims to build the “AWS for inference.” That is a bold pitch in a market where everyone wants to be the cloud, but nobody wants the outages, the bills or the blame.
Nvidia CEO Jensen Huang paid $20 billion in December to license AI-inference tech from startup chip designer Groq, and it has built a deep partnership with OpenAI, committing up to $100 billion.
Baseten’s existing backers, Bond, Greylock and Spark Capital, joined in, marking the startup’s third funding round in the past year. Elsewhere, Fireworks AI raised $250 million in October at a $4 billion valuation, and Cerebras is planning to raise $1 billion at a $22 billion valuation following a multibillion-dollar tie-up with OpenAI.
