Investors have handed a billion dollars to a San Francisco lab that launched barely a year ago and has yet to ship a product, the clearest sign yet that frontier model research remains the most expensive bet in venture capital.
Index and Lux co-led the Series A, which values the company at $6B, roughly four times its January mark, and brings total funding to $1.18B. Sequoia, GV and Kleiner Perkins participated, and Nvidia is also on the cap table. The lab introduced itself earlier this year with a $180M seed at $1.5B.
Its thesis is efficiency. Rather than buying ever larger clusters, the team argues that better training methods can extract more capability from each processor. That pitch resonates with investors worried that the compute arms race is capping returns.
The founders came out of a research background and have drawn scrutiny for the size of their raises relative to published output. In a market where the biggest labs spend tens of billions annually, a billion dollars is not enough to compete head-on, so the efficient-training angle is also a strategic necessity.
The company has said it intends to keep hiring researchers, with the money largely earmarked for talent and experiments rather than infrastructure.
No product timeline was disclosed alongside the announcement.
