Gimlet Labs raises $300M Series B at a $3B valuation, six months after its Series A
Andreessen Horowitz led the round for the multi-silicon inference-cloud startup, which says it has added billions in contracted revenue and gigawatts of datacenter pipeline since closing an $80M Series A in March.
Gimlet Labs raised $300 million in a Series B led by Andreessen Horowitz, valuing the San Francisco-based AI infrastructure startup at $3 billion -- a 10x jump from where its $80 million Series A left it just six months ago. New investors Arm, M12 and Samsung Ventures joined the round alongside Sapphire Ventures, Menlo Ventures, Tiger Global and 14 other backers. The company's total raised is now $392 million.
What it's building
Gimlet describes itself as the first multisilicon cloud built from the ground up for inference -- running AI workloads across a mix of GPUs, near-memory compute, dataflow chips and CPUs rather than treating accelerators as interchangeable. The pitch is efficiency, not just capacity: the company claims 5-10x speedups for the same power footprint versus homogeneous infrastructure. CEO and co-founder Zain Asgar tied the raise to where he sees the workload curve heading, saying "inference is the dominant AI workload and the demand for tokens is explosive."
The numbers behind the valuation jump
Gimlet says monthly token generation on its platform grew 6x over the past 12 months, and that it has signed billions of dollars in contracted revenue plus gigawatts of datacenter pipeline since its Series A closed in March. That combination -- inference demand outpacing training spend, and infrastructure startups pricing in years of future contracts today -- is the same dynamic behind Crusoe's much larger raise this month: capital is chasing the compute layer under AI, not just the model layer on top of it. For a company measuring its own AI spend, that's a reminder the infrastructure bill downstream of every inference call is itself a fast-moving target, not a fixed unit cost.