Gimlet Labs raises $300M Series B at $3B valuation to rebuild AI inference
What's the deal? Gimlet Labs has raised a $300 million Series B led by Andreessen Horowitz, valuing the AI inference startup at $3 billion. The round drew 17 additional backers, including Arm, Samsung VenturesDealroom has a profile for this one. Try Dealroom →, Tiger Global ManagementDealroom has a profile for this one. Try Dealroom →, Menlo Ventures, and Sapphire Ventures.
What's the endgame? Gimlet is building what it calls the first "multisilicon cloud," designed from the ground up for AI inference rather than training. Its approach — heterogeneous disaggregation across differentiated chip architectures — claims 3-10X faster performance for frontier workloads.
Why now? Inference has overtaken training as software's dominant workload, and demand is surging. Monthly token generation has risen sixfold in 12 months, with some projections pointing to another 20X increase by 2030.
By the numbers: Since its Series A just over five months ago, Gimlet says it has added "billions in contracted revenue" and gigawatts of datacenter pipeline, and is scaling to hundreds of megawatts in managed capacity.
The context: The startup argues that existing infrastructure — often repurposed from training or crypto mining — is orders of magnitude less efficient than it should be. Power is the tightening constraint: AI data centers consumed roughly 18GW in 2025, a figure expected to triple by 2030.
The signal: The raise ranks near the top of Series B deals in its sector, above the 99th percentile across 3,247 comparable rounds. With industry investment nearing $1 trillion a year and a cumulative $7 trillion projected by 2030, investors are betting that squeezing more throughput per kilowatt becomes the decisive edge as the AI buildout hits physical limits.
Read more: gimletlabs.ai
Image credit: Gimlet Labs
