The one-chip era of AI is ending, and Gimlet Labs is riding the split. The startup’s software spreads model workloads across many kinds of processors, and investors have now banked it with $300M that pegs the company at $3B. Andreessen Horowitz led the round, with Arm Holdings and M12, Microsoft’s venture arm, coming in as fresh backers.
The money stacks on a $12M seed that drew Intel CEO Lip-Bu Tan, Figma CEO Dylan Field and a16z partner Raghu Raghuram as angels, and an $80M Series A led by Menlo Ventures with Eclipse, Factory, Prosperity7 and Triatomic. Founder and CEO Zain Asgar and his co-founders previously built Pixie, the Kubernetes observability firm New Relic bought in 2020.
Gimlet’s pitch is simple: an AI task does not have to ride one silicon type from start to finish. Its software carves work into pieces and sends each to the hardware that fits, GPUs for some stages, CPUs or specialized accelerators for others. The company calls the approach multi-silicon inference.
With data center builders struggling to scale on a single chip line, Gimlet aims to become the layer that lets operators mix suppliers freely. The round funds its push into data center deployments as demand for heterogeneous compute climbs.
