Gimlet Labs announced on September 4, 2026 that it raised $300 million in a Series B round led by Andreessen Horowitz, valuing the AI infrastructure startup at $3 billion. The round also drew Sapphire Ventures, M12, Arm, Menlo Ventures and Factory, bringing Gimlet’s total funding to $392 million.
Gimlet Labs builds what it calls the industry’s first multi-silicon inference cloud for agentic AI: software that disaggregates a model’s inference workload and routes different phases of a single request to whichever chip architecture handles that phase best, rather than pinning a model to one vendor’s GPU stack. The company says this yields up to 10x gains in throughput and interactivity versus single-architecture deployment, and it is scaling to hundreds of megawatts of managed heterogeneous infrastructure. By March 2026 it had tripled its customer base, adding one of the top three frontier labs and one of the top three hyperscalers, and it has since signed billions of dollars in contracted revenue for its Gimlet Cloud product.
The new capital funds build-out of that multi-silicon cloud and continued hiring. The investor list matters as much as the amount: Arm and M12 (Microsoft’s venture arm) both took positions, signaling that chip designers and hyperscaler-adjacent capital see value in a neutral routing layer that sits above Nvidia, AMD, and custom silicon rather than competing with any one of them directly.
A $3 billion valuation on a company whose product is a routing layer is a data point for how tight AI inference capacity still is in September 2026: investors are paying up for software that makes existing GPU and accelerator fleets go further, not just for new chips. It does not prove the multi-silicon routing approach beats simpler single-vendor scaling at hyperscaler volumes, and the “billions in contracted revenue” figure is Gimlet’s own characterization, not an audited number.