Gimlet raises for multi‑silicon inference
Gimlet Labs raised a large round to run inference across NVIDIA, AMD, Intel, ARM, Cerebras and d‑Matrix — investors are backing multi‑silicon inference orchestration. That funding push is turning heterogeneous inference from niche to institutional. (techcrunch.com) (menlovc.com)
Gimlet announced an $80 million Series A led by Menlo Ventures, with participation from Eclipse, Factory, Prosperity7 and Triatomic. (gimletlabs.ai) The round brings Gimlet’s total disclosed funding to roughly $92 million after an earlier $12 million seed led by Factory that the company raised when it emerged from stealth in October 2025. (aidirectory.com) Gimlet describes its product as the industry’s first “multi‑silicon inference cloud” that can decompose an agentic workload and route or even slice parts of a single model across different architectures for execution. (techcrunch.com) The company says its stack delivers 3–10× speedups or efficiency gains versus GPU‑only deployments on certain agentic and large‑model inference tasks, per its product claims and customer benchmarks. (itnewsonline.com) Gimlet has announced a technology partnership with d‑Matrix to deploy Corsair SRAM‑centric accelerators alongside GPUs in the Gimlet Cloud, with a joint claim of up to 10× latency and throughput‑per‑watt improvements. (prnewswire.com) The company lists active integrations or collaborations across chips from NVIDIA, AMD, Intel, ARM and Cerebras and says its customer base has tripled since launch, now including “a top frontier lab” and a hyperscaler. (markets.businessinsider.com) Menlo Ventures framed the investment around Gimlet’s approach of decoupling AI workloads from fixed hardware bindings and routing constituent stages to optimal compute to unlock order‑of‑magnitude inference gains. (menlovc.com)
