Gimlet Labs raises $80M
Gimlet Labs closed an $80M Series A to build a multi‑silicon inference cloud that maps workloads across NVIDIA, AMD, Intel, ARM, Cerebras and d‑Matrix — claiming 3–10x faster inference. The raise underwrites a push to become the inference fabric for frontier labs while listing NVIDIA as a partner and potential training counterparty. (techcrunch.com)(x.com)
Menlo Ventures led the new round, with participation from Eclipse, Factory, Prosperity7 and Triatomic according to Gimlet’s announcement. (gimletlabs.ai) The company lists founders Zain Asgar, Michelle Nguyen, Omid Azizi and Natalie Serrino on its team page. (gimletlabs.ai) TechCrunch notes Asgar is a Stanford adjunct professor and a founder who previously exited a company. (techcrunch.com) Gimlet says it emerged from stealth five months ago and reported eight‑figure revenues while tripling its customer base and adding “one of the top three frontier labs” and “one of the top three hyperscalers” as customers. (gimletlabs.ai) Gimlet announced a partnership with d‑Matrix in March to deploy d‑Matrix Corsair accelerators alongside GPUs in Gimlet Cloud, with the joint PR claiming up to 10× gains in latency and throughput per watt versus GPU‑only deployments. (prnewswire.com) In a company benchmark post, Gimlet reported running gpt‑oss‑120B with a 1.6B‑parameter speculative decoder on d‑Matrix Corsair and found 2–5× end‑to‑end request speedups on interactivity‑optimized configs and up to 10× on energy‑optimized configs. (gimletlabs.ai) A separate Gimlet analysis showed that splitting LLM prefill and decode stages (using NVIDIA B200s and Intel Gaudi 3 in their model) can yield roughly a 1.7× TCO improvement versus single‑vendor stacks. (gimletlabs.ai) Gimlet’s filings and press coverage state the latest round brings the company’s total capital raised to about $92 million to date. (finance.yahoo.com)
