Gimlet Labs, a San Francisco, CA-based startup that provides serverless inference for AI agents, has raised $80 million in a Series A funding round led by Menlo Ventures.
The round also saw participation from Eclipse Ventures, Prosperity7, Triatomic, and Factory.
The company has raised a total of $92 million.
The company plans to use the funds to grow its operations and development.
Gimlet Labs offers the only multi-silicon inference cloud in the industry. Its software automatically runs AI workloads on the best chips without extra work for developers. It can even split a single model across different chips for maximum efficiency. This technology makes AI models 3–10 times faster at the same cost and power, even for very large models.
Gimlet Labs provides the industry’s only multi-silicon inference cloud, running its software on managed data centers with unique heterogeneous systems. Its platform automatically maps AI workloads to the best chips, even splitting a single model across multiple architectures for maximum efficiency. This approach makes AI models 3–10 times faster at the same cost and power, including for very large models, without extra work for developers.
Read More:ArtIn Energy Receives $255M Investment from Agila Investments
The rapid growth of AI workloads has revealed a key problem: using only one type of hardware is slow and inefficient. With AI processing now reaching quadrillions of tokens per month and still growing, a one-size-fits-all approach wastes performance and resources, even as the industry plans to spend $650 billion on AI data centers this year.
“We’re in a new era of computing where speed is the main limit,” said Zain Asgar, co-founder and CEO of Gimlet Labs. “To achieve 10–100× faster performance for tasks like coding agents, we use different types of hardware for faster, more efficient AI. This approach gives our customers much better performance per watt, which is essential for operating at scale with today’s datacenter limits.”
“Heterogeneous hardware is unavoidable, and Gimlet Labs is leading the way,” said Tim Tully, partner at Menlo Ventures. “Most infrastructure was built for a single-type world, costing the industry hundreds of billions in CapEx. Gimlet built the only system designed from the start for diverse hardware, made for large-scale AI. Their research and deployment experience is unmatched.”
“From our view, the biggest AI infrastructure projects—from foundation model labs to sovereign clouds—are all realizing the same thing: single-type silicon can’t deliver the needed speed and efficiency,” said Abhishek Shukla, Managing Director at Prosperity7 Ventures US. “Gimlet’s multi-silicon system fills this gap and will be a key foundation for large-scale AI deployments.”
About Gimlet Labs
Founded by Zain Asgar, Omid Azizi, Michelle Nguyen, and Natalie Serrino, Gimlet Labs aims to dramatically improve AI performance and make much more compute available for AI workloads. Their inference cloud is built from deep research across the stack to support the next generation of scalable AI infrastructure. Gimlet combines theory and practice to boost AI efficiency using techniques like automated GPU kernel generation, smart workload orchestration, and running tasks across different types of hardware.
Read More:Sam Altman Leaves Helion Energy Board During Partnership Talks
