AI Inference Software Startup Gimlet Labs Raises $80M In Series A Funding

Gimlet Labs has developed cloud-based technology that the company says can run agentic AI workloads faster and more efficiently than with current generation data centers.

Gimlet Labs, a startup developer of AI inference software, has raised $80 million in a Series A funding round led by Menlo Ventures, the company said Tuesday.

The new round brings Gimlet Labs’ total funding to $92 million, including an earlier seed funding round led by Factory. Factory, along with Eclipse, Prosperity7 and Triatomic, also participated in the latest funding round.

Gimlet Labs just emerged from stealth five months ago. The San Francisco-based company said it plans to use the additional funding to expand employee hiring and more rapidly scale out its agent inference cloud offering.

Gimlet Labs develops what it calls a “multi-silicon inference cloud” based on its proprietary software stack that automatically maps agentic workloads to the most suitable processors. The technology can even “slice and execute” a single model across different system architectures, using the most optimal chip for each portion of the model, according to the funding announcement.

“We’ve entered a fundamentally new era of computing where the speed of intelligence has become the critical bottleneck. In order to unlock the next 10-100X performance increases needed in use cases like coding agents, we’ve identified how to leverage heterogeneous hardware for faster, more efficient inference,” said Zain Asgar, Gimlet Labs co-founder and CEO, in a statement.

“At Gimlet, we’re seeing this approach deliver an order-of-magnitude better performance per watt for our customers, which is critical for anyone operating at scale given today’s datacenter capacity bottlenecks,” he said.

With the funding announcement, Gimlet Labs also disclosed that it now has “eight-figure revenues,” has tripled its customer base since exiting stealth, and now counts one of the top three frontier labs and one of the top three hyperscalers as customers. The company is also working with leading AI chip companies including AMD, ARM, Cerebras, d-Matrix, Intel and Nvidia.

The startup runs its software on “multi-silicon data centers” that the company manages due to their unique heterogeneous systems designs. Customers can also deploy Gimlet’s software to their own data centers, according to the company.

The company takes the position that the surge of agentic workloads has exposed a “critical flaw” in today’s AI infrastructure in that homogeneous hardware has reached its limits in latency and power efficiency—despite spending forecasts of $650 billion in AI datacenter capital expenditures this year.

Gimlet Labs said its technology delivers AI inference processing speeds that are three to 10 times faster than with traditional data centers, for the same cost and power utilization, for very large frontier models.

The technology is based on the company’s research that “combines theory and research” to improve AI efficiency through such techniques as automated GPU kernal generation, workload orchestration and heterogeneous execution across diverse hardware.