Gimlet Labs, a pioneering AI startup, has raised $80 million in Series A funding to address the critical AI inference bottleneck.
The company's innovative software enables AI workloads to run simultaneously across diverse hardware like CPUs, GPUs, and high-memory systems, maximizing efficiency.
Gimlet Labs' Breakthrough Technology
Gimlet Labs has developed the first multi-silicon inference cloud, orchestrating agentic workloads by slicing them based on compute, memory, or network demands.
This approach boosts AI inference speeds by 3x to 10x at the same cost and power, tackling the issue where current hardware utilization hovers at just 15-30%.
By partitioning models across optimal chips, Gimlet Labs eliminates waste from idle resources, potentially saving hundreds of billions in data center spending.
Founders' Impressive History
Founded by Stanford adjunct professor Zain Asgar and cofounders Michelle Nguyen, Omid Azizi, and Natalie Serrino, the team previously built Pixie, an open-source Kubernetes tool acquired by New Relic shortly after its launch.
Launching publicly in October with eight-figure revenues, Gimlet Labs has already doubled its customer base, including major model makers and cloud providers.
Funding and Strategic Partnerships
The oversubscribed Series A was led by Menlo Ventures, with participation from Factory, Eclipse Ventures, Prosperity7, Triatomic, and notable angels like Intel CEO Lip-Bu Tan.
Partnerships with chip giants including NVIDIA, AMD, Intel, ARM, Cerebras, and d-Matrix underscore the technology's broad hardware compatibility.
Future Impact on AI Landscape
As data center investments are projected to hit $7 trillion by 2030, Gimlet Labs' solution offers a smarter path by optimizing existing fleets rather than endless scaling.
Founder Zain Asgar envisions 10x more efficient AI workloads, transforming how labs and data centers deploy inference at scale.
Andrew Lee
