Gimlet Labs, a startup founded by Stanford adjunct professor Zain Asgar, has successfully secured $80 million in a Series A funding round led by Menlo Ventures. The company aims to address the AI inference bottleneck by introducing what it claims to be the first “multi-silicon inference cloud,” enabling AI workloads to operate across various hardware types, including CPUs and AI-optimized GPUs.
Asgar, along with co-founders Michelle Nguyen, Omid Azizi, and Natalie Serrino, developed orchestration software that enhances the efficiency of AI applications. This innovation allows for the concurrent use of diverse hardware resources, potentially increasing AI inference speeds by 3x to 10x while maintaining cost and power efficiency. The startup has also partnered with prominent chip manufacturers such as NVIDIA, AMD, and Intel.
Asgar noted that existing hardware is underutilized, functioning at only 15 to 30 percent capacity, which translates to wasted resources costing hundreds of billions of dollars. He emphasized the importance of improving the efficiency of AI workloads to capitalize on the potential of the hardware available.