Luminal, a company advancing speed-of-light inference, announced the closing of a $5.3M seed round led by Felicis Ventures, with participation from angels including Paul Graham and Guillermo Rauch.
Luminal addresses a growing challenge in AI computing: as increasingly powerful chips are released, software lags behind, leaving significant hardware underutilized. Even top-tier accelerators like Nvidia’s Hopper required years to reach software maturity, and complexity continues to grow, putting peak performance out of reach for developers.
Luminal has developed a tightly integrated high-performance compiler and inference cloud designed to overcome this software bottleneck. Large-scale kernel search enables maximum utilization across GPUs, ASICs, and other accelerators, providing AI companies with a simple deployment interface and eliminating the need to manage intricate CUDA instructions or complex inference infrastructure.
From inception, Luminal has maintained an open-source approach, building the core compiler with community collaboration. Open development allows engineers to deploy and optimize on their own hardware while contributing to the broader ecosystem.
Luminal continues partnerships with companies running custom models to optimize latency and throughput, aiming to make high-performance AI inference accessible, efficient, and scalable for developers worldwide.