While most of the AI world is focused on building bigger models, Gimlet Labs is tackling a different issue: inefficiency. The startup just raised $80M to solve the “inference bottleneck” by allowing AI workloads to run across multiple types of hardware at once — CPUs, GPUs, and high-memory systems.

The idea is simple but powerful: instead of relying on one chip to do everything, split the workload and use the best hardware for each task. With data centers currently underutilized and billions wasted on idle compute, this approach could unlock massive efficiency gains — potentially making AI faster and far more cost-effective.