Gimlet Labs Raises $80M to Build AI Inference Cloud That Runs Workloads Across Multiple Chip Types at Once
Summary
Gimlet Labs raises $80M in a Series A led by Menlo Ventures to power a multi-silicon AI inference cloud that simultaneously runs workloads across CPUs, GPUs, and high-memory systems, claiming 3x to 10x speed improvements while tackling the hundreds of billions wasted on idle hardware.
Key Points
- Gimlet Labs raises an $80 million Series A led by Menlo Ventures to deploy a 'multi-silicon inference cloud' that splits AI workloads across CPUs, GPUs, and high-memory systems simultaneously.
- The startup claims its orchestration software speeds AI inference by 3x to 10x at the same cost and power, addressing the fact that existing hardware sits idle 15-30% of the time, wasting hundreds of billions of dollars.
- Gimlet Labs, which publicly launched in October with at least $10 million in revenue, has already partnered with NVIDIA, AMD, Intel, ARM, Cerebras, and d-Matrix, and counts a major model maker and a large cloud company among its rapidly growing customer base.