Lamb Labs
ActiveAI-nativeLightning fast chips with hardcoded AI models
What it does
Lamb Labs is building MPUs (Model Processing Units), custom chips that hardcode LLM weights directly into the chip. Unlike GPUs, which move weights between memory and compute, MPUs keep the model on-chip, targeting up to 20,000+ tok/s and a 63× higher intelligence per watt.
Its site says now · captured 2026-08-23
Lamb Labs, the World's First MPU
Lamb Labs is building MPUs (Model Processing Units). We hardcode the entire model, including its weights, into silicon. The model is the chip. This solves the memory-bandwidth bottleneck GPUs face, targeting 20,000+ tokens per second and 63× higher intelligence per watt.
Next to it in AI infra and compute · AI inference acceleration hardware
- BaudY Combinator S26
AI chips for ultra-fast model training and inference
- deepsiliconY Combinator S24
Software and hardware to run neural networks faster and cheaper
- Texel.aiY Combinator W23
Run AI pipelines 10x faster
- X-SiliconPlug and Play PnP 2024
X-Silicon is developing a low-power unified compute-graphics engine for the next generation of embedded visual and compute platforms
- Dipole LabsY Combinator S26
AI-controlled optical switching for AI clusters
- OpenRelayY Combinator S26
Distributed, hardware-agnostic AI inference
- TracerY Combinator S26
Combining open-source AI models for better answers at lower cost
- Understudy LabsY Combinator S26
Self-optimizing neocloud that cuts LLM bills by 80%