femtoAI

femtoAI develops ultra-efficient AI inference technology and custom sparse processing hardware designed to run AI models directly on edge devices, without cloud connectivity. Founded in 2018 and spun out of Stanford University's Brains in Silicon Lab, the company is headquartered in Silicon Valley and raised $20.6M in total funding. Its work spans edge AI inference, neuromorphic-inspired computing, sparse processing hardware, custom AI accelerator design, model compression and optimization, software development kits and full-stack AI systems.

The company's Sparse Processing Unit (SPU-001) is a custom sparse processing AI chip aimed at battery-powered and cost-sensitive edge devices; femtoAI reports 100x energy efficiency and 10x model capacity relative to traditional AI processors. Its full-stack edge AI solution combines ready-to-deploy AI models, a software development kit and the custom hardware, reducing power consumption by 100x and memory footprint by 10x with minimal performance impact, delivering real-time AI on compact silicon.

The technology has shipped in real-world products, including the NewSound AI over-the-counter hearing aid. Target verticals include consumer electronics, smart home devices, wearables, hearing health, robotics and autonomous vehicles. femtoAI, rebranded from Femtosense in 2024, won Best-in-Show at CES 2026 and the Startup Terrace Award at InnoVEX 2025 in Taiwan, and is a member of the NVIDIA Inception Program.