Low-Power Sparse AI Processing for Enhanced ISR Communications with Femtosense’s SPU-001 Chip

Femtosense is an innovative tech startup deploying sparse AI through its breakthrough Sparse Processing Unit (SPU-001), a TRL level 7 chip designed for cost-effective and efficient real-time edge computing. This technology is poised to disrupt the $50B AI hardware market by providing significant advantages over traditional AI techniques, enabling Intelligence, Surveillance, and Reconnaissance (ISR) operations with minimal power consumption. The SPU-001 uses less than 1mW of power to filter background noise from speech using neural networks, significantly extending battery life for communication headsets—essential for soldiers who face burdensome power demands in the field. Femtosense’s SPU-001 surpasses existing audio filtration technologies, which require upwards of 100mW and multiple beamforming microphones, while using 10 times less memory and half the physical space. Capable of removing 10 to 18dB of background noise in complex environments, the SPU-001 outperforms leading commercial solutions like AirPods Pro 2, which only achieve an 8dB reduction. Its algorithm works in all directions and reverberant conditions, reducing both stationary and impulsive noises, offering critical benefits for defense applications and other high-value use cases.