POLYN Technology, a specialist in ultra-low-power neuromorphic computing, has announced the successful manufacturing and testing of the world’s first silicon-implemented Neuromorphic Analog Signal Processing (NASP) chip. The development represents a significant step towards low-energy AI processing at the Edge.
Unlike conventional digital neural processors, POLYN’s NASP platform performs AI inference in the analog domain, drastically reducing power consumption. The company said its chips process sensor signals in microseconds while consuming only microwatts of energy, eliminating the overhead associated with digital operations.
POLYN plans to showcase the first NASP chips at CES 2026 in Las Vegas, and has also displayed a limited selection at CES Unveiled Europe in Amsterdam this week.
The debut chip integrates a voice activity detection (VAD) core, marking the first stage of POLYN’s efforts to deliver advanced voice-processing capabilities for low-power applications. The company intends to follow with additional cores for speaker recognition and voice extraction, targeting devices such as smart home appliances, critical communications headsets, and other voice-controlled products.
“The successful silicon implementation proves our technology works as intended,” said Aleksandr Timofeev, Chief Executive Officer and Founder of POLYN Technology. “We have, for the first time, generated an asynchronous, fully analog neural-network core directly from a digital model. This opens the door to neural computation with digital-class accuracy, but at microwatt-level energy consumption.”
The NASP VAD chip offers performance that the company described as “breakthrough”: continuous operation at around 34 µW, inference latency of 50 microseconds, and fully asynchronous operation with no clock or digital conversions required.
POLYN’s design tools automatically convert trained digital neural network models into analog neuromorphic cores, compatible with standard CMOS processes between 40–90 nm. The company said this approach allows semiconductor developers and AI innovators to implement neural networks in analog silicon quickly and process-agnostically.
“The launch of our first NASP chip transforms the technology from concept to production-ready reality,” Timofeev added. “It demonstrates that analog neuromorphic computation can coexist with today’s digital flows, opening opportunities for chipmakers, OEMs, and AI developers.”
POLYN is now preparing evaluation kits for early adopters and expanding its NASP product families for automotive, wearable, and critical communication applications.
