At the heart of the edge AI proposition is the idea that data can be collected from distributed sensors far away from data centers for greater efficiency, cost savings and privacy. That approach requires some form of low power, memory and compute to gather and somehow analyze the sensor data — probably with AI tools -- and communicate the analysis – probably wirelessly - -to other systems or human administrators.
Many companies are involved in edge AI technology. The global edge AI market is undergoing rapid growth, valued as much as $35 billion in 2025 and projected to surpass more than $196 billion by 2035, according to Fortune Business Insights and others. The key drivers are increased demand for low-latency, real-time data processing in IoT, automotive and VR applications, which all involve a fantastic number of vendors small and large. Think also: physical AI, as in robots, drones and industrial IoT, but also smart devices for healthcare and medtech.
Key players in edge AI already include Intel and NXP with CPUs and NPUs and intelligent software, but also Irida Labs, Alif Semiconductor, Silicon Labs and Nordic Semiconductor (prominent in low-power wireless and IoT), Kinara, Syntiant, Ambiq and EMASS. Look for the list to grow and include many players in AI. Many of the above-listed companies are slated to exhibit at the upcoming Embedded World 2026 in Germany March 10-12 and future events.
EMASS, a subsidiary of Nanoveu, showed off its tech at CES 2026 and after traveling to Germany is headed to Sensors Converge in Santa Clara, CA, running May 5-7, where potential competitors such as Bosch Sensortec will also exhibit. EMASS is showcasing its ECS-DoT system-on-chip that the company describes as an on-device AI platform for always-on intelligence with ultra-low latency and heavily reduced power consumption. The SoC is based on 22nm silicon, now sampling, with 16nm silicon in the works.
In an interview, executives from EMASS told Fierce what makes the ECS-DoT competitive. One big factor is EMASS’ use of RISC-V architecture, according to analysts.
“We don’t need to have the cloud and can store all that data on-chip,” said CEO Mark Goranson. “It’s perfect for burglar alarms with no access to the cloud. There’s no problem getting a market. The market is huge.”
He predicted the ECS-DoT could also “revolutionize” the drone market with its power efficiency. The 16nm next-gen chip is “perfect for glasses and VR and object detection,” Goranson added.
Sensor data is processed on-device with ECS-DoT (an abbreviation for Edge Computer Sensor-Distributed Operational Tech) and transmitted wirelessly via Semtech LoRaWAN, making it useful for industrial predictive maintenance monitoring without the cloud. Semtech LoRaWAN also enables audio security detection, with on-device that listens for breaking glass or intrusions, even when the environment is noisy. It operates at sub-milliwatt power for use with security devices needing extended battery life.
EMASS also will demonstrate bone conduction and voice detection with a hearable that uses an inertial measurement unit for bone-conduction audio sensing. ECS-DoT is designed to detect voice activity with keyword spotting done by analyzing jaw movement through the IMU. That approach limits the need for always-on mics. In its demos, EMASS is also showing a vision-based smart meter retrofit where the ECS-DoT processes data from a camera focused on analog water and gas meters to interpret the movement of dials. If there’s a rapid increase in consumption, it can be detected locally, then transmitted wirelessly—a cost-effective smart metering approach without the need to replace water and gas meters.
EMASS, founded in 2020, employs about 30 engineers and other professionals with headquarters in Singapore, with technical R&D centers in Singapore and Cairo, Egypt. Its commercial business development is US-based in Pasadena, CA. Goranson traces his roots back to Intel with later executive roles at Freescale, ON Semi and TE Connectivity. Mohamed Sabry, the founder, is CTO and an associate professor at NTU Singapore.
According to specs shared with Fierce, the 22nm ECS-DoT relies on a Risc-V control core operates at 50 MHz with dual deep learning accelerators. On-chip memory is up to 4 MB L2 RAM, so no external DRAM is required. AI inference work is supported by 1 to 5 mW active power. The die is 7 mm2. ECS-DoT won Best in Show in embedded computer design at CES.
EMASS said its AI performance per watt is 3 TOPs on average with just 0.1 mW average power, below the power of industry peers Syntiant, Himax, Ambiq and Maxim.
For use in drones, EMASS predicts the ECS-DoT an increase in flight times of 60% without any hardware upgrades, going from 60 minutes to 96 minutes. AI logi allows minimal use of battery power, which can be preserved for propulsion. The 16nm version of ECS-DoT will be enhanced with fully integrated Low Energy Bluetooth and more memory.
Analysts took notice of the ECS-DoT SoC in comments shared with Fierce. "The ECS-DoT SoC is pushing the boundaries of ultra-low-power edge AI enablement for far edge applications," said Dan Mandell at VDC Research. "It is also a shining example of the flexibility of the RISC-V architecture for tailored acceleration to support emerging neural networks."
Mike Demler, an independent analyst formerly with Linley Group, said that ECS-DoT integrates RISC-V CPUs with custom deep-learning accelerators to produce an ultra-low-power AI processor that "sets new records for power efficiency."
Editor's Note: At Sensors Converge, May 5-7, EMASS will exhibit in the Edge AI Foundation Pavilion alongside Syntiant, Innatera, BrainChip, Ceva, EmbedUR Systems and Himax Technologies.