BM110 Analog AI Chip

Blumind describes itself as a Canadian deep-tech Physical AI startup building analog AI processors on standard CMOS. It targets always-on sensing in devices where battery life limits conventional digital AI.

Its BM110 is an always-on analog AI chip for audio that detects voice, keywords and commands so the main system can sleep. A software and ML platform trains models in standard frameworks and deploys them to Blumind silicon. A next-generation analog compute platform for vision and sensor fusion has a stated target of 250 TOPS/W.

• BM110: always-on analog AI chip for voice, keyword and command detection.
• AMPL technology: all-analog neural signal processor architecture on standard CMOS.
• Model-to-silicon toolchain: trains in standard ML frameworks and deploys to Blumind chips.
• Next-gen analog compute: platform for vision, sensor fusion and physical AI.
• Target markets: wearables, connected home, industrial, security and smart mobility.

Wearable and earbud makers can add voice wake-up with BM110. Smart home and security device assemblers can use always-on audio detection for event triggering.

Always-on voice and keyword detection lets the main processor stay asleep, extending battery life. Models are trained in standard ML frameworks and deployed through a model-to-silicon toolchain. The all-analog AMPL architecture is built on standard CMOS, avoiding exotic processes.

Wearables and smart sensors need AI that runs on microwatt to milliwatt budgets. Vietnamese consumer-electronics and IoT device makers exporting to global brands can evaluate analog AI to add voice interfaces without larger batteries.

Described on its site as a Canadian deep-tech Physical AI startup. Won the Siemens and AWS Industry Disruptor Pitch Challenge at CES 2026 per the site headline. Founding year not stated.