Blumind Launches BM110 Evaluation Kit, Putting Ultra-Low-Power Analog AI Compute on the Customer’s Bench

Evaluation platform, including board, SDK, and reference models, lets developers evaluate Blumind’s AMPL™ analog inference benefits for themselves.

OTTAWA, ONTARIO, CANADA, November 9, 2026 /EINPresswire.com/ -- Blumind today announced availability of the BM110 Evaluation Kit, a complete hardware and software platform for developing always-on audio and sensor inference on the company’s AMPL™ analog compute technology. The kit ships with Blumind’s BM110 silicon, configured as a reference always-on audio board, along with a software development kit, pre-loaded reference models for keyword spotting and voice activity detection, and current sense infrastructure for direct power measurement, everything a design team needs to validate power and performance benefits of the BM110 on its own bench. Blumind will be showcasing the kit at electronica 2026 in Munich, November 10–13, at Hall B5-560.

Always-on audio sensing is required for ubiquitous Physical AI applications enabled by continuous inference. Today’s digital architecture suffers from high-speed clocking and sequential execution that diminish battery life. In smart glasses, smartwatches, rings and other wearable and mobile applications there is no power headroom to give; efficient, intelligent analog processing is the only path to all-day, always-on products.
Blumind computes in the analog domain, storing neural network weights directly in the BM110 fabric instead of shuttling them between memory and a digital MAC array. Cutting that data movement, the dominant energy cost in digital inference, lets the BM110 neural network compute core run continuously at sub-3 microwatts with a total always-on power including the microphone of ~50 microwatts. The evaluation kit is built so that customers can confirm power numbers directly during active inference.

What Ships in the Kit
• Always-on audio AI inference evaluation board carrying BM110 silicon, with analog microphone and integrated signal processing front end.
• Battery management and series sense resistor, so always-on system current is measured on board directly in real time.
• Software development kit, with a hardware abstraction layer that isolates application code from the host platform and supports integration onto any MCU.
• Model import flow for networks trained in standard cloud and server frameworks.
• Pre-loaded reference models for voice activity detection and keyword spotting.
• BM110 datasheet and on request the AMPL™ Technology Validation Report, covering weight storage and programming accuracy across process, voltage, and temperature corners; characterized and modeled for weight drift over product lifetime; and end-to-end system power for representative always-on workloads.

Designed to Be Measured, Not just Demonstrated
The BM110 kit is built to be interrogated. Questions around accuracy, real-world power draw, drift and operation over temperature are all answerable on the board, with real-life models working in real time.

Availability
The BM110 Evaluation Kit is available now to qualified customers. Kits are supplied with the technology validation report, datasheet, and direct applications engineering support through model bring-up. Qualified customers can request a kit and engagement in Blumind’s user support program.

About Blumind
Blumind designs ultra-low-power analog AI silicon for Physical AI. Its AMPL™ technology stores weights and computes in the core neural network fabric, eliminating the memory-to-compute data movement that dominates energy use in digital inference. Blumind targets Physical AI applications where continuous intelligence must fit inside a battery budget with its market-leading compute efficiency. The company is headquartered in Ontario, Canada with operations in California.

Martin Mason
Blumind Inc.
info@blumind.ai