LeapMind

The core of LeapMind's business is its proprietary ultra-low power AI inference accelerator IP, known as "Efficiera". This product is designed for field-programmable gate arrays (FPGA) and application-specific integrated circuits (ASIC), specializing in convolutional neural network (CNN) inference processing. A key technological advantage is LeapMind's "extremely low bit quantization" technology, which minimizes the number of quantization bits to as low as 1-2 bits. This approach significantly enhances power and area efficiency without requiring advanced semiconductor manufacturing processes, allowing for the integration of deep learning functions into a variety of edge devices with power, cost, and thermal constraints. These devices include consumer electronics, industrial machinery, surveillance cameras, and small robots.

LeapMind's business model revolves around licensing its Efficiera IP to companies that design ASIC and FPGA circuits. This enables customers to develop cost-effective, low-power edge devices and accelerate their time-to-market for products with AI capabilities. To support its clients, LeapMind offers a suite of related products and services, including the Efficiera SDK (a software development kit), the Efficiera Deep Learning Model for efficient model training, and Efficiera Professional Services for creating custom, low-bit quantized deep learning models. The company has a track record of working with over 150 companies, particularly in the manufacturing and automotive sectors.

In addition to its edge AI solutions, LeapMind has announced the development of a new AI chip for training and inference, targeting the server side. This chip aims for a high level of computing performance with a significantly better cost performance compared to equivalent GPUs, with an anticipated shipment date by the end of 2025. The company has also developed an AI-powered image processing model that operates on edge devices to reduce noise and enhance image quality in real-time.