femtoAI is a US-based startup aiming to bring intelligence to everyday devices with their Embedded AI Platform: an ultra-efficient AI accelerator, end-to-end software stack, and production-grade AI models. As one of the winners of the IC Taiwan Grand Challenge Batch 3 in the AI Core Technologies & Chips category, femtoAI utilizes neuromorphic principles to enable low-latency inference with dramatically lower power and cost & power everything from wearables and appliances to robotics and vehicles.

Embedded AI Platforms Enable AI in Every Device

AI has been the fastest-growing tech trend in recent years and the need for faster, more efficient AI will lead to the development of embedded AI (EAI). With EAI, AI operates on the edge to process data in real-time for faster and more efficient operations without requiring constant connectivity. femtoAI’s core technology of an Embedded AI Platform utilizes sparsity to achieve ultra-efficient neural network inference, circumventing the massive energy and storage requirements that cloud AI systems are known to have. By leveraging sparsity, femtoAI is able to achieve 100x better energy efficiency and 10x smaller model sizes. When combined with near-memory compute to significantly reduce data movement and power consumption, femtoAI’s technology provides a critical advantage for always-on, low-power applications such as hearing aids and smart sensors.

Introducing IC Taiwan Grand Challenge

To strengthen Taiwan's position as a global semiconductor powerhouse, the National Science and Technology Council (NSTC) organizes the IC Taiwan Grand Challenge. Started in 2024, the competition originally focused purely on innovative IC designs and advanced application solutions. Today, ICTGC also honors technologies that can support and contribute to the semiconductor industry and Taiwan’s industry in general. ICTGC Batch 3 emphasizes key fields such as AI Core Technologies & Chips, Smart Mobility, Smart Manufacturing, Smart Medtech, and Sustainability.

Making On-Device Efficient, Real-Time AI Inference a Reality

Founded in 2018 by a group of researchers from the Brains in Silicon Lab at Stanford University, femtoAI has a strong academic foundation with leaders of a number of cutting-edge research at the intersection of hardware and machine learning. Combined with the decades of industry leadership from their board members and advisors, femtoAI is able to move fast while building for scale.

femtoAI’s core technology is a full-stack embedded AI platform built for maximum efficiency and real-world deployability. Its foundation is a custom AI accelerator that exploits sparsity to circumvent unnecessary computations and memory accesses to achieve 100x better energy efficiency and 10x smaller model sizes. They follow a hardware–software co-design philosophy, tightly integrating their chip architecture with an optimized software stack. Their compiler toolchain allows developers to convert standard PyTorch models into highly efficient machine code that runs seamlessly on their Sparse Processing Unit 001 (SPU-001), preserving accuracy while enabling real-time performance even on devices with limited size, weight, or power. By combining architectural innovations such as sparsity and near-memory compute with a hardware-software co-design approach, femtoAI is able to dramatically reduce energy consumption and enable AI to run where it was previously impractical.

femtoAI has a growing portfolio of strategic patents, with 3 granted patents and 7 published applications, covering their core chip architecture, parallelism mechanisms such as scoreboarding, and support for sparse instructions and memory formats. With a number of existing successful collaborations with domestic and international partners, femtoAI’s technology has been utilized by manufacturers of hearing aids, smart glasses, hearing protection, as well as premium audio OEMs and smart application OEMs. femtoAI currently has 15 design wins with 18 more in the prototype stage and a total of 74 active evaluations.

As one of the winners of the IC Taiwan Grand Challenge Batch 3, femtoAI will participate in the Taiwan Innotech Expo (TIE) 2025 which will be held from October 16 - 18, 2025 in the Taipei World Trade Center (TWTC) Hall 1. femtoAI will showcase their technologies, and visitors are encouraged to join and see how femtoAI's innovations can deploy AI at the edge with a platform designed from the ground up for efficiency, scalability, and real-world usability.

Register for the IC Taiwan Grand Challenge Batch 4

With the continued success of IC Taiwan Grand Challenge (ICTGC), the organizers will promote it as a long-term event to support semiconductor innovations from Taiwan and abroad. Registration for the next batch of ICTGC will begin in November 2025, and teams with a focus on semiconductor technologies are encouraged to join and compete.

For more details, please visit:
>>> https://ictaiwanchallenge.org/