TetraMem

Founded in January 2024 and based in Shanghai, Guangqiu Technology develops silicon-photonic AI processors for edge computing using an all-optical reservoir-computing architecture built on the nonlinear dynamics of semiconductor lasers. Co-founded by ShanghaiTech professor Wang Cheng and Li Ruiqian, its chips target sub-100-nanosecond system-level inference latency, 10GHz-class operating frequency and real-time adaptive training. Its silicon photonic chip has completed a first tape-out validation.

Sivy (时维感知, Hangzhou Sivy Technology Co., Ltd.), founded in 2025 and headquartered in Hangzhou, China, is a neuromorphic sensing-and-computing startup targeting edge and embodied-AI applications such as robotics. Building on its own neuromorphic chips and brain-inspired network models, it develops ultra-low-power sensor-processing chips based on third-generation (spiking) AI, aiming to serve as a general-purpose sensing platform that replaces traditional DSPs. It has taped out its first chip, ASC-01, a pulse-encoding chip that converts analog signals into pulse signals (up to 10 MHz sampling) and has completed packaging, power-on and testing. On top of ASC-01 it is developing multimodal sensing modules such as electronic skin and electronic nose, plus an EMG armband for next-generation human-machine interaction, toward a bionic 'perception-decision-execution' machine nervous system.

Ambarella is a fabless semiconductor company headquartered in Santa Clara, California, founded in 2004 by Fermi Wang and Les Kohn and listed on Nasdaq (AMBA) since 2012. It designs low-power video compression and computer-vision AI SoCs on its CVflow architecture, plus the Oculii AI radar platform, serving automotive ADAS/autonomous driving, intelligent security, consumer cameras and industrial/robotics edge computing.

Ubitium, founded in 2023 in Düsseldorf, Germany, develops a universal RISC-V microprocessor capable of handling various computing tasks including AI without the need for dedicated chips.

Fractile, founded in 2022 in London, UK, develops advanced AI chips that run AI models 100x faster than traditional methods at 10x lower cost.

Yezhixin Technology is a Shanghai-based edge-AI silicon company building self-developed high-efficiency NPUs from 4 TOPS to 40-80 TOPS for smart glasses, phones, robotics and autonomous driving, with a roadmap toward on-device 7B multimodal model inference; the founding team came from Silicon Valley with Apple Neural Engine, NVIDIA and AMD backgrounds.