光本位 is developing optical processing unit (OPU) chips that use a PCM+Crossbar photonic computing architecture to pursue higher integration and lower power consumption. The company says its OPU design maintains interfaces and an instruction set compatible with existing data center stacks, enabling adaptation to large AI models. 光本位 has completed three tape-outs and reports rapid product iteration after just over a year of operation, positioning commercialization as near-term. The team includes two 1995-born co‑founders: Xiong Yinjian (Chicago University, TensorFlow experience) handling management and commercial operations, and Cheng Tangsheng leading R&D with advisory support from Harish Bhaskaran. Core engineers have backgrounds at Hesai, Qualcomm and others. Management and investors argue the photonic approach offers large gains in power, latency and parallelism versus electronic chips and benefits from mature domestic silicon-photonics processes, making it well-suited to growing generative-AI compute demand.

Total raised

$14M

Funding rounds

1

Latest round

Equity

Latest activity

Mar 2024