Pioneered domestic optical interconnect GPU-OCS commercial. Full stack self-developed for extreme perf. Leadership in heterogenous mixed training standards.
High R&D intensity with awards (SAIL, national giant, Shanghai hard tech TOP100). Partnerships 100%+ perf boosts; first HK GPU IPO; 2026 Tech100 inclusion.
Market Outlook
Biren positions as key player in China sovereign AI and domestic GPU market, first HK listed GPU company. Strong in heterogenous compute, mixed training standards leadership. Partnerships with telcos and AI labs for inference/training.
Despite sanctions, pivots and funding support domestic ecosystem growth. 2026 Tech100 index inclusion signals market recognition.
Product Overview
Biren develops the 壁砺™ (Bili) 166 series of AI accelerators including 166M (wind-cooled OAM module, 550W peak for data center train/infer), 166L (liquid-cooled), and 166C (high-performance inference card). Earlier BR100/BR104 GPGPUs use dual-die Chiplet design on 7nm.
Paired with BIRENSUPA™ software platform for full-stack optimization, framework support, accel libs, inference engine, SDKs and toolchain. Supports optical interconnect for GPU-OCS comms.
