基流科技 files for Hong Kong IPO, aiming to become 'AI Infra first stock'
The AMW Read
Novelty 1: IPO filing is a step forward but not unexpected for a fast-growing infra player. Significance 2: Could establish a public-market benchmark for AI infra segment in China.
基流科技 files for Hong Kong IPO, aiming to become 'AI Infra first stock'
AI infrastructure company 基流科技 submitted its IPO application to the Hong Kong Stock Exchange on April 29, 2026, seeking to become the city's first listed AI infrastructure pure-play. The company completed Series C and D rounds totaling nearly 2 billion RMB ($275M), reaching a post-money valuation of $1.3 billion. Founded in February 2023 by CEO Hu Xiaohe, 基流科技 provides full-stack AI cluster products (Galaxy) and operations services, supporting over 90,000 GPUs and 34,000 PFLOPS of managed compute. Revenue grew from 31.8 million RMB in 2023 to 520 million RMB in 2025 (304% CAGR), with adjusted net profit of 31.1 million RMB in 2025.
Why it matters: The IPO filing signals that AI infrastructure – the networking, clustering, and operations layer between chips and models – is maturing into an investable public market category. As model labs and cloud providers race to build ever-larger clusters, specialized companies like 基流科技 that solve cluster-level performance (e.g., RoCE networking, cross-layer optimization) are emerging as critical enablers. The listing could provide a bellwether for the infrastructure segment, which has lagged behind chip and model companies in public-market recognition. With the Chinese AI compute market projected to grow from 454 billion RMB (2025) to 3.8 trillion RMB by 2030 (53.7% CAGR), the IPO tests investor appetite for this essential but less glamorous layer.
Grounded expert take: 基流科技's growth demonstrates the 'fastest-ARR-ramp' pattern common in AI infrastructure, where early partnerships with leading model labs (如智谱AI) create referenceable wins that drive rapid revenue scaling. The company's 80%+ customer retention and shrinking reliance on its largest customer (from 59% to 16.6% of revenue) suggest improving product-market fit and diversification. However, the infrastructure segment faces structural uncertainties as model architectures evolve – CEO Hu acknowledges the need to keep investing in R&D or risk falling behind. The IPO's success will depend on whether public markets share the long-term conviction that AI compute demand will sustain 50%+ annual growth.
