By Hamza Faisal
China's AI infrastructure race just got a major cash injection. Shanghai-based Infinigence AI has completed a RMB 2.3 billion — about $344 million — Series C funding round, lifting its cumulative financing past RMB 4.3 billion, or roughly $644 million, according to Chinese media reports published October 12.
The round is reportedly the largest ever raised in China's AI-native infrastructure category, underlining how investor money is flowing toward the computing layer beneath the AI boom, not just the model developers sitting on top of it.
A record raise for AI-native infrastructure
Founded in 2023 by researchers linked to Tsinghua University and Shanghai Jiao Tong University, Infinigence builds software and infrastructure that lets AI models run efficiently across different chip platforms.
The new round drew a broad investor roster including Tencent, Baidu, Lenovo, Sequoia China, Qiming Venture Partners and Shunwei Capital, alongside state-linked funds such as Hangzhou High-tech Financial Investment and the Social Security Zhongguancun Fund.
Infinigence says its platform stitches together 37,000 petaflops of computing capacity across 16 mainstream chip types, supporting mixed-chip training that combines Chinese and international processors.
It also runs model inference services and infrastructure for enterprise AI agents, with deployments in gaming, healthcare, legal services and energy — and counts Chinese model developers Zhipu AI and Moonshot AI among its customers.
Mixed-chip compute in a chip-constrained China
The company's pitch lands at a moment when compute is the constraint.
Infinigence says daily token usage on its model-service platform had risen 40-fold by July compared with December 2025, while its cross-cluster inference architecture cut the time to generate the first token by 51.5 percent and the cost per token by 37.5 percent, according to company figures cited by TechNode.
Mixed-chip capability matters because Chinese labs cannot rely on a steady supply of the most advanced American accelerators.
Infrastructure that squeezes more inference out of whatever hardware is available — domestic or imported — has become a category of its own, and investors are paying up for it.
A Hong Kong IPO as early as 2027
Infinigence had already confidentially filed for an initial public offering on the Hong Kong Stock Exchange in early October, Chinese outlet Guancha reported, with plans to raise several hundred million dollars and a potential debut as early as the first half of 2027.
Chinese reports also cited a pre-IPO valuation of RMB 14.3 billion for the company.
The timing would make Infinigence one of the first pure-play AI infrastructure companies to test Hong Kong's appetite for the sector, as Chinese tech firms increasingly look to the city's exchange to fund the next phase of the AI build-out.
By Hamza Faisal



