signal/raiseTrack rounds
Series CAI · Shanghai

Infinigence AI raises RMB 2.3bn Series C for multi-chip AI infrastructure

The Tsinghua spinout, which pools compute across 16 chip types and sells model tokens to AI agent builders, says the round lifts its funding since 2023 to RMB 4.3bn (about $643m), a record for China's AI infrastructure sector.

s/rFunding deskEast Asia · 2 min read

Infinigence AI (Wuwen Xinqiong), a Chinese AI infrastructure company, has completed a RMB 2.3bn (about $344m) Series C round, according to company materials released on October 12, 2026 and reported by Phoenix Tech and TechNode.

The round takes the company's total funding since it was founded in 2023 to RMB 4.3bn (about $643m). The company said both the single round and its cumulative total are the largest yet for China's AI-native infrastructure sector. No lead investor or valuation was disclosed.

Zhidongxi, in a report carried by Phoenix Tech, said the round has already closed and that the company is preparing for an initial public offering. IT Home, citing Bloomberg, reported that Infinigence has confidentially filed for a Hong Kong listing and could list as early as the first half of 2027.

What Infinigence AI does

Infinigence was spun out of Tsinghua University and Shanghai Jiao Tong University and grew out of a Tsinghua electronic engineering lab led by professor Wang Yu. Co-founder Xia Lixue is chief executive, and co-founder Dai Guohao, an associate professor at Shanghai Jiao Tong University, is chief scientist.

The company does not design chips or train its own large models. Instead it builds software that lets models train and run across many kinds of processors. Its compute network reaches 37,000 petaflops of capacity across 16 mainstream chip types, and it says it has trained models on a mix of Chinese and imported chips at up to 97.6% utilisation. It owns no data centres and manages capacity sitting in third-party facilities.

It also runs a model-as-a-service platform that sells tokens to companies building AI agents. The company said a cross-cluster inference design released in July 2026 cut time to first token by 51.5% and cost per token by 37.5%, and that daily token usage in July was 40 times its December 2025 level. It works with model makers Zhipu and Moonshot AI and sells industry packages for gaming, healthcare, legal services and energy.

Who invested

Phoenix Tech listed the round's backers in three groups. State-linked investors included Hangzhou High-Tech Financial Investment, the Social Security Zhongguancun Fund, Zhuhai Technology Industry Group, Xuhui Capital and Guokai Kechuang. Industry investors were Lenovo, Tencent, Baidu and Jiukun Ventures. Financial investors included Sequoia China, Qiming Venture Partners, Fortune Capital and Shunwei Capital.

Other Chinese outlets described those names as the company's broader investor base rather than a breakdown of this round, and did not give allocations.

The funding so far

Infinigence raised a Series A of nearly RMB 500m (about $70m) in September 2024 and an A+ round of nearly RMB 500m in November 2025, IT Home reported. In May 2026 it announced more than RMB 700m (about $104m), co-led by Hangzhou High-Tech Financial Investment and Huiyuan Capital.

The company organises its business around what it calls agentic infrastructure in three layers: pooled compute, its token platform and industry applications. Zhidongxi compared the round with large valuations for US inference companies such as Fireworks AI, Baseten and Together AI.

Company website: infini-ai.com