Skip to main content
Back to News
Funding
2 min read
CN

Approaching.AI raises over RMB 1.3B Series B to build inference token factories on domestic chips

The AMW Read

A sub-$500M round for a known Chinese inference-infra player, but its state-backed token-factory model on mixed domestic chips adds a meaningful datapoint for CN inference supply.
NoveltySignificance
AI Infra · Player Map

Approaching.AI raises over RMB 1.3B Series B to build inference token factories on domestic chips

Approaching.AI (趋境科技), which sells AI inference output by the token, has closed a Series B of more than RMB 1.3 billion (~$180M), Leiphone reported. Zhecai Innovation, Xiaoshan Capital, ICBC Investment, Zhongguancun Science City, Xinglian Capital and Qianjiang Century City co-led the round. SenseTime and Wangsu Science & Technology joined as followers. The round comes three months after its Series A in July and brings 2026 funding to RMB 2.4 billion (~$335M). The money will go to token capacity, its ATaaS (Token as a Service) platform, inference R&D and delivery. The company says daily token output grew more than 100x in six months to over one trillion tokens. It says revenue so far this year is close to ten times last year's total, and that the business now covers its costs.

The pitch is that inference capacity is sold as an output, measured in tokens, not as rented GPUs. Approaching.AI says it splits prefill and decode across different chips and converts KV caches between them. That lets models with more than a trillion parameters run on Chinese accelerators from partners such as Moore Threads, MetaX and Iluvatar CoreX. It claims better cost-performance than the leading foreign hardware. Local state capital drives much of the buildout. Henan runs a joint operation, a Hangzhou facility is planned at the 10,000-card scale, and Hainan uses a model where the state builds and Approaching.AI operates. The team also started the open-source KTransformers project and helped found Mooncake.

For investors, the open question is whether operating software can turn idle, mixed domestic chip fleets into profitable token supply. Revenue backed by state co-investment will need checking against demand from paying enterprise customers.

#AIInference #ApproachingAI #ChinaAI #AIInfrastructure #Funding #DomesticChips

#Approaching.AI#AI inference#Series B#token factory#domestic chips#ATaaS#related:商汤科技#related:并行科技#related:弘信电子#related:网宿科技#related:摩尔线程

How This Connects

Based on AI Infra · Player Map

  1. 8h agoApproaching.AI raises over RMB 1.3B Series B to build inference token factories on domestic chips · THIS ARTICLE
  2. 23h agoNVIDIA opens Vera Rubin rack designs for denser AI data centersNVIDIA
  3. 4d agoNscale's $4.19B H1 financing combines Series C equity and data-center debtNscale
  4. 1w agoMicrosoft plans multigigawatt Maia deployment from 2027 with inference-focused Maia 300Microsoft
  5. 1mo agoCoreWeave completes $4.2 billion convertible note offering for AI data-center expansionCoreWeave
  6. 1mo agoNvidia works to ease the electrical power bottleneck slowing AI data center expansionNvidia's

Related News

More news from 趋境科技

Stay updated with the latest news and announcements from 趋境科技.

View all 趋境科技 news

Discover AI Startups

Explore 5,000+ AI companies with VC-grade analysis, funding data, and investment insights.

Explore Dashboard