Hefei Zhongke Brain Intelligent Technology has announced a B+ strategic financing round worth hundreds of millions of yuan, with CRRC Capital leading and Ginkgo Valley Capital, Shuimu Fund, and TusPark Fund participating. The announcement was published through an authorized QbitAI release on September 7. Because the factual base is a company-provided release, the size, use of proceeds, investor rationale, and technical claims should all be read as Zhongke Brain’s own disclosures rather than independently audited conclusions.
The company is trying to position itself around a term that is appearing more often in China’s AI infrastructure debate: the token factory. In Zhongke Brain’s telling, a token factory is not merely a cluster of GPUs or domestic accelerators. It is a system that manages computing and electricity together, aiming to reduce the cost of producing model output while coordinating heterogeneous hardware and power resources.
The financing is significant because it links a state-owned industrial investor to a company working on the economics underneath AI services. EastFrontier recently covered Bolun Zhihui’s angel funding for heterogeneous AI clusters, another example of capital flowing toward the infrastructure layer rather than only toward general-purpose models. Zhongke Brain’s announcement adds an energy-management angle to the same broad theme.
CRRC Capital Joins a Power-and-Compute Thesis
According to the company release, CRRC Capital led the B+ round. CRRC Capital is connected to CRRC Corporation, a central enterprise with a substantial industrial and transportation equipment footprint. Zhongke Brain said Ginkgo Valley Capital, Shuimu Fund, and TusPark Fund also participated.
The identity of the lead investor is relevant because Zhongke Brain’s stated business depends on bringing together energy systems and computing systems. The company says it wants to optimize inference across Nvidia hardware, domestic chips, and supernode-style systems. It also says it will work on what it calls power-computing synergy, a model in which scheduling decisions consider both available computing resources and electricity conditions.
The company described the round as another milestone after a 2025 strategic investment by China Mobile’s Beijing China Mobile Digital New Economy Industrial Fund. That earlier backing is a company-reported fact, but it provides a useful picture of the type of partners Zhongke Brain is seeking. China Mobile represents the communications and data-infrastructure side of the market. CRRC Capital brings a different industrial and energy-adjacent profile.
The investment does not by itself prove that Zhongke Brain’s technical approach will succeed. It does show that multiple Chinese industrial funds see enough potential in the thesis to provide strategic capital. For a young infrastructure company, that can matter as much as an ordinary financial investment because prospective customers may need confidence that a supplier can remain present through long data-center or industrial projects.
The Token Factory Is an Inference-Economics Argument
Zhongke Brain says the new capital will support three areas. First, it plans to deepen core inference optimization across heterogeneous systems. Second, it plans to work on a decision-making layer for coordinating electricity, computing, and token output. Third, it says it will expand project-delivery and ecosystem capabilities for token factories.
The focus on inference is important. Training an AI model may require a concentrated burst of high-end computing, but inference is the continuing cost of serving the model. Every user query, agent action, document-processing task, and generated image requires resources. As AI applications spread, the cost and efficiency of producing tokens can shape whether a service is affordable.
The company’s phrase “power-computing synergy” refers to a scheduling problem. Electricity conditions may change by time and place. Different accelerators may be available at different prices or performance levels. A system designed for this environment would try to decide which workloads should run where, on which chips, and at what time. Zhongke Brain says it intends to use multi-agent game methods, reinforcement learning, and real-time optimization for that purpose. These are the company’s technical plans, not results from an independent performance evaluation.
The broader policy environment gives the idea some relevance. China’s new information-and-communication blueprint targets much more intelligent-computing capacity by 2030. EastFrontier also reported that Greater Bay Area AI computing providers had described a tenfold revenue jump. More capacity creates a larger need to decide how capacity is allocated, powered, and monetized.
Funding Tests Whether Token Factories Can Become a Business
The word “factory” implies repeatability. A startup can build a custom data-center project for a single customer, but that is not necessarily a scalable product. Zhongke Brain says it intends to create a replicable standard for token factories and build an ecosystem around planning, construction, and operations.
That ambition will be hard to execute. A useful inference platform needs compatible hardware, reliable energy access, scheduling software, security controls, and customers with workloads substantial enough to justify integration. The company also has to compete with cloud providers, telecom operators, chip vendors, and other AI-infrastructure startups that are all building versions of the same stack.
Zhongke Brain’s CEO, Liu Haifeng, said in the release that token industrialization has become a central issue for the AI industry. The statement captures a real commercial question: can AI output be produced and sold with enough efficiency to support large-scale services? But it remains a company’s framing, and the answer will depend on customer deployments and independently measurable economics.
The B+ round gives Zhongke Brain more resources to test that proposition. Its value will not be measured only by the amount raised. Investors will want to see whether the company can demonstrate lower inference costs, reliable heterogeneous scheduling, and real projects that connect power planning with AI workloads.
For China’s AI ecosystem, the significance is that financing is moving deeper into the operational layer. Model companies need chips and data centers, but they also need ways to make every unit of computation economical. Zhongke Brain’s round is a bet that the next infrastructure opportunity lies not only in adding more servers but in making the relationship between electricity, hardware, and token output more intelligent.