Greater Bay Area AI Computing Tech has reported a sharp first-half shift from a mixed infrastructure business toward AI computing services. In an HKEX filing released on August 28, the Hong Kong-listed company said revenue for the six months to June 30 reached RMB2.55856 billion, an increase of 1,020.9% from a year earlier. Its AI business contributed RMB2.382821 billion, or 93.1% of total revenue, while period profit was RMB281.365 million.
The figures are unaudited interim results, and the company’s statements about market position, order strength, and future strategy remain issuer disclosures. Still, the filing offers a detailed look at the scale of China’s AI infrastructure buildout. Rather than selling a model or a finished application, Greater Bay Area AI Computing is positioning itself as a provider of AI computing technical services, token services, and high-performance server integration and delivery.
The company says it has more than 50,000 PFLOPS of FP16 dense compute capacity and is investing in a scheduling platform called Quantumpai, a heterogeneous-computing system, and an inference-optimization engine. Together, those details show how the business is trying to compete in a market where customers want access to different chips, flexible capacity, and increasingly, AI services priced around token generation rather than rented hardware alone.
AI Services Become 93.1% of Greater Bay Area AI Computing Revenue
The headline number in the filing is the revenue increase from RMB228.252 million in the first half of 2025 to RMB2.55856 billion in the first half of 2026. The company attributes the shift to its AI business, which contributed RMB2.382821 billion during the latest period. Its remaining infrastructure business, including property sales, property management, rental income, and data-center development and operations, contributed RMB175.739 million.
The mix is significant because it indicates how quickly the company has changed its commercial identity. The filing says the group’s AI business includes AI computing technical services, token services, and high-performance-server integration and delivery. Those are related but different revenue streams. A customer may pay for a managed computing service, use capacity to generate tokens through an inference platform, or buy a server solution integrated for a specific workload.
The company’s gross profit was RMB584.731 million, while the filing reported RMB1.973829 billion in sales costs. Greater Bay Area AI Computing listed an AI-business gross margin of 35.5% in its management discussion. The financial statement also shows a period profit of RMB281.365 million. These are useful snapshots, but they should not be treated as a prediction of full-year profitability. The results are unaudited and reflect a company undergoing a rapid shift in business mix.
China’s AI infrastructure industry has increasingly emphasized computing as a service rather than a static asset. EastFrontier’s earlier reporting on regional AI compute clusters expanding across China described the broader contest to place capacity near energy, industrial, and user demand. Greater Bay Area AI Computing’s filing adds a company-level view of that process, with AI services now representing more than nine-tenths of reported revenue.
More Than 50,000 PFLOPS Supports a Token-Service Strategy
The company says its operational FP16 dense compute capacity exceeds 50,000 PFLOPS. A PFLOPS is a measure of floating-point computing speed, and the use of an FP16 measure is relevant to many AI training and inference workloads. Capacity alone does not reveal utilization, chip mix, power costs, or customer demand, but it signals the type of scale the company believes is necessary to serve large AI workloads.
The filing also identifies a move from selling raw computing capacity toward providing token services. In the language of the company’s management discussion, the business is seeking to shift from an “AI computing warehouse” to a “token factory.” That phrasing is an issuer’s strategic description, but it captures an important commercial change. Instead of charging only for server time or installed equipment, an AI infrastructure company can try to charge for the useful output produced by a model’s inference process.
To support that shift, Greater Bay Area AI Computing says it is investing in Quantumpai, its own compute-scheduling platform, a heterogeneous-computing scheduling system, and an inference-optimization engine. Heterogeneous scheduling is important because AI workloads may rely on different kinds of processors, accelerators, networks, and storage. A scheduling system that can allocate work across those resources can matter as much as the raw number of machines.
The company’s approach echoes wider efforts to make AI capacity easier to use across different hardware environments. EastFrontier’s coverage of Huawei’s Ascend ecosystem and its Wuxi token factory examined a similar attempt to connect domestic compute hardware with model-serving demand. Greater Bay Area AI Computing is presenting itself as a separate platform player in that same transition from hardware supply to managed AI output.
The Filing Shows Both Rapid Scale and Financial Constraints
The interim results contain evidence of rapid growth, but they also show the capital intensity of the business. Greater Bay Area AI Computing reported total assets of RMB16.812555 billion as of June 30 and bank loans and other borrowings of about RMB5.121 billion. Financing costs in the first half were RMB155.531 million, compared with RMB23.05 million a year earlier. The company says the higher costs were associated with borrowing for AI computing equipment purchases, capacity expansion, and related operations.
The filing says new financing in the first half carried a weighted average cost of about 3.5%, compared with 5.45% a year earlier. It also says the company had received more than RMB50 billion in credit lines from financial institutions. Those are company disclosures, and credit availability does not guarantee future revenue or cash flow. But they illustrate the financial scale required to build an AI computing business in which equipment, facilities, electricity, networking, and financing all matter.
The reported AI revenue growth is especially striking because the prior-year comparison started from a much smaller base. That makes the 1,020.9% increase mathematically dramatic but also means readers should pay attention to recurring demand, service margins, and collection of receivables as the business grows. The company reported trade and other receivables of RMB2.14957 billion as of June 30, alongside cash and cash equivalents of RMB1.007245 billion.
For investors and industry observers, the filing provides a concrete indicator of how China’s AI infrastructure race is spreading beyond foundation-model developers. Greater Bay Area AI Computing has reported RMB2.55856 billion in first-half revenue, RMB2.382821 billion from its AI business, FP16 dense capacity above 50,000 PFLOPS, and R&D focused on scheduling and inference optimization. Those facts do not prove that every compute provider can translate capacity into sustainable returns. They do show that one Hong Kong-listed company is trying to build a business around the full path from servers and data centers to tokens and AI services.