Runjian’s $200 million contract highlights a new way to sell Chinese AI services overseas, but thin margins and heavy upfront investment remain major challenges.
A $200 million AI services contract has put Chinese listed company Runjian Corp. (002929.SZ) in the spotlight, highlighting a new way for China’s AI industry to generate overseas revenue: selling access to computing power through token-based services.
Signed on Oct. 18, the one-year agreement allows overseas customers to access large language model (LLM) inference services through an API, using computing resources located entirely in China. The contract includes a minimum purchase commitment: if the customer’s spending falls short of $200 million over the contract year, it must make up the difference under the agreed terms.
The deal is equivalent to about 35.6% of Runjian’s revenue in the first half of 2026. Rather than selling servers or chips, the company is selling the computing capacity used to run AI models, measured and billed in tokens.
But the headline contract value is only part of the story. The real test is whether Runjian can deliver the services profitably, collect payments and maintain margins amid intense competition.
A new route for China’s AI exports
Tokens are the units used to measure the text processed and generated by AI models. Customers can access models through APIs, integrating them into applications without having to build their own inference infrastructure.
Chinese models including DeepSeek, Moonshot AI, Z.AI (Zhipu AI) (2513.HK), and MiniMax (0100.HK) are expanding overseas. Model performance, pricing, response speed and reliability all influence customers’ choices.
Yet developer interest does not automatically translate into sustainable revenue. Trials, paid usage, enterprise procurement and contract renewals represent different stages of commercialization. Rising token volumes must also be assessed alongside prices and delivery costs.
Runjian offers a different route into the market. Traditionally focused on telecommunications network management and operations, the company has expanded into digital infrastructure, energy networks and computing services.
Under the latest agreement, its subsidiary Wuxiang Cloud Valley will provide overseas customers with LLM inference services through APIs, while the computing infrastructure remains in China. The arrangement illustrates how companies without their own foundation models can participate in the international AI market through computing resources and service delivery.
Thin margins and heavy investment
Two broad groups are pursuing opportunities in the token economy. Model developers sell API access to their own models, competing on capability and inference efficiency. Computing service providers supply infrastructure, deployment, API connectivity and operational support. The two can also work together, with different companies supplying models, computing capacity and overseas customer relationships.
But as similar services become more widely available, customers can compare prices, speed and reliability. Token volumes alone therefore offer an incomplete picture of business performance. Customer retention, inference costs, gross margins and payment collection matter just as much.
Runjian’s financial results illustrate the challenge.
The company reported revenue of 3.76 billion yuan ($562 million) in the first half of 2026, making the contract value equivalent to about 35.6% of its half-year revenue. The agreement runs from Oct. 19, 2026, to Oct. 18, 2027, with usage-based billing settled monthly. Actual revenue recognition will depend on service delivery and accounting treatment.
According to its first-half report, Runjian’s computing network business generated about 519 million yuan in revenue, up 50.3% year over year. Operating costs rose 55.1% to about 483 million yuan, leaving a gross margin of 6.97%, down 2.86 percentage points from a year earlier.
The company attributed the relatively low margin of its token factory business partly to high depreciation costs as fixed assets entered service.
Servers require substantial upfront investment, followed by ongoing depreciation, electricity and maintenance expenses. Higher demand will not necessarily translate into stronger profits unless companies improve equipment utilization.
As a simple illustration, if the full $200 million contract value were recognized as revenue at the existing 6.97% gross margin, gross profit would be about $14 million, or 93.2 million yuan. This is a hypothetical calculation, not a forecast of the contract’s actual profitability.
Cash flow and repeat business will be critical
Runjian’s operating cash flow highlights the funding pressures involved. Net cash outflow from operating activities reached about 2.88 billion yuan in the first half of 2026. The company attributed this partly to investment in computing and energy businesses, as well as customer payments concentrated in the second half of the year.
Computing providers often have to commit substantial funds before recovering costs through service revenue. Strong demand and higher utilization can spread depreciation across more business; disappointing purchases risk leaving expensive equipment underused.
Cross-border services also bring compliance requirements involving customers, data handling and applicable regulations. Keeping computing infrastructure in China does not eliminate these risks.
Runjian has warned that token services require significant capital, technical capabilities and customer-development efforts. Competition is intense, and future orders and large-scale commercialization remain uncertain.
Can token exports become sustainable?
For Chinese AI companies, token exports offer another way to reach international customers. Model developers can sell access to their technology, while computing providers can use domestic data centers to supply inference services.
For companies that have invested heavily in AI infrastructure, overseas demand could help put existing computing resources to commercial use. But the opportunity’s long-term value will depend on three factors: repeat purchases and renewals, improving gross margins, and the conversion of contract revenue into reliable cash flow.
Runjian’s $200 million deal shows that cross-border token services have moved beyond industry discussion into commercial contracts. The next test is whether the company can turn that headline figure into collected revenue and acceptable returns.
Ultimately, token exports will become a durable source of overseas income only if Chinese companies can deliver services that customers continue to pay for while maintaining healthy margins and cash flow.
Source:
