Runjian's $200 million token contract puts delivery and collection in focus
An annual purchase minimum can strengthen a supplier's contractual position. It still leaves a gap between signed demand, delivered AI services and collected cash.
What does a large AI services contract tell you before the customer starts paying? For Runjian, the useful answer is in the billing clauses.
In an October 19 issuer notice, the Shenzhen-listed company, stock code 002929, says its wholly owned subsidiary Wuxiang Yungu recently signed a one-year AI token services agreement with an unnamed overseas customer. The contract sets a $200 million annual purchase minimum for October 19, 2026 through October 18, 2027. Runjian's company homepage lists a related announcement dated October 18; the exact signing and first-disclosure times are unknown.
The floor and the monthly bill do different jobs
The customer accesses AI inference through an application programming interface, or API, backed by computing resources that the notice says are located in China. Token usage measures the amount of model input and output processed. Prices are set in written orders and are not disclosed in the notice.
Monthly charges follow actual usage. The subsidiary issues a bill by the fifth day of the following month, with payment due by the fifteenth. Prices may be expressed in dollars or equivalent renminbi, but settlement is entirely in renminbi.
The annual floor adds another obligation. After the contract year ends, the parties have ten working days to reconcile usage charges against the $200 million minimum. If usage falls short, the buyer owes the difference. That reconciliation period is not a stated deadline for paying the shortfall.
Contracted demand still has to become cash
Our reading: the minimum reduces the supplier's contractual exposure to low usage. It does not prove the customer will consume that much service, pay every invoice or cover the supplier's costs. The notice withholds the customer's identity under an internal confidentiality exemption, so readers cannot independently assess that counterparty from the disclosure.
Runjian says AI, computing and token businesses remain a small share of total revenue. It also warns about delivery interruptions, contractual liability and substantial upfront investment. Recognized revenue will depend on audited figures.
The next useful evidence is therefore operational and financial: services delivered, bills collected, costs incurred and revenue recognized. A large purchase floor gives investors a concrete contract to follow. Those later disclosures will show how much of it becomes a durable business.