How far have domestic computing-power projects progressed as overseas construction faces obstacles?

  • The report says that regulatory measures and electricity-price protection measures are slowing the expansion of US data centers, while calls to halt the construction of new large-scale projects have also emerged in several parts of Europe. By contrast, China is seen as a potential beneficiary of computing-power investment, thanks to advantages in electricity costs, infrastructure and engineering delivery.
  • The Ulanqab project in Inner Mongolia has a planned IT capacity of up to 800 MW, with the first 120 MW reportedly scheduled for delivery within a year; this remains a planned timeline. Operating data offer another point of reference: UCloud’s utilization rate for cabinets built in the first half of 2026 exceeded 95%, and more than half of the 3136 cabinets planned for Building D had been booked. Dongfang Guoxin’s Helingeer project recorded revenue of 104 million yuan through the third quarter of 2026; four buildings have been delivered and are operating, while two more are expected to be delivered by year-end.

What steps remain before the cost advantages of space computing can be scaled up?

  • The report lists a Starship launch-cost target of 20 to 40 dollars per kilogram, far below the approximately 5000 dollars for Falcon 9. The material puts the economic viability threshold at 200 to 500 dollars per kilogram. However, this cost advantage depends on Starship achieving reusable recovery; the report says its 15th test flight is planned to attempt catching the booster and spacecraft simultaneously.
  • The estimates of space demand remain institutional forecasts: the report lists annualized computing-power rental revenue of 41.16 billion dollars in 2027, computing capacity of 5 to 6 GW, and demand for around 1.5 million computing cards. Based on 20 thousand cabinets and titanium plates worth 40 thousand yuan per cabinet, revenue for the titanium-plate segment is estimated at about 800 million yuan; applying a net margin of 20% gives estimated profit of about 160 million yuan. Meanwhile, radiation-hardened HBM and GPUs remain payload bottlenecks. The first batch of Starmind satellites is planned for launch in the third quarter of 2027, meaning there is still a time gap between demand expectations and supply-chain delivery.

As Token services go global, how should orders be distinguished from revenue and expansion?

  • Wuxiang Yungu, a subsidiary of Runjian Co., signed a one-year Token-service order worth 200 million dollars with an overseas company, equivalent to approximately 1.338 billion yuan. The model deploys computing capacity in China and provides API-call services overseas, with settlement based on Token consumption; revenue is linked to actual call volumes.
  • Beyond the order, institutions expect Wuxiang Yungu’s first phase to generate annual revenue of more than 3 billion yuan once it reaches full capacity, and estimate a gross margin of 50%, a net margin of 40% and a payback period of under two years. The planned scale of phase two is 10 times that of phase one, with capacity for 200 thousand cards. The target is to obtain energy-consumption quota approval in 2026, with production expected to ramp up gradually in the second half of 2027. These expansion and profitability figures are based on future construction and full-capacity estimates; they are not equivalent to the value of orders already signed.