The haul trucks that carry ore out of the pit at a Shougang mine in Tangshan, Hebei, have nobody in the cab. They run fixed routes between the shovel and the dump, and at one pit battery trucks and diesel trucks wait in the same dispatch queue.
What changed there was not a model release. It was a quota in a planning document for how much of the dangerous, dirty and repetitive work must be done by machines.
That quota is the part foreign readers usually miss about industrial AI in China: the customer is often a safety regulator, and the order is written against a number.
Key takeaways
- A first national list of 136 flagship scenarios, issued by six central government bodies, makes smart mining (智慧矿山) item 23, run by the National Mine Safety Administration.
- The mining scenario covers more than 2,800 mines and sets a 40% substitution rate for dangerous posts and 70% intelligent capacity in coal.
- Coal already runs above 65% intelligent capacity, per the China National Coal Association, so the policy margin is thinner than the headline.
- Nearly 8,500 autonomous haul systems (无人矿卡) had shipped by June 2026 and nearly 6,000 were running, CCTV reported; 2026 is forecast above 10,000.
- The RMB 324 billion (≈US$45.6bn / HK$356bn) attached to the scenario is an industry-chain estimate, not a procurement budget.
Three documents, three numbers
The list was issued by the National Development and Reform Commission with the science, industry, human-resources, emergency-management ministries and the state-asset commission. Mining is described as a comprehensive scenario rather than a single product: 21 lead organisations take part, including 15 state-owned backbone firms, 14 of them in mining or coal equipment.
The targets reported alongside it are exact — coal capacity qualifying as intelligent at no less than 70%, intelligent-scenario coverage at no less than 50% of large metal and non-metal mines, and a 40% substitution rate for posts classed as dangerous, dirty, tiring or repetitive.
Set those against two other numbers. The China National Coal Association puts current intelligent coal capacity above 65% (CCTV, 6 November 2026). The industry's fifteenth five-year plan, issued by the development commission and the energy administration, targets 75% by 2030. The 70% is about five points from where the industry says it stands, so most of what remains is the expensive part: thin seams, deep shafts, difficult geology.
Why haulage went first
An open-pit haul route is the easiest place to put a machine: closed site, fixed roads, no pedestrians. The hard problems are energy and dispatch rather than perception, which is why the winning designs are boring — a battery pack, a swap station, and a scheduler.
At Shougang's Tangshan mine 43 trucks run without drivers, and at one pit 17 battery trucks and 11 diesel trucks are dispatched as one fleet; the company reports energy use down about 15% and haul volume up about 10% (CCTV, 25 October 2026).
Underground the same logic appears as distance. At a mine in Heze, Shandong, where seams sit about a kilometre down, operators control equipment more than 200 metres away, and at a trial base in Ordos a quadruped robot surveys beside an intelligent heading machine while the operator works from a desk (CCTV, 6 November 2026). Beijing Longruan's Chen Huazhou told CCTV that mapping the heading face with lidar and radar cuts physical work intensity by more than half.
Suppliers grew with it: Yutong Heavy Industries reported autonomous shipments up 12% in the first half of 2026, Sany Intelligent Equipment said shipments doubled with overseas orders up nearly 90%, and a Changsha-based supplier reported 213% growth (China Financial Information Network, 25 October 2026).
The number that decides the purchase
A mine does not buy autonomy. It buys cost per tonne moved, and that moves with utilisation. Of everything shipped to June, roughly seven in ten units were still in operation — my arithmetic on the two CCTV figures, not a published ratio.
The 40% target is looser than it sounds. It counts posts covered by smart equipment or robots, not only driverless haulage, which leaves room for remote operation and fixed automation. That is a lower bar than autonomy.
The effect shows up in the safety record instead. Shandong Energy's Yunding Technology says its mine model, applied to gas, excavation and transport, raised the speed of hazard handling by more than 80% and cut non-compliant behaviour by more than 75% (CCTV, 6 November 2026).
What travels, and what does not
Exports are a line item: Chinese autonomous trucks operate in Southeast Asia, Africa and Latin America, adapted to local geology and climate.
Sun Zhidong, a senior researcher at Anxin Fund, expects domestic shipments to pass 20,000 a year within two to three years, with smaller mines following after 2028. He cites a Thai project where the Chinese system cost about 40% less and deployed about 50% faster than European or American equivalents (CCTV Finance, October 2026).
The caveat is physical: a swap station needs grid capacity, spare batteries and a depot within reach. The truck exports; the conditions that make it cheap do not.
Honest limitations
The 136-item list is reported consistently by People's Daily, Chongqing Daily and Guangming Online, but I could not retrieve the document itself from the commission's site, so the 70/50/40 targets rest on those reports; the energy administration's regional office states the same 65% current level, corroborating the association but not the targets. The 8,500 and 6,000 figures are CCTV's, credited to unnamed statistics, and count systems shipped rather than mines automated. Shougang's energy and volume changes are company figures given on camera, unaudited. No ministry has published how the 40% rate is measured or how much has been achieved.
What readers can do now
- When a Chinese policy sets a substitution rate, ask what counts as a post. If remote operation is included, the quota can be met without autonomy, and equipment demand is smaller than the headline implies.
- For buyers, request availability and cost-per-tonne for a specific site rather than fleet totals; the industry's own numbers show roughly a tenth of shipped units not in service.
- For analysts, track the gap between intelligent-capacity share and the 2030 target of 75%. As it narrows, demand shifts from new fleets to retrofits in difficult geology, which is a different product mix.