In 2021, two twenty-something agronomists arrived at a 3,000-mu cotton plot on the edge of the Taklamakan Desert and were told they were crazy. One veteran farmer put it plainly: it took his whole family to work 300 mu, and these two kids wanted to run ten times that with almost no one in the field. Four years later, their "super cotton field" (超级棉田) in Yuli County has become the template for an unmanned-farming push now covering more than 2 million mu across Xinjiang.
The project is run by XAG (极飞科技), a Guangzhou-based agricultural-robotics company, and it is one of the clearest examples of AI and autonomy moving from demo video to real acreage in Chinese agriculture.
Key takeaways
- XAG's Yuli "super cotton field" is managed by two people across 3,000 mu (about 200 hectares), using drones, autonomous vehicles, and IoT sensors.
- By 2024 the plot averaged 529 kg of seed cotton per mu, with premium-grade cotton above 96% — among the best in the region.
- Management tasks now run at 75% unmanned operation rate, and by the end of 2025 the model had been extended to over 2 million mu across Xinjiang (Xinhua).
- A newer X150 drone mixes, refills, recharges, plans routes, and sprays on its own — no pilot, no remote control.
What "mu" means, and why it matters
A mu (亩) is roughly 0.067 hectare, or about 1/15 of an acre. Scaling the numbers: 3,000 mu is close to 200 hectares, and 2 million mu is about 133,000 hectares. Cotton is Xinjiang's signature crop — the region produced 92.8% of China's cotton in 2025, leading the nation in both total and per-unit yield for 32 consecutive years (Xinhua). Automating even a fraction of that land is a large deployment by any standard.
How two people run a farm
The Yuli field started as a bet on unmanned management. The two agronomists, Ai Haipeng and Ling Lei, first attacked irrigation — the most labor-hungry step — then layered on autonomous monitoring. Today the loop looks like this: soil sensors report moisture and salinity every ten minutes; a small weather station tracks wind and light; drones fly preset grids daily, scanning for gaps, pests, and weeds; and a smartphone app lets a manager open irrigation valves and tune fertilizer from anywhere. A farm "brain" ingests weather, soil, and crop data and returns precise task instructions.
The latest upgrade is the X150 agricultural drone. On a September 2026 spray run, Ai selected a template on his phone and tapped "start." The drone took off, worked the field, and recharged itself — mixing chemicals, refilling, and planning routes without a pilot or a remote. "In past years we still needed a pilot for the trajectory and to swap batteries," he said. "Now you leave the aircraft in the field and come back the next morning; it has finished the plant-protection work."
The economics that convinced farmers
The pitch is not just novelty — it is cost. XAG's team calculates the super-cotton-field model saves about RMB 254 (≈ US$36) per mu across water, fertilizer, pesticide, and labor, a 22% reduction versus traditional practice. Independent reporting from People's Daily frames the gains broadly: water and electricity costs per mu down about 47%, pesticide use down 30%, and fertilizer efficiency up 40% in comparable smart-rice setups. A single XAG drone can manage roughly 500 mu of farmland.
Yields held up. The Yuli plot hit 529 kg of seed cotton per mu in 2024 with over 96% premium grade, and even in a tougher 2025 climate still out-yielded neighboring fields by about 50 kg per mu. That combination — lower cost, stable or better yield — is what turned skeptics into adopters and pushed the model past 2 million mu region-wide by late 2025.
Beyond cotton
XAG says it has trained more than 140,000 smart-agriculture professionals in China and deployed its robots in nearly 70 countries and regions, ranking second globally in agricultural-drone and robotics sales (People's Daily). The Yuli playbook is being adapted to corn, grapes, and soybeans. A new generation of "tech-savvy cotton farmers" is emerging — younger operators who manage thousands of mu from a phone, a stark contrast to the back-breaking manual harvest of a generation ago.
Honest limitations
- Yield and cost figures (529 kg/mu, 96% premium grade, RMB 254/mu savings, 75% unmanned rate, 2 million mu extension) come from Xinhua and state-media reporting of XAG's project; they describe the company's showcase field and its licensed model, not a randomized regional average.
- The "-47% water/electricity, -30% pesticide, +40% fertilizer" numbers are reported for XAG smart-rice systems, not the Yuli cotton field specifically; cotton results may differ.
- "Second globally in ag-drone sales" and "70 countries" are company claims echoed by state media, not independently audited market rankings here.
- mu-to-hectare conversions use 1 mu ≈ 0.0667 hectare; RMB uses ~7.15 per US dollar.
- Automation covers management and plant-protection; harvesting still uses manned cotton-picker machines in the reported setup.
What readers can do now
- If you farm or advise farms, pilot one autonomous task first — drone spraying or smart irrigation — before betting on full unmanned management; the Yuli story scaled one problem at a time, starting with irrigation.
- Watch the labor math, not the headlines. The real unlock is "one manager per hundreds of mu," which only pays off once input savings exceed the tech and service cost.
- For supply-chain watchers, track agricultural-drone makers' service networks — deployment at this scale lives or dies on local repair, refilling, and training, not just the hardware.