DJI Agriculture — the world's largest agricultural drone fleet, and the numbers nobody can check

East-Asia (origin); deployed globally

Content

DJI Agriculture is the largest agricultural drone operation in the world. What the corpus can now say precisely is how much of that is verified: none of it.

Scale — vendor-stated, and inconsistent between DJI’s own 2026 documents.

The environmental claims do not reconcile. DJI’s earlier pair (222 Mt water, 30.87 Mt CO₂, end-2024) is consistent with the 2026 pairs as a cumulative series, but the two 2026 documents disagree with each other by 70 Mt of water and 10.32 Mt of CO₂, and neither publishes a methodology, a baseline year, or an as-of date beyond the word “cumulative”. The likely explanation — different reporting windows — cannot be confirmed from public sources. Western trade press repeated both sets without noticing; one outlet rendered 30.87 Mt as “nearly 31 tons”, a three-order-of-magnitude error that shows the level of scrutiny these figures receive. G-033 is closed: vendor-stated, methodology undisclosed, no independent verification found.

The national-scale anchor is MARA, not DJI. For China specifically, use the ministry’s January 2026 baseline: over 300,000 agricultural drones, over 460 million mu treated annually, overall crop mechanisation rate 76.7%. The 15th Five-Year-Plan target is 350,000 drones and 3.5 billion mu-times per year — a 17% fleet increase against a 7.6-fold increase in annual operating area, implying per-drone annual use rising from roughly 1,500 to 10,000 mu-times. NDRC projects hilly and mountainous-region drone numbers doubling to about 200,000 by 2030. These are government figures with a defined scope; DJI’s global totals have neither.

Product line (2025–2026). China launch 18 November 2025: T100S, T70S, T55 at first-release prices of CNY 40,999 / 35,999 / 28,999. T100S: 149.9 kg maximum take-off weight, 85 L spray, 90 kg spread, 95 kg lift. T55: 50 L spray, 55 kg spread, 40 kg lift. Global launch 1 July 2026: T55 and the T100 dual-battery spraying system (90 L, 80 kg lift). DJI’s stated 2026 posture is safety-led: laser and millimetre-wave radar, quad-vision, and operator principles of “certified to work, graded by task” (持证上岗、分级作业).

Operator certification — the corpus’s earlier framing was wrong. Agricultural drone pilots in China are not CAAC-licensed by default. CAAC’s Civil Unmanned Aircraft Pilot Management Rules (AC-61-FS-2018-20R3) allows a plant-protection operator to hold either a Class V pilot licence or a certificate issued by a manufacturer authorised by MARA under its own training and examination scheme. That carve-out was formalised by the joint CAAC–MARA 《农用无人驾驶航空器操控员培训管理规定（试行）》 (10 February 2025). Flying your own farm requires no aviation licence at all; operating units (companies, cooperatives) need an operating certificate. So DJI’s “700,000 trained operators” are manufacturer-issued certificate holders, not state-licensed pilots, and the two are not interchangeable.

Independent evidence on effectiveness — specific and positive. Li, Ma, Shen & Li, Food Policy 136 (October 2025), a study of 520 rice growers in Yunnan: UAV adoption cut pesticide expenditure by CNY 77.3 per mu and raised rice yield by 196.9 kg per mu. This is the corpus’s best peer-reviewed evidence that Chinese agricultural drones deliver, and it is narrower than any vendor claim.

Competitors — verified, replacing corpus errors. Beyond DJI and XAG: EAVISION / 苏州极目机器人 (Suzhou, founded 2016; EA-J150/J100/J70, EA-30XP; exporting to Brazil and Latin America); 无锡汉和航空 (Wuxi Hanhe), which built China’s first commercial plant-protection drone (the 3CD-10, exhibited 2010); 深圳天鹰兄弟 (Shenzhen Tianying Brothers, est. 2015); 珠海羽人农业航空; 安阳全丰航空植保; plus traditional machinery majors entering (潍柴雷沃, 第一拖拉机). A 2022 peer-reviewed industry survey lists DJI, XAG, Hanhe, Anyang Quanfeng and Tianying Brothers as the main manufacturers. The names “TTA” and “Anyvision” recorded as Chinese agri-drone competitors in earlier corpus material do not exist as such actors — the plausible near-namesakes are 天鹰兄弟 (Tianying) and EAVISION (极目). No purge was needed in the current corpus text, and this note prevents the error reappearing.

Data privacy posture (unchanged, and substantive). No flight logs, photos or videos sync to DJI by default; data is stored locally on the remote controller and aircraft; operators opt in to share; non-China-mainland user data is housed on US, Japanese and European servers under TLS; AES-256 for personal data; third-party security audits by Booz Allen Hamilton, FTI Consulting and Kivu Consulting since 2017.

What this unit is doing in the taxonomy

Anchors the robotics — aerial × on-farm-production-open-field cell with the globally largest fleet, and now also anchors the corpus’s verification-posture thread: it is the canonical V0 case adjacent to XAG’s V2, in the same market and the same country.

Distinct from:

• xag-china-drone-leader.md — the regulated-disclosure counterpart; use XAG for drone economics and DJI for global deployment reach, and note which of the two a given number came from.
• china-agricultural-drone-drift-liability.md — the harm and governance layer.
• naio-technologies.md — French ground robotics, specialty crops.
• john-deere-see-and-spray.md — US vision-driven ground spraying, equipment-integration model.

Why it matters for talks

• The reconciliation failure is the story. A company that publishes two contradictory environmental totals five weeks apart, with no methodology, is the corpus’s clearest demonstration of why vendor-scale figures need a verification column. Use DJI as the example when explaining the V-tier.
• MARA’s baseline, not DJI’s, is the number to quote for China — over 300,000 drones, 460 million mu, from the ministry, with the 15th-FYP target of 350,000 units and 3.5 billion mu-times.
• Per-drone utilisation is the hidden assumption in the 15th-FYP target. Going from ~1,500 to ~10,000 mu-times per drone per year implies a shift from occasional to near-continuous operation — an operational claim worth naming rather than accepting.
• “700,000 trained operators” is not a licensed-pilot figure. China certifies agricultural drone operators through manufacturers, not the aviation authority; this makes the Chinese drone workforce structurally different from every other jurisdiction in the corpus.
• The one peer-reviewed Chinese result (−CNY 77.3/mu pesticide, +196.9 kg/mu rice) is small, specific and more useful than any global total.

Critical context

• Everything in the environmental table is vendor-stated and mutually inconsistent. maturity-verification: V0 is a statement about the claims, not an accusation.
• Cumulative sales are not an installed base. >700,000 cumulative units sold says nothing about how many are flying, in which country, or how often — the same conflation the sector’s operator figures suffer from.
• Geopolitics now shapes the market. FCC “Covered List” restrictions have blocked 25 DJI launches (about USD 1.5 bn at stake) and FAA waivers still gate spray operations in the US, while Brazil’s RBAC 100 (June 2026) and simplifications in Argentina, Spain, Switzerland and Canada opened other markets. Market access is a regulatory variable, not a merit-based one.
• No independent employment figure exists for Chinese agricultural drone work. DJI’s “approaching 500,000 people in flying-defence services” (2024) and “700,000 certified operators” mix trained, active and service-adjacent populations; MARA publishes fleet size, not employment.
• The 2022-2024 corpus case studies (Brazilian coffee cost reduction, Romanian vineyard chemical reduction) are vendor-reported and remain useful only as illustrations, not as measurements.