The main federal AI in agriculture statistics come from the USDA and the US Government Accountability Office (GAO), and they show uneven adoption. According to GAO, citing 2023 USDA reporting, only 27 percent of US farms or ranches used precision agriculture practices, yet USDA data show automated guidance on well over half the acreage of six major crops. Market forecasts for agricultural AI are estimates, and two current ones differ by more than a factor of two for the same year. The savings GAO reports come from cutting inputs such as herbicide and water, and the barriers it names are cost, data ownership and missing standards.
AI in agriculture statistics: US adoption by technology and crop
The table below collects the adoption figures with a named source and year. Read each one against its denominator: farms, acres or herd size.
| Measure | Figure | Source and year |
|---|---|---|
| US farms or ranches using precision agriculture practices | 27% | GAO 2024, citing 2023 USDA reporting |
| Automated guidance (auto-steer) | Well over 50% of planted acreage for corn, cotton, rice, sorghum, soybeans and winter wheat | USDA ERS, ARMS data 1996–2019 |
| Yield maps, soil maps and variable rate technology (VRT) on winter wheat, cotton, sorghum and rice | 5–25% of planted acreage | USDA ERS |
| Yield maps, soil maps and VRT on corn and soybeans | Substantial adoption for many years | USDA ERS |
| Dairy farms using automatic takeoffs | 49.1% overall; 93.5% of farms over 500 cows; 19.9% under 100 cows | USDA 2014 data, via GAO |
| Computerised milking systems, farms with 500–1,999 cows | Almost half | USDA 2016 data, via GAO |
| CNH farmer and rancher customers (US and Canada) using AI capabilities such as automatic equipment adjustment | 14% of 217 surveyed | CNH survey, May 2026, reported by Mordor Intelligence |
| Farms reporting internet access | 85% | NASS, August 2023, via FCC 2024 |
The 27 percent and the “over 50 percent” figures differ in denominator (farms versus acres), in technology (any precision practice versus automated guidance on six crops) and in year (2023 versus data to 2019). USDA counted 2 million US farms in 2022: farms producing up to $100,000 in goods were about 81 percent of farms but only about 30 percent of farmland, while farms producing $500,000 or more were about 7 percent of farms and 41 percent of farmland. A technology that large farms adopt can therefore cover a far larger share of acres than of farms.
Herd size works the same way in dairy: automatic takeoffs are nearly universal above 500 cows and uncommon below 100. Regionally, GAO reports higher adoption in the Midwest, and the top five adopting states (North Dakota, Nebraska, Iowa, South Dakota and Illinois) accounted for 52.6 percent of 2022 US corn cash receipts and 45.7 percent of soybean receipts.
One caveat matters more than the rest. USDA’s Agricultural Resource Management Survey (ARMS) mainly tracks yield monitors, VRT, GPS-enabled technologies and mapping, and added drone questions in 2015. The FCC’s 2024 precision agriculture task force report says the FCC, NTIA and USDA should include consistent reporting metrics for precision agriculture adoption. The CNH figure above comes from a survey of 217 of the company’s own customers, not from a federal statistic.
How is AI used in agriculture today, from precision farming to automation?
AI shows up in agriculture as an analysis technique added to precision agriculture and as computer vision in automated machines. GAO describes precision agriculture as collecting data tied to a location or time and using it to apply water, fertiliser and feed more precisely, and notes that the 1990s package of GPS soil sampling, variable rate fertiliser and yield monitoring has since moved to new sensors and techniques such as AI. A 2021 PRISMA review of machine learning in agriculture (Benos et al., Sensors) organised the 2018–2020 journal literature into four areas: crop, water, soil and livestock management.
Machine learning also drives farm automation. GAO explains that these systems rely on computer vision: sensors collect data, onboard computers analyse it, and the machine makes real-time decisions without a person involved. The examples GAO gives are targeted spraying, automated mechanical weeding and autonomous vehicles, each available for both field and specialty crops. Our piece on computer vision in agricultural machinery covers where the cameras and models sit on that equipment, and a broader explainer looks at how computer vision is helpful in agriculture outside machinery.
Which AI technologies are the most widely used? It depends on what you count. By acreage, automated guidance, a GPS technology, covers well over half the planted acreage of six major crops. By spending, two market research firms independently put machine learning at about 47–48 percent of the agricultural AI market in 2026 (Mordor Intelligence: 47.98 percent; Future Market Insights: 47.0 percent), with Mordor forecasting computer vision as the fastest-growing technology segment at 22.2 percent CAGR over 2027–2032.
Two market estimates that disagree on size
Market size figures for AI in agriculture come from commercial research firms, and the two current ones do not agree. Mordor Intelligence estimates USD 3.11 billion in 2026, rising to USD 3.79 billion in 2027 and USD 10.02 billion by 2032, a 21.46 percent CAGR over 2027–2032. Future Market Insights values the market at USD 5.9 billion in 2025 and USD 7.5 billion by the end of 2026, growing at 26.3 percent CAGR to about USD 77 billion by 2036. For 2026 that puts one estimate at roughly 2.4 times the other (7.5 ÷ 3.11). An older Statista forecast, cited in a 2022 Frontiers in Artificial Intelligence review, projected growth from $1.1 billion in 2019 to more than $3.8 billion by 2024.
The two firms also define the market’s scope differently. Mordor counts annual spending on AI products across farm, livestock, drone, greenhouse and supply-chain operations, and excludes generic machinery and sensors without an AI layer. Future Market Insights covers hardware, software and integrated solutions spanning precision farming, crop monitoring, livestock, predictive analytics and autonomous equipment. Their shares for precision farming also differ: 43.07 percent against 34.0 percent. Treat both as evidence that spending is growing quickly, not as a size you can plan around.
What are the main benefits of AI in agriculture for farmers and agribusinesses?
GAO’s 2024 report names the main benefits as higher profits (more yield from the same inputs, or the same yield from fewer) and less fertiliser, herbicide, fuel and water. GAO adds that precise application can prevent excess chemicals and nutrients in a field, which may reduce runoff into soil and waterways. Stakeholders also told GAO that guidance accounting for current weather and crop growth stage can raise irrigation efficiency and cut fertiliser use.
The FCC’s 2024 report describes a farm where a network cut operating costs by 20 percent and water use by 10 percent in the first month. That is one farm over one month, useful as an example and nothing more.
The FCC report says autonomous vehicles could work day and night and may reduce fuel, labour, repair and maintenance costs. The Frontiers review points to predictive maintenance models that recommend servicing equipment before it fails, which can reduce unplanned downtime.
How does AI improve pesticide use?
Targeted spraying applies herbicide only where a weed is detected. GAO describes the mechanism: cameras on the sprayer capture images as it moves, an onboard computer runs a machine learning model to detect weeds, and a nozzle fires a small dose only at that spot. GAO also notes that a targeted sprayer that came out of USDA Agricultural Research Service work saves farmers time and money because they refill their tanks less often.
John Deere’s See & Spray has published acreage figures. AgWeb reported in December 2024 that Deere said the system ran across more than 1 million acres in the 2024 season, with an estimated saving of 8 million gallons of herbicide mix. That figure is Deere’s own estimate, and it counts herbicide mix. Mordor Intelligence reports that Deere put the 2025 figure at more than 5 million acres across its See & Spray tiers.
The same AgWeb report sets out the limits: in corn, soybeans and cotton the crop can be at most 16 inches tall for effective application, and tillage, crop rotation and camera maintenance all affect performance. Deere also ran a webinar on the practical problems of switching from broadcast to targeted spraying.
Automated mechanical weeding is another route. GAO describes automated mechanical weeders that use computer vision to tell crops from weeds and start and stop blades to avoid the crop, as an alternative to herbicides in specialty crops. Carbon Robotics says its LaserWeeder G2 uses a model trained on more than 40 million labelled plants from three continents, and claims a payback period of 1–3 years on equipment with a 7–10 year lifetime. Those are the vendor’s own claims.
What are some real-world examples of AI being used on farms?
Several deployment figures are reported by the companies running the systems. The list below gives each one as reported, with the year, as compiled by Mordor Intelligence unless stated otherwise.
- Bayer Climate FieldView: more than 275 million subscribed acres in its main markets in 2025.
- Syngenta Cropwise: 14.8 million hectares monitored in Brazil, reported in August 2026, with satellite imagery and AI analysis turned into crop-health insights.
- TerraClear: machine learning rock maps delivered across more than 1.5 million acres to over 1,000 US farm operations by August 2026.
- OPTiM (Japan): AI-timed drone spraying across about 32,000 hectares and 130,000 rice fields in 25 prefectures in the FY2025 season.
- FBN Norm AI Ag Advisor: more than 140,000 members, reported in September 2026.
- Haswa Mango Farm (Saudi Arabia): around 50 percent lower irrigation water use in the 2026 season through AI-supported, soil-moisture-based scheduling.
- Automated rotary milking parlours: GAO reports these can handle about 500 cows an hour, compared with about 60 cows a day for an automatic milking system.
A platform can cover millions of acres while the decisions on them still come from the farmer.
How is AI used in livestock farming and herd health monitoring?
In livestock, precision cattle farming aims to identify and count cattle, detect disease early, automate milking and feeding, and manage feed efficiently. The stakes are large. A systematic review in Computers and Electronics in Agriculture notes that cattle accounted for 18 percent ($66.2 billion) of US agricultural cash receipts in 2019. The FCC’s 2024 report describes dairy operations combining computer vision, AI, and edge and cloud computing to track weight gain, feed efficiency and animal health, and to plan breeding.
A 2025 study cites a systematic review estimating that 23–25 percent of dairy cows worldwide have clinical or sub-clinical lameness at any given time, which lowers milk yield, delays reproduction and raises culling. The study’s 3D convolutional network classified lame versus normal cattle from video with 90 percent accuracy. The dataset was 50 videos of 42 cattle.
Our article on AI in livestock management and precision livestock farming explains how the cameras, sensors and edge models work.
Does AI in agriculture require constant internet connectivity?
No, not for every task. Some agricultural AI runs on the machine itself, but some uses depend on reliable, field-wide connectivity. GAO describes targeted sprayers and weeders that process camera images on an onboard computer and act in real time, and Mohanty et al. (2016) reported that their trained disease classifier ran in under a second on a CPU, fast enough for a smartphone.
The FCC’s 2024 task force sorts farm connectivity into three profiles: real-time heavy processing such as AI and autonomous machines, which need high bandwidth and low latency; real-time telematics for machinery faults or livestock welfare; and asynchronous bulk transfers, such as drone field maps used to plan the next few days. Autonomous tractors need seamless, field-wide connectivity. Autonomous tractors stop at the first abnormality and send an image of the problem, and a connectivity lapse can mean the tractor has to be driven or moved somewhere with signal before it can restart.
Camera-based livestock monitoring shows the data problem. The FCC report estimates that 3D imaging of a 200-cow dairy herd can generate up to 130 TB a day, about 300 GB with edge computing compression, and about 1 GB with edge computing, which still takes more than 20 minutes to upload on a typical 7 Mbps farm connection. The report also notes that 85 percent of farms have internet access but that this falls well short of coverage across working land, and that some cellular “coverage” cannot sustain more than 2 Mbps down. Our article on edge AI deployment trade-offs covers deciding what runs at the edge and what has to leave the site.
What are the limitations and disadvantages of AI in agriculture?
One technical limitation is that accuracy reported in controlled conditions does not carry over to the field. Mohanty et al. (2016) trained a model on 54,306 controlled-condition leaf images that reached 99.35 percent accuracy on a held-out test set, but scored about 31 percent on images taken under different conditions. Why that happens, and what it means for deployments, is covered in our article on where AI in agriculture works and where it stalls. Later work reported a best generalisation accuracy of 81.60 percent with background removal, and 77.50 to 80.33 percent when field images were added. That is a large improvement, but still well below the 99.35 percent lab figure.
The Frontiers review points out that a wrong soil-nutrition recommendation can lead to the wrong fertiliser, lost product and financial losses, which raises accountability problems. The same review cites an argument that most farmers are unaware they use AI-enabled technology, and reports that some providers have made farm data collection a condition of service.
Data trust is also a barrier. The Frontiers review cites Stock et al. finding that 78 percent of farmers worry about corporations sharing and selling their data without consent. A 2024 Ag Data Transparent survey found that farmers overwhelmingly believe they own their data but are less sure who can access it or how it is stored, and only a third know who to ask for help using it.
What stops a working farm from adopting AI?
GAO lists three challenges limiting broader adoption: acquisition costs prohibitive for farmers with limited capital, concerns about farm data sharing and ownership, and a lack of uniform standards hampering interoperability. Mordor Intelligence reports similar findings in Europe: an April 2026 European Commission report on interoperability and vendor lock-in, and a 2025 EU CAP Network report on AI-robotics costs and uncertain returns for small and medium farms.
Complexity also slows adoption. Stakeholders told GAO that easy-to-use technologies spread faster, and that data-intensive tools, which need farmers to collect, analyse and act on data, face a higher barrier. That is the stated reason auto-guidance outpaces VRT: guidance responds to GPS with no farmer input, while VRT needs variability maps built and acted on. GAO also reports that younger farmers adopt faster.
Connectivity is another barrier. The FCC task force says precision agriculture needs a “last-acre” wireless network across the farm as well as a last-mile connection to it, and that any last-acre option will be expensive enough to hold back small and mid-sized operations. Federal support exists: GAO counts $253.6 million in NRCS conservation assistance for practices that may use precision agriculture, fiscal years 2017–2021.
Taken together, these barriers weigh most on smaller, capital-limited farms and on data-intensive tools. For teams building the models, edge inference and deployment pipelines behind such tools, TechnoLynx sets out its engineering approach on its page on AI engineering for agriculture.
Reading the next agriculture AI statistic
Federal surveys track GPS, yield monitors and VRT, market firms define the market’s scope differently, and some large deployment figures are company estimates. Until survey metrics catch up, ask three things of any new number: what the denominator is (farms, acres or animals), who measured it, and whether the accuracy was measured in a field or in a lab.
FAQ
What are some leading agricultural AI companies?
Future Market Insights groups the market into equipment manufacturers adding AI, dedicated agri-tech software firms, large technology companies entering agriculture, and specialist startups. Named deployments in the sources include John Deere (See & Spray), Bayer (Climate FieldView), Syngenta (Cropwise) and CNH.
