Deere paid $305M for a sprayer that sees, then shipped it as See & Spray

Sprayers historically blanketed whole fields with herbicide because weeds are scattered unpredictably, herbicide-resistant weeds keep spreading, and growers face economic and environmental pressure to cut use. Researchers had chased machine-vision weeding for decades, blocked by computing that was too weak and fields too variable — plant shapes shift, and weeds hide behind crops.

Blue River Technology cracked the loop: its See & Spray equipment analyzes each plant against an image library and squirts herbicide only on actual weeds. The hard engineering shows in the details of later machines: operators tune detection sensitivity to balance missed weeds against off-target spray, and nozzles are inclined 30 degrees rearward to buy time between detection and actuation.

Deere agreed in September 2017 to pay $305 million for the 60-person startup — five months after abandoning a purchase of Monsanto's Precision Planting unit on antitrust concerns — with CTO John May promising to integrate the vision stack across Deere products while the team stayed in Sunnyvale.

The bet became a product line in three tiers — Select for green-on-brown burndown, Premium, and Ultimate — and by 2021–2022 university trials in Arkansas and Mississippi, run on research platforms provided by Blue River's own agronomists, targeted sprays delivered biologically comparable control to broadcast programs, with all treatments reaching at least 98% control by season end while saving herbicide. The startup's founders, in effect, sold Deere an eye for every nozzle it already owned.

See every plant; spray only the weeds

Detection changed the unit of spraying from the field to the plant, so savings scale with how sparse the weeds actually are.

Buying beat building: Deere had manufacturing and distribution, Blue River had the vision stack, and antitrust had just blocked the alternative acquisition route.

The sprayer already rolled over every acre, so adding sight upgraded an existing pass instead of requiring a new operation.

Type I errors were treated as the core design constraint: a missed weed costs less than off-target chemistry, and operators can tune that trade-off.

When a machine already passes every acre, the cheapest upgrade is sight, not strength. Instrumenting an existing routine can beat inventing a new one, and it turns a commodity pass into a data loop.

See & Spray moved from startup demo to Deere's commercial lineup, and targeted spraying became a normal subject of agronomy research: the 2024 Weed Technology study with Blue River agronomists among its authors, a 2025 overview of the technology, and 2025–2026 trials of targeted applications in corn. Adoption-cost analyses have begun asking how much herbicide prices must rise before the machines pay for themselves, which is the usual sign a technology has stopped being a bet and become a line item.