US specialty-crop weeding robotics — Advanced Farm Technologies (strawberry), FarmWise (Vulcan weeder), Carbon Robotics (LaserWeeder), Harvest CROO (berries)

NA-US (Advanced Farm Technologies Davis CA; FarmWise San Francisco CA; Carbon Robotics Seattle WA; Harvest CROO Plant City FL); Salinas Valley California (the deployment concentration point)

Content

US specialty-crop weeding robotics is the most-concentrated commercial-scale AI cell in US specialty-crop agriculture. Four vendors anchor the cell: Advanced Farm Technologies (strawberry), FarmWise (Vulcan weeder), Carbon Robotics (LaserWeeder), and Harvest CROO (berry prototype). All four are US-headquartered; three are California-based with Florida (Harvest CROO) and Seattle (Carbon Robotics) as additional hubs. The Salinas Valley is the primary geographic concentration point.

This unit is a consolidated cell anchor — distinct from single-vendor units. The cell-level observation is the substantive finding: US specialty-crop weeding AI has reached commercial scale, while US specialty-crop harvest AI has not (per units/sullivan-salinas-valley-specialty-crop-critical.md’s “zero percent” head-lettuce finding).

Advanced Farm Technologies (Davis CA) — strawberry robotics

Founded 2015. First commercial-scale robotic strawberry deployment in the United States, operating in California’s Oxnard and Santa Maria during 2019. The platform uses large autonomous vehicles with four custom-designed gantry arms (two per row of strawberries) to pick strawberries at commercial scale.

2025 academic validation: Ochoa, E., Mo, C. Design and Field Evaluation of an End Effector for Robotic Strawberry Harvesting. MDPI Actuators 14(2):42, 2025 (cited 14 times per MDPI metadata). The paper documents the deployment scale and validates the academic literature on Advanced Farm Technologies as the US strawberry-robotics reference case.

Structural significance: Advanced Farm Technologies is the commercial-scale strawberry deployment reference. The 2019 commercial scale has continued as of 2026.

FarmWise Labs (San Francisco CA) — Vulcan intra-row weeder

Per farmwise.io (verified July 2026):

• Vulcan is FarmWise’s next-generation intra-row weeder and precision cultivator. Designed for durability and precision across day/night operation, varying weed pressure levels, and crop growth stages.
• Performance: Up to 3 acres/hour coverage.
• Modular architecture for service and maintenance simplicity; software updates over the air.
• Vendor-reported value proposition: “Save $250+ per acre”; fewer passes across the field; runs in all field conditions.
• Customer reference: Larry Santos, Senior Farm Manager, Taylor Farms — a major US fresh-produce company.
• Distribution partnership: RDO Equipment Co. (announced 2025) for exclusive distribution in specialty-crop precision weeding.

Strategic significance: FarmWise is the specialty-crop weeding commercial-deployment anchor. The Taylor Farms customer reference is substantive (Taylor Farms is one of the largest US fresh-produce companies); the RDO Equipment distribution partnership gives FarmWise a substantial equipment-dealer footprint.

Carbon Robotics (Seattle WA) — LaserWeeder

Carbon Robotics builds the LaserWeeder — uses computer vision to identify weeds and high-powered lasers to kill them without chemicals. Per public reporting, the LaserWeeder has been deployed across multiple US farms at operational scale; specific acreage and customer references vary by source.

Distinctive design choice: The laser-based approach eliminates herbicide application entirely. This is a different input-reduction proposition than mechanical weeding (FarmWise Vulcan) or chemical reduction (John Deere See & Spray) — Carbon Robotics positions as chemical-free weeding.

Note: Carbon Robotics has not been pulled in depth for this scan; the public reporting on customer base and acreage deployment is fragmented. G-168 (new, per US industry scan) is the specific gap: substantiated customer base and deployment scale data for Carbon Robotics at the same granularity as Deere See & Spray.

Harvest CROO (Plant City FL) — berry prototype

Founded 2013 — the longest-running US specialty-crop robotics effort. Focused on strawberries and berries. Harvest CROO’s berry harvester has been in development and prototype deployment for over a decade.

Structural significance: Harvest CROO is the longevity reference case — a vendor that has worked on the same problem (berry harvest) for over a decade without yet reaching commercial scale comparable to Advanced Farm Technologies strawberry. This longevity itself is a structural finding: even with substantial sustained effort, US berry harvest AI has not yet matched the commercial deployment of US strawberry weeding AI.

Differentiation from Advanced Farm Technologies: Advanced Farm Technologies focuses on strawberry picking (commercial scale 2019); Harvest CROO focuses on berry picking (still in prototype per public reporting). Different crops and different deployment models.

The consolidated cell finding

The four vendors together populate the specialty-crop weeding + selective-picking AI cell:

• Weeding at commercial scale: FarmWise Vulcan, Carbon Robotics LaserWeeder, Advanced Farm Technologies (some configurations). Commercial scale achieved.
• Selective picking at commercial scale for strawberry: Advanced Farm Technologies (commercial since 2019). Commercial scale achieved for strawberry.
• Berry harvest at commercial scale: Harvest CROO (still in prototype per public reporting). Commercial scale not yet achieved.

The cell-level observation: weeding and selective strawberry picking have reached commercial scale; berry harvest remains in prototype.

Sullivan counterweight — the structural inversion

Per units/sullivan-salinas-valley-specialty-crop-critical.md:

• “Zero percent” head-lettuce harvest automation (Salinas Valley automation expert, April 2025).
• “Cheaper and more reliable” farmworker labour vs. existing automated harvesters for romaine lettuce.
• “Holy grail” framing is industry-side, indicating industry actors themselves name harvest automation as the hardest problem.

The Sullivan finding inverts the consolidated cell’s commercial-scale success: weeding is automated at commercial scale; harvest is not. This is a substantive distinction worth surfacing in any talk about US specialty-crop AI.

Salinas Valley as deployment concentration point

Per Sullivan: Salinas Valley (Monterey County) accounts for at least 70% of US leafy greens production and just shy of $5 billion in gross specialty crop production value in 2024. Romaine lettuce alone is approximately $1 billion. Strawberry production in adjacent Santa Maria and Oxnard regions is substantial. The geographic concentration of US specialty-crop AI deployment mirrors the geographic concentration of US specialty-crop production.

Comparative technology approaches

Three distinct modality approaches (mechanical picking, mechanical cultivation, laser) within one cell. The cell is not technology-convergent — vendors are taking structurally different technical paths to similar commercial objectives (input reduction, labour substitution, yield optimisation).

Connection to broader US agritech AI

• vs. John Deere See & Spray (units/john-deere-see-and-spray.md): Deere is row-crop locked-in equipment vendor at operational scale (5M+ acres 2025). Specialty-crop weeding is different cell.
• vs. AGCO PTx (units/agco-ptx.md): AGCO is broadacre retrofit; specialty-crop weeding is different cell.
• vs. CNH + Raven (units/cnh-raven-autonomous-portfolio.md): CNH-Raven is broadacre autonomous equipment; specialty-crop weeding is different cell.
• vs. Indigo Ag (units/indigo-ag.md): Indigo was biologicals + digital (pre-restructuring); distinct cell.
• vs. Pivot Bio (units/pivot-bio-proven-g3-biologicals-ml.md): Pivot Bio is biotech + ML discovery; distinct cell.

The specialty-crop weeding cell is a distinct US AI cell that operates in parallel with — but does not converge with — broadacre row-crop AI, input-vendor digital platforms, biotech discovery, or hyperscaler substrate.

What this unit is doing in the taxonomy

Anchors the specialty-crop weeding × commercial-scale AI × US-concentrated deployment cell of the matrix. Distinct from:

• Single-vendor units (Deere, Bayer, Corteva, etc.) — this is a consolidated cell.
• Sullivan critical-voice unit (units/sullivan-salinas-valley-specialty-crop-critical.md) — Sullivan is the counterweight; this unit is the vendor-side cell anchor.
• Harvest CROO (units/sullivan-salinas-valley-specialty-crop-critical.md references Harvest CROO but Harvest CROO is not yet standalone) — Harvest CROO is part of this cell; the cell is the unit.

The consolidated cell anchor pattern is structurally important — it names the cell-level finding (weeding commercial scale, harvest not) rather than treating each vendor as an isolated unit.

Why it matters for talks

• The most-concentrated commercial-scale AI cell in US specialty-crop agriculture. Distinct from broader row-crop AI, input-vendor platforms, biotech discovery, etc.
• The cell-level finding (weeding commercial, harvest not) is the substantive analytical observation — backed by Sullivan ethnography on the harvest side.
• The three distinct modality approaches (mechanical picking, mechanical cultivation, laser) within one cell show the cell is not technology-convergent.
• The geographic concentration in Salinas Valley and adjacent California regions mirrors US specialty-crop production concentration.
• The venture-funded posture is structurally similar to Indigo Ag’s biologicals trajectory — boom-bust risk worth flagging.

Critical context

• Vendor-reported figures dominate across all four vendors. Acreage and customer-deployment scale varies by source; independent verification at Deere See & Spray-equivalent granularity not surfaced.
• Carbon Robotics customer base and acreage data are fragmented — G-168 specifically named this gap in the US industry scan.
• Harvest CROO’s decade-plus prototype phase suggests the commercial-scale trajectory for berry harvest is harder than the strawberry picking deployment scale — worth tracking as a longevity-counter signal.
• Salinas Valley labour-context dependency: California’s overtime regulation (2024), H-2A worker costs, and farmworker labour availability drive the deployment urgency. Sullivan documents the labour context in detail.
• The Sullivan ethnography is the substantive critical voice — vendor and investor framing should be cited alongside Sullivan for any talk that addresses this cell.
• IP and right-to-repair implications: Each vendor holds proprietary ML models; farmer-data flows to vendor via cloud-mediated telemetry. Field-guide hygiene applies — cite the figure, name the source, flag the integration.