S10.1 · Food, Agriculture & Beverage
The base commodity-production layer of the food system, valued on land appreciation more than crop income, where AI is eroding the local agronomist's advisory role.
US farmland carries a value of roughly $3.4T (USDA/American Farm Bureau Federation, 2025 land values report), sitting on a global arable base of roughly 1.2-1.4B hectares (FAO). The operating economics are almost beside the point. Crop margins are thin, ownership is fragmented even where the farming itself runs at industrial scale, and the return that has justified capital here is land appreciation, not farm income. The live question is different: satellite imagery and AI recommendation engines are taking over the scouting and diagnosis work that used to require a paid visit, and the local agronomist's advisory franchise is eroding with it.
Values compounded at a double-digit annual clip through 2020-2022 and have since flattened to low-single-digit gains, with cash rents near record highs (AFBF, 2025). Farm income has softened as large 2025/26 corn and soybean crops press on prices (USDA WASDE, November 2025). Production concentrates in the US Midwest, the Brazilian Cerrado, the Argentine Pampas, the Black Sea region and the EU. Ownership does not concentrate anywhere. Family operators dominate, even on farms big enough to run as industrial operations, and the listed farmland REITs — Farmland Partners and Gladstone Land are the visible ones — hold well under 1% of US farmland between them. No scaled institutional aggregator exists; the ownership base is an extreme long tail.
The segment buys from seed, chemical, fertilizer and equipment suppliers and sells into grain elevators, traders and processors. What separates it from nearly everything else in the sector is where the return sits: the asset being held matters more than the annual operating result. Capital intensity is high, and the earnings line is hostage, year to year, to weather and to trade policy — export restrictions, tariffs and biofuel mandates move commodity prices in ways no operator controls. USDA subsidy programs cushion the downside without changing the underlying volatility.
Operators are responding to margin compression in two ways that should not be read as one. Aggregating acres — for scale and for the data access that comes with running more of them — defends: a bigger operation absorbs input-cost inflation better and spreads the fixed cost of precision-agriculture tooling, but it adds no new dollars to the system. Bolting on input distribution does add dollars to the operator, though they are reallocated dollars — margin pulled away from the cooperative or dealer that used to earn it.
The structural AI effect lands on advisory labor. Satellite and drone imagery paired with AI recommendation engines now flag nutrient deficiencies, pest pressure and irrigation needs at close to zero incremental cost — work that used to require a paid site visit from a cooperative agronomist. Routine advice reprices toward zero, and the agronomist's earning power migrates to what the model cannot do: unusual weather events, novel pest outbreaks, decisions with real financial stakes attached. Data-enabled scouting and recommendation tools are already in commercial use across large operations. Full operational autonomy — equipment making both the diagnostic and the execution decision — remains 5-10 years out, gated by the capital cost of autonomous machinery and by regulatory approval for unsupervised equipment operation in the field.