S13.5 · Transportation, Logistics & Mobility
Six fixed-infrastructure Class I railroads using predictive maintenance and scheduling AI to improve operating ratio, not market structure.
US Class I railroads generate approximately $80B in aggregate revenue (Association of American Railroads data) across six carriers that together control the vast majority of US freight rail ton-miles — Union Pacific at roughly $24B, BNSF at roughly $23B, CSX at roughly $14B, Norfolk Southern at roughly $12B, plus a combined operator running a single North American network across the US, Canada and Mexico. No segment in this sector has higher barriers to entry: the infrastructure is fixed and the approvals are regulatory. That frames what AI can be here — a margin tool working a fixed asset base, not a force reshaping who competes.
The growth story is a mix shift, not a volume story. Overall volume growth is low-single-digit, with intermodal traffic the primary growth vector against flat-to-declining coal and merchandise carload volume — a shift that favors carriers with strong intermodal terminal networks over those still weighted toward legacy bulk commodities. North America is the concentrated developed market, and a single-line network now spans the US, Canada and Mexico, giving shippers continuous cross-border service without an interchange handoff. For nearshoring-driven Mexico-US freight, that is the structural advantage pulling volume onto rail rather than truck.
For the six Class I carriers, barriers to new entry are extreme: rights-of-way are fixed and effectively non-replicable, and Surface Transportation Board oversight gates rates and service commitments. Once scale is reached, the economics are high-fixed-cost and high-margin — operating ratios in the 60s are typical — supported by long-duration shipper contracts. Upstream sit locomotive and rolling-stock manufacturers and rail-car lessors; downstream, bulk shippers directly, and intermodal and drayage trucking operations at the terminal interface.
Both of AI's primary levers in rail — predictive maintenance and network-scheduling optimization — work the same target: operating ratio. Better maintenance scheduling reduces unplanned downtime and asset-failure cost. Better network scheduling improves utilization across a fixed set of locomotives, cars and track. Every dollar either lever produces drops to margin on an unchanged revenue line, and neither touches the fixed-infrastructure moat that defines competitive position — that moat is a function of owning the rights-of-way, not of software.
Of the two strategic builds in motion, continued single-line cross-border service captures freight that would otherwise require a truck-to-rail-to-truck handoff or move by truck entirely — genuinely new volume for rail. Precision-scheduled railroading investment defends rail's cost position against trucking competition on lanes where the two modes overlap. Both are already under way, alongside predictive-maintenance deployment across the major carriers. Rail's AI adoption is mature relative to the rest of the sector, and for a structural reason: a fixed-asset base at this scale rewards marginal efficiency gains more than almost any other business model does.