Rail connectivity contracts can run for 25 years, yet the performance figure at their heart is written for the traffic volumes of the year they are signed. Everything that follows in the agreement lifecycle—upgrade pressure, budget exposure and the defensibility of the decision itself—flows from how much room that specification leaves for growth. 

The pattern is systemic. An agency publishes a bandwidth requirement that reflects its current needs; the winning bidder meets it and the service launches successfully. A few years into the contract, that initial requirement has become a hard ceiling. The agency is then left choosing between a slowly degrading service and a second procurement process midway through the contract, complete with the trackside construction, track outages and premium night labor required when working on a live railroad.  

Where per-train demand is heading 

European regulators have conducted some of the most direct modeling of future per-train demand. In 2018, Ofcom, the UK telecommunications regulator, concluded that within 10 years, a busy mainline train could require around 1 Gbps while a crowded commuter train could require two to three. Deutsche Bahn projects a median demand of 1.15 Gbps and an upper limit of 3.6 per 1,000 passengers by 2028. Demand in the U.S. will follow its own trajectory, but the underlying forces are the same everywhere—streaming, video calls and passengers who now expect a train seat to double as a workstation. 

Passenger devices, however, account for only part of the total demand, and the passenger experience is only part of the business case. Safety and security systems rely on the same network. Agencies are increasingly streaming onboard video to the ground in real time so that incidents can be seen and acted upon as they unfold. With artificial intelligence (AI)-enabled analytics also operating onboard, each 4K camera may require roughly 20 Mbps of continuous upload capacity. Ten cameras, therefore, place 200 Mbps of demand on the uplink before a single passenger connects. A specification designed solely around passenger downloads risks leaving insufficient capacity for safety-critical traffic. 

Four things to settle before the ink dries 

1. Treat the published minimum as a starting line: A requirement based on current traffic identifies the bidders who can serve the railroad as it is today. The margin a system can demonstrate above that figure identifies those who can serve the railroad for decades to come. 
2. Test on the railroad and in real-world weather conditions: Paper evaluations tend to group bidders together; live trials separate them. Shortlisted suppliers should conduct trials in the actual corridor before the contract is awarded because operational testing is where the gaps that can shape a 25-year commitment will emerge. 
3. Stress-test the economics alongside the engineering: Site works account for most of the capital cost of a trackside build, far more than the radios themselves, so tower spacing has a significant effect on the budget. Ask how costs will change as data volumes multiply. Per-gigabyte pricing converts rising demand into a rising invoice, whereas owned infrastructure operating on unlicensed spectrum behaves more like a fixed asset. Coordinating deployment with planned construction and using structures that are already being built can reduce the cost further. 
4. Write technology refreshes into the contract: No hardware bought today will last the full term, and radio technology is likely to move through at least one generation. What the agency is purchasing is a long-term partner and a credible upgrade path. Periodic technology renewal should therefore be included in the agreement as a planned line item rather than left to a future negotiation. 

What this looks like in the field 

Caltrain's recent procurement, conducted alongside the electrification of its San Francisco-San Jose corridor, applied these principles almost to the letter. The tender set a modest floor of 100 Mbps to the train and attracted widespread expressions of interest. Rather than award on paper, the California-based commuter line invited two finalists onto the live track in spring 2022. Over a 0.7-mile section in Redwood City, the winning millimeter-wave system sustained around 2 Gbps through heavy rain, 10 times the performance of the competing trackside Wi-Fi bid and some 20 times the published requirement. 

The network now operates across 65 masts. Route testing reported by the project team shows average speeds above 500 Mbps with peaks approaching 1.5 Gbps while observed passenger demand tops out near 300 to 380 Mbps, a margin by design. The agency can lift per-passenger allowances as ridership builds and the service strengthens over the asset's life. The radios ride on masts erected for electrification and connect over fiber already in the ground, which contained both the construction cost and the disruption. 

Field data from the U.K. reads the same way. A 2026 measurement study commissioned by Ofcom found that onboard Wi-Fi on the one British route with trackside high bandwidth backhaul met the regulator's good performance threshold 83.2%of the time. Across the rest of that journey, the figure fell below 1% and the national average for onboard Wi-Fi was 1% . Independent measurement of live deployments on two continents is the grade of evidence a 25-year signature deserves. 

The forecast inside every specification 

Every specification makes a forecast, irrespective of whether its authors intend one. Agencies that accept it pay in installments, first through a service that falls a little further behind expectations each year, then through a construction program to catch up. 

The agencies that will look farsighted a decade from now are the ones asking harder questions before awarding the tender—how much margin above the floor, demonstrated on whose track, at what cost when volumes double and on what schedule it will be renewed. The technology supplying the headroom will keep changing, but the discipline of demanding it should remain constant.