The Cost of Running the Train
What it costs to run a high-speed corridor every year — and the ridership it would take to pay for it.
The debate over a high-speed corridor usually fixes on the construction price tag. But a corridor that is built still has to be run — maintained, staffed, energised, and periodically re-equipped — for as long as it operates. That recurring cost is a separate question from the capital cost, and it is answered by a separate methodology.
This brief sets out that methodology in three parts: the cost of keeping the fixed assets in service, the cost of running trains on them, and the cost of replacing the trains when they wear out. It then asks the single question those three costs raise together: how many passengers would the corridor need to carry to cover them?
For a 1,000 km dedicated high-speed corridor under Canadian operating conditions, the three recurring cost streams sum to approximately $2.15 billion per year at baseline service. To cover that from fare revenue at the modelled fare and load factor, the corridor would need to carry approximately 12.5 million passengers per year. At the modelled baseline service level, fare revenue recovers only 80 per cent of recurring cost — a $439 million annual deficit, incurred before a single dollar of construction debt is serviced.
This brief builds each of the three cost streams from international benchmarks, stacks them, and derives the break-even ridership. The point is not a verdict on the project. It is to give the reader a structure for testing any published operating-cost or ridership claim against the arithmetic that governs it.
Three cost streams, three different shapes
Recurring lifecycle cost is not one number. It is three streams with fundamentally different drivers, and they respond to traffic in opposite ways. Modelling them as a single line item — the common “O&M” or “lifecycle cost” figure — hides the structure that decides whether cost recovery is achievable at all.
The first two streams have opposite sensitivity to traffic. Maintenance is dominated by the cost of having the assets there at all: patrol, inspection, and age-based renewal continue whether eighty trains run or two hundred. Operations is dominated by the cost of activity: more trains mean more crew-hours, more energy, more servicing. The third stream, fleet capital, is set by the size of the fleet needed to deliver peak service — it does not scale with utilisation at all.
This opposite-shape structure is why a single bundled cost figure cannot be audited. A reader given only a total cannot tell how much of it is fixed — and the fixed share is precisely what determines how the cost behaves as ridership changes.
The cost of keeping the assets in service
Infrastructure maintenance has two parts that must be modelled separately. Routine maintenance is annual recurring spend on inspection and preventive and corrective work. Renewal is the periodic capital replacement of long-life components — rail, ballast, contact wire, signalling electronics — annuitised over each asset’s useful life. Conflating the two is the most common business-case error in long-life infrastructure analysis; omitting the renewal annuity understates real lifecycle cost by 40 to 60 per cent.
Applied to the worked example — a 1,000 km dedicated double-track corridor at 300 km/h, under an Eastern Canadian climate-and-terrain uplift of 1.375 — the maintenance-plus-renewal total is approximately $1.27 billion per year, or $1.27 million per route-kilometre. Stripping the Canadian uplift leaves an underlying figure of about $920k per route-km, which sits at the top end of the European HSR range — the appropriate position given Canadian labour rates and the absence of a domestic HSR supply chain.
The structurally important fact is the fixed-cost floor. About $980 million of the annual total is incurred regardless of how many trains run. No ridership scenario reduces it. This is the single most important number for the alternative-framework comparison: a corridor that already exists and is already being maintained for other traffic does not add a fresh fixed-cost floor of this size merely because passenger services are layered onto it.
The cost of running the trains
Operating cost decomposes into eight categories. Three — traincrew, traction energy, and rolling-stock light and intermediate servicing — scale directly with train-kilometres. Three — station operations, network control, and insurance — are largely fixed. One (commercial) scales with revenue, and one (general and administrative overhead) is applied as a markup on direct costs. Where infrastructure is dominated by the existence of assets, operations is dominated by the act of running trains.
At the baseline 80 trains per day, total operating cost is approximately $700 million per year, or $24 per train-km after an Ontario-grid climate uplift. Three categories — traincrew, rolling-stock servicing, and station operations — account for just over half the total. Any cost-reduction strategy that does not touch those three addresses only half of operating cost.
Two findings cut against common assumptions. Energy is small: traction power is only about 6 per cent of operating cost, so grid decarbonisation or efficiency gains will not materially move the operating line — the environmental argument for high-speed rail rests on modal shift and embodied emissions, not on operating-energy savings. And stations are the largest fixed line: at roughly $18 million per staffed station per year, each additional intermediate stop adds about that much to the fixed-cost floor regardless of how many trains call there. Station-count decisions are not cost-free.
The alternative-framework comparison matters less here than it does for maintenance. Operating cost per train-km is largely independent of whether the corridor is dedicated high-speed track or shared with other services — so the structural cost advantage of the High Performance Rail (HPR) framework lives in the infrastructure line, not the operations line.
Stacking the three — and the break-even it implies
The third stream is the fleet itself. Trainsets retire after 25 to 35 years; the acquisition cost is therefore the first cycle of a recurring recapitalisation. For a 30-trainset fleet at roughly $70 million per set — about $2.1 billion of fleet capital — annuitised over a conservative 25-year life at the Treasury Board reference discount rate, the annual fleet-replacement annuity is approximately $180 million per year. Whether the assumed life is 25 or 35 years moves this by only about 10 per cent; what matters is that the cost exists at all, not the exact horizon.
Summing the three streams at the MID baseline gives the full recurring picture:
Collected into a single function of service frequency, combined cost is approximately $1.38 billion in fixed cost plus $9.6 million per train-per-day. Revenue rises along a different line, set by fare yield, seats, load factor, and corridor length. Whether the two lines cross — and at what passenger volume — is the cost-recovery question.
At the modelled fare yield of $0.20 per passenger-kilometre and a 65 per cent load factor, the lines cross at approximately 12.5 million full-corridor passenger trips per year. Below that ridership, the corridor cannot cover its recurring cost from fares — before any allowance for construction debt.
| Service / metric (MID) | Value |
|---|---|
| Total combined recurring cost (M + O + F) | $2,147M / yr |
| Fare revenue at 80 trains/day ($0.20/pkm, 65% LF) | $1,708M / yr |
| Annual deficit at baseline service | −$439M / yr |
| Cost recovery ratio at baseline | 0.80 |
| Break-even ridership | 12.5M pax / yr |
Including fleet replacement raises the break-even by about 15 per cent — from 10.9 million pax/yr on an operations-and-maintenance-only basis to 12.5 million once the trains themselves are paid for. The effect is mechanical: every dollar added to the fixed-cost floor needs roughly 8.5 cents of additional annual contribution to recover.
It moves sharply with fare and load factor
The 12.5-million figure is not a constant. It depends heavily on two assumptions a business case can set at will unless they are disclosed and benchmarked: the average fare yield, and the average load factor. A modest reduction in either pushes the required ridership up steeply.
| Fare yield ($/pax-km) | LF 55% | LF 65% | LF 75% |
|---|---|---|---|
| $0.15 | 31.4 | 22.9 | 19.1 |
| $0.18 | 18.7 | 15.3 | 13.5 |
| $0.20 (MID baseline) | 14.7 | 12.5 | 11.3 |
| $0.23 | 11.1 | 9.8 | 9.1 |
| $0.26 | 9.0 | 8.1 | 7.6 |
A 25 per cent cut in yield — from $0.20 to $0.15 per passenger-kilometre — nearly doubles the break-even ridership at baseline load factor, from 12.5 to 22.9 million. This matters because $0.20 per passenger-kilometre is already above the European average: SNCF’s TGV and Trenitalia’s Frecciarossa run nearer €0.14 with higher load factors on long-haul routes. A Canadian assumption above the European benchmark requires explicit justification from route economics, demographics, and competing-mode pricing — it cannot simply be asserted.
Why this matters
The international record on rail demand forecasts is not encouraging: across a large sample of projects, nine in ten rail forecasts overestimated ridership, with an average overestimation around 100 per cent in the first decade. A break-even at 12.5 million leaves little margin to absorb that kind of forecasting error — and the margin shrinks further at any fare below the modelled $0.20.
Can the corridor pay to run itself?
At the modelled baseline, no — not from fares alone. The corridor would need to carry roughly 12.5 million passengers a year to cover its recurring cost, and at the baseline service level it recovers only 80 per cent, running a $439 million annual deficit. And this is the easy half of the cost question. Break-even here is computed on recurring lifecycle cost only.
The construction cost has not entered yet. At the proponent’s own $60–90 billion estimate, construction debt service alone would add on the order of $2.5 to $5 billion per year — several times the entire operating-and-maintenance surplus available at any plausible service level. The recurring cost recovers, at best, the cost of running the corridor; it does not begin to recover the cost of building it.
This is not, in itself, an argument against the project. Most large rail systems in the world close their gaps through public subsidy and have done so for over a century. The question the methodology forces is narrower and more answerable: is the recurring cost being disclosed honestly, separated into its three streams, with the fare and load-factor assumptions stated and benchmarked — so that a reader can check whether the ridership forecast clears the break-even the arithmetic requires?
A reader who knows the cost has three streams, knows the fixed-cost floor cannot be reduced by running more trains, and knows where the break-even sits can ask, at every turn, what the missing terms are. That is what this brief is for.
Three questions to ask of any operating-cost claim
Each follows directly from the methodology. None presupposes opposition to any project. Each is the kind of question the arithmetic requires to be answered before a reader can form a judgment.
1. Are the three streams disclosed separately?
Maintenance, operations, and fleet capital have different drivers and opposite sensitivities to traffic. A single bundled “O&M” or “lifecycle cost” figure cannot be audited. In particular: is rolling-stock replacement amortised into the recurring line, or quietly treated as one-time acquisition capital? Omitting it understates recurring cost by around 10 per cent.
2. What fare yield and load factor are assumed?
Both must be stated and benchmarked. A yield above $0.20 per passenger-kilometre sits above the European average and requires demographic, competitive, and route-specific justification. Without these two numbers, a ridership figure cannot be tested against break-even at all.
3. What is the cost-recovery ratio at the central ridership forecast?
Below 1.0, recurring cost cannot be self-funded from fares. Between 1.0 and 1.2 is a thin margin highly exposed to the normal range of forecasting error. And whatever surplus exists above break-even is the only resource available to service construction debt — which is the far larger number.
None of these questions presupposes a view about whether the corridor should be built. Each is the kind of question a reasonable reader would ask before forming one — and each is a question the published materials have so far not been pressed to answer in the terms the arithmetic requires.
The three notes and their evidence base
This brief synthesises the three operating-cost research notes produced by the Initiative. Each is available in full below, with the complete derivations, parameter tables, sensitivity analyses, and diagnostic checklists summarised here.