CSX Completes $495M Baltimore Howard Street Tunnel Expansion, Unlocking Double-Stack Container Rail Service

The Analysis

Spread the map of the eastern seaboard out on the table and find Baltimore, then follow the water inland until the city closes over it. For a hundred and thirty-one years there has been a freight tunnel under those streets, a mile and seven tenths of brick and rock and soot, and for most of that time it did one job and did it badly enough that everyone with a container and a railcar learned to work around it. On 24 June 2026 the state's governor stood at a ribbon-cutting in front of it and the reconstructed bore was declared done, at a cost of four hundred and ninety-five million dollars, carrying eighteen additional inches of vertical clearance and a claim that took nine years to become true.

Eighteen inches. That is the entire number, and I want to resist the temptation to laugh at it. The project also raised clearance at twenty-one other locations across Maryland, Delaware and Pennsylvania, which tells you the tunnel was never the only pinch point, only the most famous one. Baltimore can now run full double-stack container rail to Midwestern and East Coast markets, and the port's annual cargo capacity is expected to rise by a hundred and sixty thousand containers. Over thirty years, the projections attached to the project put the savings at a hundred and thirty-seven million gallons of truck fuel and one and two tenths billion truck vehicle miles, which is the kind of number that sounds like a press release until you divide it and find out it is arithmetically honest.

I have watched this corridor for a long time, and the thing I keep thinking about is not the engineering. It is the calendar. CSX walked away from this tunnel project in 2017, saying the investment no longer justified itself. It came back in 2018 with a ninety-one million dollar commitment. Double-stack service through a northern route debuted in 2024. The tunnel reopened in September 2025 with work still outstanding into 2026. The final ribbon was cut in June. Nine years, one tunnel, eighteen inches, and a network that quietly rerouted itself around the obstacle in the meantime. If there is a lesson for anyone signing a freight contract this year, it is that the people who benefit from a project like this are the ones who started aligning their inland plan while the boring machine was still running.

Let me be precise about what changed, because the word double-stack gets used loosely and the looseness costs money. A double-stack train carries two containers stacked one on top of the other on the same well car. The constraint has never been the idea, it has been the space above the rail: tunnels, bridges, overpasses, signal gantries, and the occasional catenary that was hung at a height nobody thought about again for a century. Adding eighteen inches to a bore that is a mile and seven tenths long is not a question of digging more, it is a question of whether a stacked pair now clears every obstacle on the route. The twenty-one additional clearance sites in three states are the honest part of the announcement, because they admit that one tunnel was never going to be enough.

So the physical change is real, and it is not the same thing as a commercial change. The port's annual capacity is expected to rise by a hundred and sixty thousand containers. Divide that by fifty-two and you get roughly three thousand and eighty containers a week. Divide by twelve and it is about thirteen thousand three hundred a month. Those are the units you should carry into your planning conversations, because a headline number of a hundred and sixty thousand sounds like an ocean and three thousand a week sounds like something you can either capture or miss.

Here is a cross-check I like, because it tells you the projections hang together. One and two tenths billion truck vehicle miles over thirty years is forty million a year. If an avoided truck move averages three hundred miles, which is my assumption and not a sourced figure, that is about a hundred and thirty-three thousand truck trips a year taken off the road. Against a hundred and sixty thousand additional containers a year, that works out at roughly one and a fifth containers per avoided truck trip, which is exactly what you would expect if a truck is moving either one forty-foot box or two twenty-foot boxes. The two headline numbers are telling the same story from different ends.

The fuel figure survives the same treatment. A hundred and thirty-seven million gallons over thirty years is about four and a half million gallons a year. Spread across those hundred and thirty-three thousand trips, it is roughly thirty-four gallons per avoided trip, and three hundred miles at thirty-four gallons implies something like eight and a half miles to the gallon. That is in the right neighbourhood for a loaded Class 8 truck on a highway run, under my assumption about the trip length. I am not going to put a dollar figure on a gallon or on a truck-mile, because I do not have sourced numbers and because the point here is not the total. The point is that the environmental case and the capacity case are two views of one physical change, and the physical change is the only part that matters to your plan.

Now translate it into what an importer or exporter actually faces. Baltimore sits closer to the Midwest than almost any other East Coast port by rail mileage, and that was true before the tunnel and remains true after it. What the clearance change alters is not the distance, it is the cost and the rhythm of running boxes over that distance on rail rather than on tyres. For a shipper whose inland points sit inside the double-stack network, the change is a capacity and reliability story: fewer trains to move the same volume, better asset turns, and a rail product that becomes competitive against a long truck move on more lanes than before. For a shipper whose inland points sit outside it, the change is approximately nothing, and that is the distinction that will decide whether this project shows up in your cost line at all.

Let me put a specific case on it, with the assumptions written out. Assume you bring four hundred containers a month into Baltimore and move them inland by a mix of rail and truck to three distribution points, one of which is in the Midwest. Assume that before the clearance work, the rail option to that Midwest point was effectively capped at fifty-five boxes a week for you, because the service existed on a route that did not have full clearance and the operator priced and scheduled accordingly. Assume the new capacity on the corridor is the three thousand and eighty containers a week I derived above, and that your share of it could plausibly be three percent, which is about ninety-two boxes a week.

Under those assumptions, the increment available to you is thirty-seven boxes a week, which is roughly a hundred and nine boxes a month, or about one thousand three hundred a year. Now assume, again my assumption, that each of those boxes avoids a hundred and twenty miles of truck movement because the rail ramps you use are closer to the destination than the alternative. Call the cost of a truck-mile T, with the value left for you to fill from your own contracts. Your annual benefit is one thousand three hundred boxes, times a hundred and twenty miles, times T, which is a hundred and fifty-six thousand truck-miles times T. I am not going to assert T. Multiply it by your own number and you will know within five minutes whether this corridor is worth a pilot or worth ignoring, and you will have done it with arithmetic rather than enthusiasm.

There is a second number in that case that matters more than the saving. Moving a hundred and nine boxes a month off truck and onto rail changes your yard profile, your chassis requirements, and the shape of your appointment schedule at the destination. Those are the costs people forget to book when they chase a transport saving, and on a lot of lanes they eat the entire benefit. Rail is not a truck with more wheels. It is a different operating model with different failure modes, and the failure modes show up at the interface, in the drayage window and in the container availability at the ramp, not on the line haul.

One more word about scale, because it changes how you should read the whole thing. Nobody digs a tunnel for one shipping season. An asset like this is built against a thirty-year demand case, which is why the projections attached to it are stated over thirty years and why the number that matters to the port is annual capacity rather than this quarter's bookings. That mismatch of horizons is not a criticism of anyone. It is simply how infrastructure works, and it tells you something useful: the operator is going to optimise for filling that capacity steadily over years, not for rescuing your peak season. If you want the corridor to work for you in a particular month, you need to be on the schedule as a known quantity well before that month arrives, which is another argument for a volume commitment rather than for buying the lane spot.

Everything so far is written from the importer's side of the dock, and exporters deserve their own paragraph. If you ship out of Baltimore, the clearance change does not automatically help you, but it does change the economics of the equipment your boxes sit in. Inbound volume that arrives doubled up and moves inland on rail releases equipment differently from the same volume trucked out, and where that equipment ends up is largely what decides whether your export booking gets confirmed at the price you were quoted. Watch the balance rather than the headline. A corridor whose inbound economics just improved will, over a few quarters, make export equipment either easier or harder to find in the hinterland depending on which way the flow balances, and you will see it in booking confirmations long before you see it anywhere else. If your export programme depends on boxes being available at an inland point, put that question to your carrier now, in writing, while the corridor is still being rebalanced.

A word on how to run the inside-or-outside test, because the obvious version of it is wrong. Do not ask your rail contact whether a destination is served. Almost every destination is served by something. Ask what the routing is, how many times a week it runs, and whether the stacked pair clears the whole path. A destination that sits nominally on the network but is reached by a route that was not among the twenty-one clearance locations will behave like a single-stack destination no matter what the map says. Then apply a volume filter: if a destination takes fewer than twenty boxes a month from you, the change probably is not worth rebuilding your inland plan around, because the fixed work of switching a lane is larger than the saving. Concentrate on the destinations where the volume is big enough that the avoided truck mileage per box actually moves your annual number.

Which brings me to what I would actually do, and when.

The first job is boring and it is not optional. Before the end of August, whoever owns your North American inland plan should produce a list of your top twenty inland destinations by volume, and against each one a single label: inside the improved double-stack corridor, or outside it. This is not a data science project. It is a phone call to your rail contact and an afternoon with a map. Until you have that list, every conversation about Baltimore is a conversation about someone else's port.

By the end of September, get written confirmation of the clearance and loading limits on the three routings you actually intend to use, from the operator, in writing, for the specific equipment types you ship. Not a brochure. Not a map with a highlighted line. A written statement of what heights and weights and box types the route accepts, and which segments are still subject to outstanding work. The tunnel reopened in September 2025 and there were still projects running into 2026, which means the network was not finished on the day the ribbon was cut, and a written confirmation dated this month is worth more than a general assurance dated last year.

Set the pilot in the fourth quarter of this year, not next year. The capacity is available now, the service is running now, and the value of being early on a corridor like this is that you get the scheduling slots and the ramp relationships before everyone else reads the same press release. Run a hundred boxes a month through it for a full quarter, and measure three things: door-to-door transit against your truck baseline, the number of times a box missed its appointment at the destination, and the dwell at the ramp. Three measures, one quarter, and a written decision at the end of it.

By the end of March next year, commit to a share. If the pilot held up, move at least thirty percent of the eligible Midwest volume onto rail, where eligible means the destinations on the inside-the-corridor list you built in August. If it did not hold up, write down why, in the same document, so that the next person who proposes it in eighteen months does not have to re-learn it. Either outcome is worth more than a corridor you keep meaning to test.

And then the uncomfortable one, about time. This project took nine years from the decision to walk away to the ribbon. Your freight contract is probably twelve months long, maybe twenty-four. That mismatch is the real subject of this piece. A corridor improvement is a ten-year asset; a contract is a one-year instrument; and the shippers who capture these things are the ones who sign multi-year agreements that reference a corridor rather than a rate, with volume commitments that give the operator a reason to protect your slots. If you are negotiating this winter, ask for twenty-four months with a volume band, and ask what happens to your slots in year two. You will learn more from the answer than from any projection.

Now the alternatives and the traps, which is where most of the money actually gets lost.

Changing ports is the obvious move and the most overestimated one. Norfolk, New York and New Jersey, Savannah, Charleston: each of them is a legitimate entry point, and each of them puts your container a different distance from your destination. Baltimore's advantage has always been rail mileage into the Midwest, and the clearance work sharpened that advantage rather than creating it. If your inland points are in the corridor, switching ports to chase a different service is likely to cost you the very thing you came for. If your points are outside it, no amount of switching helps, because the constraint is not the port, it is the last hundred miles.

Changing mode is the real decision and it carries its own traps. Rail gives you a different cost shape and a different risk shape, and the risks are concentrated at the ends. The drayage window at the origin, the availability of chassis and containers at the destination ramp, the appointment discipline at your distribution centre: these are where a rail move fails, and none of them appear in a line-haul quote. Check free time and the per-diem basis on both the box and the chassis before you sign anything, and check what the clock runs from. A generous free-time allowance measured from a gate event you do not control is not generous.

Watch the service commitment wording too. A rail transit figure and a door-to-door commitment are different animals, and the gap between them is exactly the drayage and dwell that I just told you to measure. If a provider quotes you an impressive transit number, ask what the two ends are, and ask whose clock it is. I have seen teams sign up to a rail product on the strength of a line-haul figure and then discover their door-to-door performance was unchanged, because the ends ate the improvement.

There is also the trap of treating the eighteen inches as if it applied everywhere. It applies to a tunnel and to twenty-one other specific locations in three states. It does not apply to your route by contagion. If any segment of the path your boxes take was not in that list, the stacked pair still does not clear it, and your routing will quietly be built on a single-stack assumption while everyone in the meeting says double-stack. This is the version of the mistake that costs the most, because it is invisible until a train is built and your boxes are not on it.

The last trap is the one the calendar teaches. In 2017 the project was judged not worth the money. In 2018 it came back with public partners and a ninety-one million dollar commitment, and the final figure is four hundred and ninety-five million. Projects of this kind get delayed, descoped, resumed and re-scoped, and they are financed by a mix of public and private money that can change with an election. Anyone who builds a single-point plan on a corridor that took nine years to deliver is taking a risk they have not priced. Keep a second routing qualified, even if you never use it, because the cost of a qualified alternative is a phone call and the cost of an unqualified one is a quarter.

Let me close on the picture I started with. A tunnel dug in the eighteen nineties was too low by eighteen inches for the boxes the world eventually decided to ship, and it took a decade, four hundred and ninety-five million dollars, three states and a change of corporate heart to fix it. The port's ceiling went up. The network on the map changed shape at one specific place, and the routes that used to bend around that place will slowly unbend over the next few years. What does not change is the mismatch in clocks: infrastructure moves on a decade, contracts move on a year, and the gap between those two is where the money sits. Ten years ago this corridor was a single-stack story with a bottleneck under a city. Today it is a different drawing. Find out which of your boxes are on it before your competitor does.

↑ Back to the key points

— By Tomás Ferreira