Toyota expands Texas Tacoma plant $3.6B; Ford, GM sign Micron memory deals

The Analysis

Two announcements landed in the same week and look unrelated until you put them on the same dock. Toyota is putting $3.6 billion into its San Antonio plant and moving some US-bound Tacoma output back from Mexico. Ford and GM signed memory-supply agreements with Micron to lock chips and storage for next-generation vehicles while semiconductor supply stays tight.

One of those is about a plant that has to receive parts. The other is about a component that weighs almost nothing and can stop a line inside a shift. Both come down to the same question, and it is not how much money is moving. It is where the queue forms.

A plant expansion is a yard project before it is a building project. That is the part the press release does not photograph.

Start with what physically crosses. A finished Tacoma coming north from Mexico rides an auto rack. One train, one crossing, hundreds of units, one set of paperwork and one queue at the border. A Tacoma built in Texas crosses the border as parts, and parts do not ride in one trailer. They ride in many trailers, on many days, each with its own entry and its own place in the line at the Laredo crossing. Moving output north does not reduce the number of border crossings. It changes the unit of crossing from a train to a trailer, and it multiplies the count.

That is the first thing to get straight, because it inverts the usual assumption. People hear that production is coming back to the US and picture less border friction. What actually happens is more frequent, smaller crossings, and frequency is what a border is bad at. A rail crossing is scheduled and consolidated. A trailer crossing is an appointment, a queue, and a driver sitting in it.

A plant's throughput is set at the inbound dock, not on the line. The line can be speeded up with a shift. A dock door cannot. A door takes a trailer, the trailer gets unloaded, the door turns, and the turn time is set by how the freight was packed and what the receiving system does with it. A yard spot is the same story: it is a physical rectangle, and when it is full, trailers park on the access road, and when they park on the access road, the gate backs up, and when the gate backs up, the appointments slide, and the slide is what everybody downstream calls a supply problem. Three billion dollars buys a lot of building. It does not buy a fourth dimension of yard.

Now put the Micron deal next to it, because the contrast is the useful part. Memory is small, light and expensive. Freight cost is a rounding error against the value, which is why that material moves expedited rather than cheap, and why its binding constraint is allocation rather than transport. Nobody is short of airplane space for memory. Somebody is short of memory. That is a completely different kind of shortage from a trailer shortage, and it fails on a different clock: a missing trailer costs you a day of production, a missing allocation stops the line in a single shift, because the buffer at the line for sequenced parts is measured in hours of build, not weeks of inventory. I am not quoting a number from either announcement here. That is just how an assembly line works, and it is why the two announcements sitting side by side are a study in mismatched clocks.

The layering matters more than the total. Tier-one suppliers clustered near San Antonio get the volume and inherit the appointment problem, because their trailers now compete with everybody else's for the same doors. Small suppliers shipping less than truckload lose dock windows first, since a full door is worth more to the plant than a half-empty one, and a supplier who cannot hold a window gets pushed to nights. Ocean and rail carriers see a mode shift rather than a volume loss: fewer finished vehicles riding north on racks, more parts riding north in trailers, and the difference lands on two different networks with two different capacities. Importers of aftermarket parts should expect the sourcing pattern to follow the production pattern, because service parts are built where the vehicle is built.

Here is my own arithmetic, and every assumption in it is mine, not the announcement's. Assume the expansion eventually adds 100 inbound trailers a week, assume an average three hours of waiting per trailer at the crossing, and assume a driver cost of $30 an hour. That is $90 a trailer, $9,000 a week, roughly $470,000 a year. The money is not the interesting number. The interesting number is 300 driver-hours a week sitting still, which is seven and a half shifts of a driver doing nothing. That is what decides whether you need a fourth tractor on the shuttle, and it decides it long before the finance model notices. Run the door math the same way: assume a door turns eight trailers in a shift, 100 trailers a week over five days is 20 a day, which is two and a half doors running flat out with no recovery time. Two and a half doors is not a plan. It is a hope with a forecast attached.

The point the coverage misses is that this is a relocation of the bottleneck, not an addition of capacity, and the two bottlenecks have different shapes. In Mexico the constraint sat at the border rail ramp and in the pipeline of finished cars riding north, which is inventory on wheels: slow, visible, insured, and easy to explain to a board. In Texas the finished-vehicle pipeline shortens and the parts pipeline gets faster and more frequent, and the constraint moves to a dock door and a yard spot that nobody put in a press release. Companies measure the first thing because it shows up in transit days. They miss the second because a dock appointment that slides by two hours does not show up anywhere until the line stops, and by then the conversation is about expediting, which is the most expensive way to buy back a planning failure.

The counterfactual is not hypothetical, it is the road not taken. Had the volume stayed in Mexico, there would be no new doors to build and no new yard to pave, but the finished-vehicle pipeline would stay long and the plant would keep depending on a rail ramp at the border it does not control. Had the plant expanded without adding doors and yard, output would rise until the yard saturated, and the symptom would look like a labor problem, with overtime and detention and drivers refusing the lane, when the actual constraint is a rectangle of asphalt. Had Micron allocation loosened, the expedited premium on memory would fall, but the discipline of hourly line-side buffers should not fall with it, because the next shortage will not announce itself either.

Two clocks, then, and they should not be managed with one calendar. The $3.6 billion is a multi-year build, and its freight consequences arrive with the building, which means the dock design decisions being made now are the ones that will bind later, when changing them costs ten times as much. The memory agreements are a this-quarter exposure: allocation is tight now, so the buffer decision is now. Building a five-year plan for a problem that bites this quarter is how companies end up expediting parts by air while a perfectly good yard sits half empty.

The traps are dull. Counting square footage as capacity and never counting doors. Assuming a parts move and a finished-vehicle move need the same appointment structure, when one is a train and the other is a hundred trailers. Reading a supply agreement as a transport solution, when it only guarantees you a place in the allocation queue and says nothing about how the part reaches the line. And forgetting that a shuttle waiting at the border burns driver hours whether or not the trailer moves, so a queue is a cost even on the days nothing ships.

Go stand at your receiving dock this week and count two things: how many doors you have, and how many trailers are parked waiting for one. Then call the carrier who runs your border shuttle and ask what appointment availability looks like for the next two weeks. If the answer is that they will get you in sometime, you already know what the next expansion will cost you, and it is not in the $3.6 billion.

↑ Back to the key points

— By Ömer Kaya