A column about curious stories from the worlds of science and technology. Today—a story about the biggest rocket on the planet, one that does not have a single proper door. It sounds like a joke about an architect, but it is a deliberate engineering compromise, and its consequences reach surprisingly far: while SpaceX carries its satellites through a narrow slot in the side, design bureaus on the other side of the continent are beginning to design their spacecraft around that foreign slot. This week, even Canada’s Telesat said so publicly—a company whose constellation is meant to compete with Starlink itself. Let’s unpack why the largest cargo bay in the history of spaceflight has also turned out to be one of the most inconvenient, what PEZ candies have to do with it, and how a launch shortage turns someone else’s design flaw into an industry standard.
Let’s start with geometry. Every space “truck” needs something to cover its cargo with: a rocket scours through the atmosphere at thousands of kilometers per hour, and there is a standard strategy for that—a fairing. On conventional rockets, it is a two-piece shell: at altitude, locks release, the halves swing apart, the payload is exposed to sunlight, and the halves fall into the ocean and become expendable hardware. Cheap and cheerful, if the rocket is disposable.
Starship plays by different rules: it is fully reusable, including the nose cone with its payload bay, and the entire structure is covered on the outside with thermal-protection tiles. You cannot jettison the fairing—that would literally mean stripping the ship of its armor. So it needs a hatch that can open and close repeatedly without compromising the heat shield. Now pay attention: tiles cover most of the hull—the entire windward “belly” that faces the flow during atmospheric entry, and a substantial part of the side. You simply cannot put a door in the belly: that is the zone of maximum heating, and any gap there is an invitation for plasma to get inside. That leaves the leeward side, but there too the door has to sit flush with the tiles, have no protruding hinges or locks, and still hold vacuum and withstand launch loads.
Every attempt to make that door larger ran into the same old problems: a bigger panel means more edges, seals, and oddly shaped tiles around the cutout, more chances that something will jam during atmospheric entry. The practical test came early: the third test flight in 2024 already exposed problems with the door’s operation, and SpaceX has been methodically refining the mechanism ever since. Nobody has expanded the opening in the process. Quite the opposite: in Version 3 ships, the payload bay was redesigned from scratch—more compact and denser, while the opening in the side became a neat slot sized to Starlink satellites, and not a centimeter more.
The bay itself, meanwhile, is enormous: the ship is nine meters in diameter, with several hundred cubic meters of usable volume inside—unheard-of luxury by market standards. That is what makes the situation absurd: the planet’s biggest cargo bay accepts payloads through a mail slot.
A little on how it works at all. Musk himself coined the term back in 2022: a “PEZ dispenser”—after the Austrian candy dispenser in which little candy blocks sit in a stack and come out one by one. Starlink satellites lie inside Starship exactly that way: in a flat stack, like candies in a dispenser. The mechanics are more kitchen refrigerator than space station: chain drives, clamps; each slab in turn is fed to the slot in the side, catches on the grab mechanism, and glides away into free flight. There is almost no machinery in the bay itself—the entire “smart” part lives in the stack and in the slot.
The flat geometry is not a whim but a solution built around a standard. Starlink’s new-generation satellites, according to available data, are panels roughly 7 by 3.5 meters, around 30 centimeters thick; two such panels fit side by side in a single layer of the stack. Early enthusiast estimates suggested the hold could accommodate around four dozen satellites from the previous generation. Every candy is standardized: the same attachment points, the same interfaces, the same sequence. One type of cargo—that was the logistics SpaceX was after.
If this story sounds familiar, that is no coincidence. In the 1950s, American shipping operator Malcolm McLean proved a simple point: it does not matter how much the cargo itself costs—the problem is that every time it changes hands, someone has to manually move it from a barrel into a hold and back again. The standardized container eliminated the handling, the cost of shipping piece goods by sea fell by orders of magnitude, and global trade got a boost we can still feel today.
In the mid-2020s, SpaceX is pulling the same trick, only vertically: making the satellite into the container. Flat, standardized, with no protruding parts, with fixed attachment points. Then all the individual work disappears—adapters, docking rings, resonance testing, weeks of integration. The satellite simply travels in the general sequence, like a box in an Amazon warehouse.
And here a second conflict arises, an economic one. Integrating a satellite for a specific rocket costs money, and not a little, but against a spacecraft worth hundreds of millions of dollars and a $70 mln launch, it was always a minor expense. Falcon 9 has long had it all worked out: five standard adapters to choose from, collaboration with SpaceX engineers—a few weeks of fuss against the backdrop of a project that takes years, and no one worried about that routine. But when there are not enough launches on the market, the equation changes: the queue matters more than the details, and customers are willing to redesign a satellite around someone else’s slot just to get onto the schedule.
Last week, analysts at Quilty Space published a report that Eric Berger of Ars Technica accompanied with the headline “The launch crisis is worse than we thought.” The numbers are these: outside China, the world is short around 150 launches between now and 2030—given demand from satellite operators and civilian agencies. Add U.S. military demand—and the Space Force has already raised its own projections, from 30–60 to roughly 170 launches in one program line and from 54 to 79 in another—and the deficit could double. Quilty’s list of causes: the war in Ukraine, which pushed Russian rockets out of the Western market; explosive demand from Amazon Leo and other constellations; delays to the new generation of rockets—New Glenn, Vulcan, Neutron, Terran R; and a fresh factor—the withdrawal of Falcon 9 from commercial service. Analysts expect the shortage to peak in 2028. Quilty research director Caleb Henry offered no reassurance in an interview with Ars: “The launch crisis is more serious than we thought. It will not go away this decade—and may not ease until the end of the 2030s.” And then, in a separate line apparently addressed to everyone with a satellite in their plans: “If you haven’t booked a launch yet, you’re already late.”
Against that backdrop, two candidates are converging on the task of easing the market: New Glenn, which is gathering momentum, and Starship. But even Quilty’s cautious forecast gives New Glenn only a dozen to a dozen and a half launches a year toward the end of the decade, while Starship will only suit those who fit through the slot.
On October 30, during a conference call with investors, there came a moment that deserves an article of its own. Dan Goldberg, CEO of Canadian operator Telesat, was discussing the fate of the Lightspeed project—a constellation of 225 low-Earth-orbit satellites on which the company is staking its future in the race against Starlink. Launches have been booked for nearly all those spacecraft. Except two. The very two that are meant to fly first and prove that the design works.
Launching the first vehicles without a place of your own on the schedule is an awkward position in itself. Goldberg explained frankly: the company is trying to “squeeze” two test satellites onto a SpaceX rocket booked by another firm, and even under that arrangement, full Lightspeed service is being pushed back another quarter—to mid-2028. Behind the scenes lies Telesat’s loss of more than $157 mln in the last quarter: the price of entering the race is rising before one’s eyes.
Then came the line that made the entire call worth listening to. Of Starship, Goldberg said this: first, SpaceX will focus on its own satellites and its own needs; second, beside them stands a priority customer, the U.S. government. And only then: “I firmly believe that if our next-generation satellites are configured to fit in their dispenser—the very one they use to deploy their own—we have every reason to expect that Starship will be available to Telesat for future missions.” Think about that: the head of a multibillion-dollar satellite project is publicly describing the condition of admission not in terms of a contract, but in terms of the geometry of someone else’s payload bay. Design for our slot—get a place in the queue.
That is why the logic of a standard taking shape is easy to read. SpaceX will not build a larger door: Ars Technica’s analysis ends pessimistically for those expecting a conventional fairing—there is no work underway on a full swing-open hatch, no sign of one appearing either, and the Starlink-optimized slot will remain the main configuration for years to come. Yet every incentive converges at one point. Every dollar invested in a universal door is a dollar taken away from the main customer, the company’s own constellation. And conversely: every customer that redesigns a satellite around the PEZ slot strengthens standardization and makes launches cheaper for everyone already inside the standard.
Breakdowns on independent platforms boil down to one formula: “design for the candy standard—and get a discount and a slot on the schedule.” Some have already heard it: the Muon Space orbital data center is being designed around PEZ dimensions from the start. And the recently unveiled Starfall—SpaceX’s orbital platform for outside customers—is built in exactly the same form factor. The logic has closed in on itself: the truck carries passengers only in its own standard, and the market is now discussing not rates, but dimensions.
The lesson from the world of containers now has to be read in reverse. The strength of McLean’s standard was that every port adopted it freely. The PEZ slot is accepted by one—and only one—“port,” but without it you cannot get into any queue at all. In a world short of launches, geometry works like a tariff: if you do not fit the slot, then please, fly on something else—and that something else is already gone.
To be fair, the “slot” has a natural limit. Anything oversized—large antennas, telescopes, orbital stations, spacecraft that need to unfold after deployment—simply will not fit into a 7-by-3.5-meter box, no matter how much you flatten it. For such payloads, the market is nearly empty, and that is a separate niche that someone will eventually try to occupy. But hand on heart: the share of such missions in the overall flow is small, while the network effect of constellations is enormous. When hundreds of standard “slabs” a month begin flying to low orbit, tailoring yourself to the slot will cost less than waiting for an alternative.
In the end, this story is older than rockets: standardization always begins when someone dictates the dimensions of the box, and everyone else adapts their goods to it. The only question is who owns the box. In the 1950s, its dimensions were determined by competition among many carriers. In the 2020s, they are dictated by a single seller—and the consumer, not always noticing, designs their product around someone else’s warehouse.
That’s PEZ for you.
Sources: Ars Technica (30.10.2026) on the Quilty Space report “Quantifying the Launch Shortage”; The Logic (30.10.2026) on the Telesat briefing; Ringwatchers—technical breakdowns of the PEZ dispenser and Ship 33 configuration; the analysis “Starship changes the rules for commercial payloads” (eagerspace.net). Estimates of Starlink V3 dimensions and stack capacity are independent enthusiast estimates; SpaceX has not published official figures.