Counting launches is the wrong way to keep score

The launch count is the space industry’s most quoted number and its least examined one. This year the world has attempted 186 orbital launches to 6 August, of which 180 worked, against 330 across the whole of 2025. Our own Analysis desk publishes the table, because it is the figure readers ask for first. It is also a figure that rewards a little suspicion.

Two honest counts, fifteen flights apart

Start with China. On the itemised tables the total is 55 so far this year. Other trackers put it nearer 70 over the same period, and China’s own stated ambition for 2026 sits near 140. None of those numbers is a lie. They differ because the people compiling them draw the boundary in different places: what counts as orbital, when a suborbital test article belongs in the tally, whether a failure that reached space is an attempt or an event.

Starship makes the point sharply. Its 2026 test flights fly deliberately suborbital profiles and splash down. One respected catalogue folds them into orbital launch attempts; the itemised tables keep them separate. Both are defensible. A reader comparing two headlines built on different choices is comparing nothing at all.

What a flight actually delivered

The deeper problem is that a launch is a unit of activity rather than of capability. Flight 13 in July put twenty Starlink V3 satellites out of a Starship for the first time, a genuine engineering milestone. Those satellites are not in service, because the vehicle was on a suborbital path and the catalogue records them as such. Counted as a launch, it looks identical to a Falcon 9 delivering a working batch to a usable orbit. Counted honestly, one is a demonstration and one is a delivery.

Run the same test on the constellation numbers and the picture sharpens. There are 10,904 working Starlink satellites on the independent count, and roughly 16,000 active satellites in orbit in total, two thirds of them from one operator. Those figures describe what exists and does something. A launch tally describes how often somebody lit an engine.

The number to publish next to it

None of this argues for dropping the count. It is a real measure of industrial tempo, it is comparable to itself year on year when one methodology is held steady, and it is the number the public understands. It should simply travel with company: mass delivered to orbit, working spacecraft added, missions that met their stated objective.

We will keep the cadence table on the Analysis desk, held to one source so its own comparisons stand up. Readers should treat any single-figure claim about who is ahead, ours included, as the opening of an argument rather than the end of one.