Between 1970 and 1988, Swiss watchmaking lost 62,000 jobs. Of 90,000, 28,000 were left. The number of firms fell from 1,600 to 600 in thirteen years. The reason was simple: Seiko and Casio could measure time more accurately, and more cheaply.
The Swiss lost the race for the function itself. Today no mechanical watch competes on accuracy with a quartz oscillator, a sliver of synthetic quartz that vibrates about 32,000 times a second and is so stable that its error is counted in seconds a month, not a day. A mechanical movement, where the pace is set by a balance wheel on a spiral spring, can't compete with that. Let alone with a phone that keeps pulling the time from the network.
And yet the watchmaking industry is in a shape it wasn't in fifty years ago.
The thesis of this piece is simple. When a layer gets commoditised, demand doesn't disappear. It moves one floor up, to the layer that couldn't be commoditised. And that layer is often driven by taste.
What actually happened in Switzerland
An industry that had lost its main argument, accuracy, did the only thing that made sense. It stopped competing on technical specs and started competing on what a machine can't make cheaply.
Three things.
- Construction and design. Decisions about the proportions of the case, how wide the bezel should be, whether the dial carries indices or numerals, and when there's enough on it already. None of it has a functional benefit. And yet all of it decides whether the piece is worth more or less twenty years later.
- Progress in manufacture. Silicon hairsprings that don't magnetise and don't change their elasticity with temperature. Balance wheels that resist strong magnetic fields. Chronographs that measure a hundredth of a second. An industry that lost on technology spent the next thirty years on materials research that is useless next to a phone. Accuracy stopped being a reason to buy, but it never stopped being the standard of the craft.
- A brand's accumulated decisions. Patek Philippe doesn't hold its value because it has a better movement than the competition. It holds it because for a hundred years in a row it didn't make a stupid decision about what not to make.
What it looks like in numbers
In 2025 Switzerland exported watches worth CHF 25.5 billion. Volume fell 4.8% to 14.6 million pieces, value only 1.7%. Noticeably fewer watches were sold for almost the same money.
The split inside is even more telling. The segment above CHF 50,000 holds 37.3% of export value. And 89% of all the growth in 2025 came from that segment. The rest of the market stagnated or fell, pieces under CHF 500 by 4.5%.
Meanwhile, according to estimates by Morgan Stanley and LuxeConsult, Apple sold about 37 million watches in 2023, more than twice the volume of the whole Swiss industry, at an average price of around CHF 415. Rolex sold about 1.2 million pieces in the same year at an average price of CHF 12,500.
What works against it
The secondary market fell and hasn't recovered. The WatchCharts indices have been below their peak since spring 2022, and 2025 didn't save it: of the 25 brands WatchCharts tracks, only a fifth were up for most of the year. Blancpain −12.6%, Piaget −12.5%, Jaeger-LeCoultre −8.5%.
The honest reading is probably that part of the growth between 2020 and 2022 wasn't about taste but about helicopter money in the pockets of people looking for status. For two years watches became an asset and behaved like one, including how it ended.
What didn't fall is the primary market. If it were only speculation, the segment above CHF 50,000 would be falling fastest. It's falling slowest, and it's carrying the growth.
The second caveat, one I like to think about as a watch enthusiast and an important argument, is concentration. Four brands, Rolex, Patek, AP and Richard Mille, hold 49.1% of the market and added 2.2 percentage points year on year.
Why I'm writing this
Because exactly this shift is now happening a few industries over, in the segment that media around the world sell hardest every day, and most people read it completely differently. It's AI.
The common view is that once intelligence is free, we'll all be pragmatists. Everything gets optimised, price and speed decide, and soft things like taste and judgement lose their point.
Watches say the opposite. When accuracy was commoditised, people didn't start buying the cheapest way to find out what time it is. They started paying an order of magnitude more for what couldn't be commoditised. Demand for taste didn't appear in spite of cheap technology. It appeared because of it.
And here's what I take from it.
The market for bringing AI into large organisations is, I think, where the future of the whole field will be decided in the next few years. How governments, banks, insurers, hospitals and energy companies really start working with AI will decide how we as a society work with it and how we see it.
Today that market looks almost exactly like the watch market in 1975. Everyone sells the same thing, and everyone sells it as if it were what matters most. We can save you tokens. We'll build you a personal dashboard (and creatively call it Jarvis). It's an argument about accuracy at a time when accuracy no longer costs anyone anything.
In two years none of that will make a difference. The difference will be in two things you can't download from Hugging Face:
- Knowing what not to build. Most AI projects die because someone built what could be built, not what should have been. Throwing out use cases is the main work, not the preparation for it. Remember Patek Philippe a few paragraphs up.
- Understanding the operation, not the tool. Whoever doesn't know how the customer's company really runs, who sends what to whom, where it gets stuck, who has to sign it off, will build a nice clickable prototype. Knowing the segment is the layer nobody can copy. Prompt engineering isn't.
No model will do either of them for you.