Every advanced chip in the world is drawn with light from a machine that one company builds. In April it guided to €36–40 billion of sales. By July that was €43–45 billion.
A company raising its full-year guidance by roughly a sixth at the midpoint, three months in, is not responding to demand. It is rationing it — and it is planning 30% more capacity for the following year.
Most moats are a matter of degree: a brand that is a bit stronger, costs that are a bit lower. This one is a matter of kind. There is no second supplier.
It is not one invention, it is a few hundred
The usual explanation — "they have the patents" — is the weakest part of the story. Patents expire and can be designed around. What cannot be designed around is the shape of the problem.
- The physics is at the edge of what is possible. Extreme-ultraviolet light is absorbed by air and by glass, so the whole optical path must run in vacuum, using mirrors polished to a smoothness where the error budget is measured in atoms.
- The supply chain was co-developed and does not sell to anyone else. The optics, the light source, the stages — these come from a small number of partners who spent decades building to one customer's specification. A rival cannot buy the parts, because the parts do not exist as products.
- The knowledge only accumulates in use. Every machine running in a customer's fab returns information that improves the next one. A new entrant starting today would be twenty years of installed-base learning behind, and the gap would widen while they worked.
- The customer cannot help a challenger. A chipmaker switching to an unproven tool risks its entire process node. The rational move is always to wait for someone else to go first, which means nobody goes first.
A competitor would have to rebuild the optics, the light source, the metrology and the suppliers at the same time — and then wait twenty years for the experience.
What a real monopoly looks like in the numbers
| ASML · 28 Sep 2026 | What it indicates | |
|---|---|---|
| 2026 guidance | €36–40bn in April → €43–45bn in July | Raised by a fifth at the bottom of the range, three months in |
| 2027 capacity plan | +30% | Demand is not the constraint; the factory is |
| Customer position | TSMC at 67.7% gross margin, revenue +40% | The customers are themselves extremely profitable — they can pay |
| Valuation | 40× 2026 · 29× 2027 | Earnings yield of 2.5%, against a 5.12% Treasury |
| The year's range | −11% from the high, +89% from the low | The business did not change this much. Opinion did. |
The third row is the one people skip. A monopoly is only valuable if its customers can afford it — a sole supplier to a dying industry is a sole supplier to nothing. Here the customer earns two-thirds of every revenue dollar as gross profit and is growing over 40%, with 77% of its wafer sales on nodes where it has few rivals.
Two monopolies stacked on top of each other, each one's pricing power resting on the other's.
Neither of them is a competitor
China was about a third of last year's sales, is being cut back, and American lawmakers are pushing to restrict it further. That is a third of the revenue base whose availability is decided in legislatures rather than by customers.
And the second risk is ordinary: semiconductor capital spending is cyclical. Customers order in waves, and a company selling the most expensive tool in the fab feels the trough hardest.
Note what both of these have in common. Neither touches the monopoly. They hit the earnings, possibly severely, while the competitive position stays exactly where it was — which is the specific reason a share price can fall 40% in a business nobody can enter.
The uncomfortable part
Everything above argues for owning it. The price argues back.
At 40 times this year's earnings you are accepting an earnings yield of 2.5% when a government bond pays 5.12%. Every bit of that gap is a claim about the next decade — and the claim is probably right, which is exactly why it is priced in.
Our own conclusion was accumulate slowly, and the adverb is doing real work. A wonderful business bought at a full price gives you the business's returns minus the premium you paid; a wonderful business bought in a trough gives you both. The cycle is the only thing here that reliably hands out second chances.
When you think you have found a moat, ask: what would a well-funded competitor actually have to do, and how long would it take even if everything went right?
For most "moats" the honest answer is "spend a lot and wait three years". When the honest answer is "rebuild four industries simultaneously and then wait two decades for the experience", you are looking at something different in kind — and the market will have noticed too.
The only company that makes the machine, and the company that owns most of the machines.
The monopoly, the cycle, the export rules and the valuation, in the right currency.
Frequently asked
Why can't anyone else build EUV lithography machines?
Because the machine is not one invention that could be patented around — it is a few hundred of them, assembled over two decades, held together by a supply chain that was co-developed with the company and does not sell to anyone else. A competitor would have to rebuild the optics, the light source, the metrology and the suppliers simultaneously, then accumulate the process knowledge that only comes from machines already running in customers' fabs.
Is ASML really a monopoly?
In extreme-ultraviolet lithography it is the sole supplier, which is about as close to a true monopoly as capitalism allows. The qualification is that it depends on a handful of customers who are themselves concentrated, and on export rules set by governments — China was roughly a third of last year's sales and is being cut back, with American lawmakers pushing to restrict it further.
Is ASML expensive?
Yes, on any ordinary measure. At our September 2026 report it traded at 40 times 2026 and 29 times 2027 consensus earnings — an earnings yield of 2.5% against a Treasury paying 5.12%. The shares were 11% below their high and 89% above their low in the same year, which tells you how much opinion changes about a business whose competitive position does not.
What would break ASML's moat?
Not a competitor, on any visible timescale. The realistic threats are that its customers stop needing more machines for a few years, which is the ordinary semiconductor cycle, or that governments restrict who it may sell to. Both affect the earnings badly and neither touches the monopoly.