The Machine Is Only the First Sale

ASML beat the top of its own guidance by about 326 million euros this week. Roughly 92 percent of that beat came from machines it had already sold, in some cases years ago. The headline is that AI demand is strong. The lesson is that the best equipment businesses get paid twice, and the second payment is the one that compounds.

ASML reported Wednesday morning. Sales of 9.3 billion euros, guidance raised for the full year, the stock up as much as 5.6 percent before the open. Every wire ran a version of the same sentence: AI demand is strong.

That sentence is true and it is nearly useless. We already knew AI demand was strong. What we did not know is where the money actually came from, and the answer is not where the headline points.

The number nobody put in the headline

ASML told the market to expect somewhere between 8.4 and 9.0 billion euros. It delivered 9.33 billion. Call the beat 326 million euros over the top end of its own range.

Then management explained the composition. Installed Base Management, the line that captures service and field upgrades, came in roughly 300 million euros above plan. Do the division. About 92 percent of the money above ASML’s own ceiling came from machines already sitting on customer floors.

Not from selling more machines. From the ones it sold in prior years, running harder, needing more service, and getting upgraded in place. The CFO named the reason plainly: customers are looking for productivity enhancements.

The business most people skip

An ASML lithography system prints the circuit patterns that make advanced logic and memory possible. It is the machine no hyperscaler can route around, which is why the company gets written about as a monopoly on the hardest step in chipmaking.

But once that machine is installed, ASML does not stop earning. Installed Base Management is four things: maintenance and service, replacement parts, software and performance improvements, and upgrades installed at the customer’s fab. It reached 2.762 billion euros in the quarter, or 29.6 percent of total sales, up about 31 percent from a year ago.

Say the important part out loud, because it is where most write-ups get sloppy. This revenue is recurring in character. It is not a subscription. Nobody signs a contract promising ASML a decade of upgrade purchases. Calling it an annuity overstates the case and would set you up to be surprised later.

Why running the machines hard matters

Here is the mechanism. When fabs run near capacity, downtime gets expensive fast. A tool sitting idle during a shortage costs the customer real money, so service intensity rises and the economics of a productivity upgrade get easier to justify.

The customer stops asking only one question. It is no longer just “do we need another machine.” It becomes “can the machines we already own produce more wafers per hour.” ASML has an answer for sale, and shipping it does not require building anything new.

That second question is the whole essay. It creates a revenue path that does not depend on the capex cycle turning again.

The flywheel, and what it did to the margin

The mechanism is simple enough to hold in your head. More systems shipped means a larger installed base. A larger installed base means more service and more upgrade opportunities. More of both means deeper integration into the customer’s roadmap, which makes the next machine easier to sell.

Now look for the evidence in the margin, because that is where a story becomes a fact. ASML guided gross margin to 51 or 52 percent. It printed 54.0 percent. Upgrades and service carry better economics than building and shipping a new system, so when the mix tilts that way, the margin tells on it.

The scale of the shift is easy to miss. At its 2024 investor day, ASML laid out a 2030 opportunity of 44 to 60 billion euros in sales. Its new 2026 guidance midpoint is 44 billion. The company just reached the floor of its own 2030 target four years early. The margin target has not been pulled forward with it, which is the honest asterisk, but the revenue line arrived in 2026 wearing 2030’s clothes.

The other side

A thesis you cannot argue against is not a thesis. Here is the case against.

So the question is not whether Installed Base Management is a good business. It plainly is. The question is how much of that durability the price already assumes.