Applied's 39% Equipment Margin Is a Bet on the Next Order

The AI machine rush has paid Applied. What happens after the machines are installed will determine what those earnings are worth.

Applied Materials earned a 39.0% non-GAAP operating margin in Semiconductor Systems during the quarter ended 25 October 2026. A year earlier, it earned 29.6%. Those are the sort of numbers that make a factory look like a software business, until you remember who has to order the next machine.

The boom in equipment spending is real. The case for treating its profits as permanent is much harder. The Capacity Test asks whether chipmakers keep ordering equipment after the first wave of manufacturing capacity is installed. It is a tougher standard than beating an earnings estimate, because today's excellent results measure machinery already sold, while tomorrow's margins depend on the appetite to buy more.

In its 12 November 2026 results, Applied disclosed Q4 FY2026 revenue of $10.409 billion, up from $6.800 billion in Q4 FY2025. That is ($10.409 billion − $6.800 billion) / $6.800 billion = 53.07% growth. Semiconductor Systems supplied much of the acceleration: its revenue rose from $4.879 billion to $7.996 billion over those same quarters. These are sales, rather than hopeful speeches about artificial intelligence. The next sale remains a customer decision.

Now put a price on the margin. At Q4 FY2026 Semiconductor Systems revenue of $7.996 billion, a scenario in which the segment's non-GAAP operating margin returns from 39.0% to its Q4 FY2025 level of 29.6%, with sales held constant, removes $7.996 billion × (39.0% − 29.6%) = $751.624 million from quarterly non-GAAP operating income. That is a sensitivity, rather than a loss Applied has booked. Yet when you pay for a supplier as though exceptional margins are its new normal, you are buying an assumption about its customers' future capital spending.

The mix is changing too. In Q4 FY2026, DRAM represented 33% of Semiconductor Systems revenue, against 28% in Q4 FY2025; flash memory rose from 6% to 9%, while foundry and logic fell from 66% to 58%. More memory spending is excellent business for the company selling the tools. It also concentrates the bet on manufacturers that invest in waves: capacity arrives, the machines start producing, and the next purchasing decision depends on the economics of everything built in the previous round. You can sell extremely well into a cycle that later rests.

Applied has seen this movie in its own numbers. In a previous equipment cycle, its 14 November 2019 fiscal results put Semiconductor Systems net sales at $9.027 billion in FY2019, down from $10.577 billion in FY2018, while non-GAAP adjusted operating margin dropped from 34.3% to 27.8%. Memory fell as a share of Semiconductor Systems sales (DRAM 27% to 22%, flash 37% to 26%); the segment lost both sales and margin. The comparison has a limit. Those are annual numbers from a conventional memory cycle, whereas management now emphasizes AI-related device complexity alongside capacity investment. The older cycle explains the way margins contracted without predicting when the next contraction arrives.

The counterargument arrives with serious numbers. SEMI's 14 July 2026 outlook projects wafer-fab equipment growth of 21.8% in 2027 and 14.1% in 2028, and projects total equipment sales of $229.5 billion for 2028. Lam Research's results for the quarter ended 27 September 2026 give that optimism an operating counterpart: $8.219 billion in revenue and a 40.2% non-GAAP operating margin. Lam CEO Tim Archer described "AI driving higher semiconductor demand and greater device complexity" as the backdrop.

ASML supplies the inconvenient counterexample: weaker equipment orders and an intact business. ASML's results released 29 January 2025 show new lithography systems sold falling from 421 in FY2023 to 380 in FY2024, and annual net bookings slipping from €20.040 billion to €18.899 billion. Still, total sales rose from €27.559 billion to €28.263 billion, gross margin held at 51.3%, and Installed Base Management sales climbed from €5.620 billion to €6.494 billion. Installed Base Management sales rose while new-system units fell. ASML's uniquely constrained EUV franchise makes this an imperfect comparison with Applied. The resilience is real all the same.

ASML gives the clearest objection to treating every equipment boom as a future margin collapse. Applied itself has an installed business: in Q4 FY2026, Applied Global Services recorded $1.896 billion in revenue against $1.506 billion a year earlier, and $578 million in non-GAAP operating income against $434 million. Machines already installed can keep generating service work after a fresh equipment order disappears. But the weight of the business still sits with Semiconductor Systems, where the Q4 FY2026 39.0% margin is attached to $7.996 billion in sales. Services help the landing; they do not eliminate the height.

Cash also deserves its proper place. Applied generated $2.616 billion in non-GAAP free cash flow in Q4 FY2026, up from $2.043 billion in Q4 FY2025, an increase of ($2.616 billion − $2.043 billion) / $2.043 billion = 28.05%. Its free cash flow margin went the other way: $2.616 billion / $10.409 billion = 25.13%, against $2.043 billion / $6.800 billion = 30.04% a year earlier. That divergence is worth watching, but the company still made substantially more cash. Cash conversion measures money generated by the current business; future equipment orders still depend on customer budgets.

The more revealing comparison comes from a past boom. In the Q4 FY2021 release dated 18 November 2021, Applied disclosed fiscal 2021 Semiconductor Systems orders up 78% year over year and a Semiconductor Systems backlog rising from $5.5 billion to $6.7 billion during the fiscal quarter ended 31 October 2021, giving shareholders a line of sight into orders beyond machines already recognized as revenue. The 12 November 2026 release offers neither order intake nor backlog. That is the kind of evidence The Capacity Test demands. You learn something different from a queue of equipment orders than from the margin on equipment shipped. One is a claim on future customer budgets; the other is a reward for budgets already spent.

For now, management is offering another strong quarter. The 12 November 2026 outlook gives Q1 FY2027 revenue of $11.100 billion plus or minus $500 million and non-GAAP diluted EPS of $4.35 plus or minus $0.20. Gary Dickerson said Applied is on track to grow its Semiconductor Systems business more than 40 percent in calendar 2026. Those are guidance figures, not customers signing up for another capital cycle. A higher revenue run rate matters, particularly with Semiconductor Systems so profitable, but even a guide that holds only buys the company time to demonstrate that spending remains robust after the first wave of capacity arrives.

I would price Applied as a highly profitable equipment supplier with a cyclical option attached, rather than treat its Q4 FY2026 39.0% Semiconductor Systems margin as an annuity. For your money, future orders matter more than an encore earnings beat, because the cost of being early on a margin peak is larger than the satisfaction of admiring a record quarter. The falsifier is the Q1 FY2027 earnings release for the fiscal quarter ending in January 2027: Semiconductor Systems orders or backlog above the level at 25 October 2026, or, if Applied again discloses neither, Semiconductor Systems revenue above $7.996 billion at a non-GAAP operating margin at or above 39.0%, would show customers still buying after the first wave of capacity and overturn this view.