The machine behind the boom — and the oven-maker’s problem
Everyone is building data centres — that is the story of this decade, and our master theme, the Data Centre Boom, tells it in full. This page sits one step up the supply chain: data centres are filled with chips, chips are made in fabs, and fabs are filled with these three companies’ machines. For sixty years, every advance in chips has meant more work for the machine makers. But a seat at the front of the chain cuts both ways. These are the city’s oven makers in a restaurant boom: as long as new restaurants open, the phone rings. The week the city decides it has enough restaurants, the ovens stop selling — even though every restaurant keeps serving dinner. In 2023 the chip industry merely paused for breath — and Lam’s sales fell 14% in two fiscal years.
The stock market has seen this movie. In March 2000 the most valuable company on earth was Cisco, seller of the internet buildout’s plumbing. The buildout slowed. Cisco’s business survived and kept growing — but a buyer of the shares at the 2000 peak waited twenty-five years, until December 2025, to get back to even. The end of a buildout doesn’t have to kill the builder. It kills the price people will pay for the builder.
So the question for these three is not whether a slowdown ever comes — someday one does — but what each company keeps earning when it arrives. Their money comes in two kinds. New-machine sales stop when the building stops. Service revenue — parts, upgrades, support on machines already installed — keeps arriving, boom or pause: it is the rent on the installed base.
None of this is a forecast — the watch list in Part four is how a reader spots the turn. GARPify covers all three with full company reports — KLAC LRCX AMAT — built from the companies’ own filings and our standard data pull, every claim checked against its source, and no advice anywhere.
Three toolmakers, three jobs, one summer
Think of a fab as a construction site: KLA is the building inspector, Lam the specialist mason, and Applied the general contractor with a crew on almost every floor. Three different businesses — and one summer that cut all three share prices by 38–45% from their 52-week highs while the order books stayed full.
Every tag on this shelf reads high — even after the markdown
A price-earnings tag — what you pay for one dollar of a company’s yearly profit — only means something against that company’s own history. All three toolmakers fell hard this summer, and all three still wear tags far above their own five-year averages. Before you read that as one verdict, check the yardsticks:
| Company | The tag today | Its 5-yr typical | Premium | The state of its yardstick |
|---|---|---|---|---|
| KLA Corporation | 45.8× | 28.9× | 1.58× usual | The cleanest stick on the shelf — no windfalls or fines in its record. A 45% fall from its 52-week high still leaves a 58% markup. |
| Lam Research | 47.0× | 26.6× | 1.77× usual | The dearest tag against its own history — on the most cyclical earnings of the three. Memory spending swings hardest, and its report shows what the last downcycle did to the profits under this tag. |
| Applied Materials | 36.3× | 22.9× | 1.59× usual | Reads as the cheapest tag on the shelf — but about a dollar in nine of the trailing earnings under it is investment-portfolio marks, not tools. Strip the marks and the tag reads higher. Its report keeps the two ledgers apart. |
What survives the straightening is the real story: even after falls of 38–45% from the 52-week highs, the whole shelf still trades at 1.6 to 1.8 times its usual charge — all three names marked up together. The filings offer real reasons: record order books, AI-driven fab construction, margins at or near records. Every reason is real. Every reason is also still in the price — and a markup this broad is a verdict on the industry, not the companies. Each report prices its own company against its own history and hands the reader the choice; none of them tells you what to pay.
Four dials, all published quarterly
The fabs’ nerve. This is the Cisco-moment detector. Equipment spending breaks the day the fabs lose theirs. The warning will not come from the three toolmakers; it comes from the capital-spending guidance of TSMC, Samsung, Micron and SK Hynix — the first quarter one of them talks about deferral instead of ramp is the quarter to re-read everything on this page.
Washington’s rules. China is the largest single geography for all three — 33% of KLA’s revenue, 34% of Lam’s, 30% of Applied’s — and export rules keep tightening. Applied is the live reminder of the stakes: a February 2026 settlement with the Commerce Department — $253 million paid, and a suspended denial order that lifts only after three years of clean compliance audits. One of the three is, in plain terms, on probation.
The order books. KLA reported $7.9B of contracted work at its fiscal-2025 year-end, Lam $2.4B of deferred revenue at fiscal-2026 year-end, Applied a $15.0B order backlog — signed demand, published every quarter. The dial to watch is not the level; it is whether the books refill as they drain.
The tags themselves. Three different businesses — an inspector, a sculptor, a general contractor — wearing markups of 1.6 to 1.8 times usual at once. That closeness says the market is currently pricing the industry, not the companies. All three report again within weeks of each other this autumn — Applied on November 12 — and each report scores its company the day the filings land.
Own the theme in one purchase — and the one thing to know first
If the thesis convinces you and you prefer one purchase to three, all three toolmakers appear in the major semiconductor funds — they are large enough, and distinctive enough in their roles, that most broad chip funds hold all of them. The trade-off is dilution: every fund below mixes the three operators with dozens of chip designers and foundries, so a fund is a bet on the whole chip complex, with the toolmakers along for the ride.
| Fund | Fee | How it holds the three |
|---|---|---|
| SMH VanEck Semiconductor | 0.35% | Holds all three. Cap-weighted toward the biggest chip names (NVDA, TSMC), so the toolmakers sit lower in the weights than their role would suggest. |
| SOXX iShares Semiconductor | 0.35% | Holds all three, with capped weights — less top-heavy than SMH, so the toolmakers get a modestly larger seat. |
| SOXQ Invesco PHLX Semiconductor | 0.19% | The 30 largest US-listed chip names — all three included. Cheapest fee on this shelf. |
| XSD SPDR S&P Semiconductor | 0.35% | Equal-weight — every holding gets roughly the same seat at each rebalance, which favours the smaller equipment names. Weights drift between rebalances; check the provider’s current sheet. |
Want the businesses, not a blend? That is what the three company reports are for — the curve, the track record, the capital allocation, and the promises-versus-delivered record on each: KLAC LRCX AMAT
What the stock charts say — our standard trend test
One test, one definition, every page: weekly closes against the 52-week rolling average, with the average’s direction over the last 14 weeks.
| Company | Price, 52-week low – high | Earnings per share, 52-week low – high | P/E, 52-week low – high |
|---|---|---|---|
| KLAC +3% above a rising 52-wk avg | $96.70now $168$307 | $3.04now $3.67$3.67 | 31xnow 46x82x |
| LRCX +11% above a rising 52-wk avg | $117now $271$438 | $4.16now $5.76$5.76 | 29xnow 47x78x |
| AMAT +11% above a rising 52-wk avg | $169now $421$740 | $8.39now $11.60$11.60 | 20xnow 36x68x |