The chip names that led the market for two years are coming apart. The companies that supply them are doing the opposite.
Tuesday's failed bounce in semiconductors rolled into a second selling day this morning. Nvidia is down more than 2% near $203. The broad chip group extended its breakdown, and a hot inflation print that keeps the discount rate high gave sellers another reason.
But look one layer down the supply chain and the tape flips. Applied Materials, KLA, and Lam Research each printed a new 52-week high in the first hour of trading. At midday all three are still green while the Nasdaq is down more than 1%.
The Picks and the Shovels
These three do not design chips. They build the machines that make chips possible.
Applied Materials handles deposition and materials engineering. Lam Research owns the etch and deposition steps that carve circuits into silicon. KLA inspects the wafers and finds the defects before they ruin a batch. Every leading-edge fab on earth, whether it belongs to TSMC, Samsung, Intel, or Micron, runs their equipment.
That is the structural point. The market spends its days arguing about which chip designer wins the AI race. The toolmakers get paid no matter how that argument ends. When the constraint on AI is how fast the world can build capacity, the companies selling the capacity capture it first.
Why the Spending Keeps Climbing
The demand behind these stocks is not a single quarter of orders. It is a multi-year build-out.
Hyperscalers like Microsoft, Amazon, Google, and Meta have each committed to record capital budgets to expand data centers. Those budgets eventually land as fab orders, because more AI compute means more advanced chips, and more advanced chips mean more etch, deposition, and inspection steps per wafer. The move to ever-smaller transistors and stacked memory does not just add demand, it adds complexity, and complexity is what these three companies sell.
That is the difference between this cycle and past chip booms. The equipment makers are tied to the pace of construction, not the price of any one chip. A memory glut can crush a designer's margins while the fabs keep buying tools to stay on the leading edge.
The Catalysts Underneath the Move
The fundamentals are backing the price action.
Applied Materials just opened a $500 million campus in Singapore aimed squarely at AI chip demand, more than doubling its advanced cleanroom capacity there. The company also lifted its quarterly dividend to $0.53 a share this week. It carries a market value near $400 billion and trades close to its all-time high after a roughly 167% run over the past year.
KLA is up about 46% in three months on rising demand for process control and advanced packaging, the steps that let chipmakers stack memory next to logic for AI workloads. It pays a small dividend, earns a net margin around 36%, and sits at a $284 billion market cap. Lam Research holds a near-monopoly in etch tools and posted record margins in its latest quarter. Its shares have gone from $88 to $348 over the past year, a move of nearly four times off the low.
What the Valuation Is Telling You
The catch is the price tag.
KLA and Lam each trade around 60 times trailing earnings. KLA changes hands at roughly 22 times sales. These are not the multiples of sleepy industrial suppliers. They are growth multiples, and they assume the capital-spending cycle keeps climbing.
That works in both directions. KLA carries a beta near 1.5 and Lam near 1.9, meaning they move faster than the market in either direction. If the chip-designer selloff turns into a broad de-rating of anything tied to AI, the equipment names will not be immune. Their earnings also swing with the spending plans of a handful of customers, so a single delayed fab build can dent a quarter. The divergence on display today is real, but it is not a hedge.
What to Watch From Here
The signal to track is order flow, not the daily chart.
The next read on wafer-fab equipment bookings, plus the capital-spending guidance from TSMC, Samsung, and Intel, will tell investors whether the toolmakers can keep diverging from the names they supply. As long as the AI bottleneck is physical capacity, spending flows toward the equipment layer. If demand cools or a customer trims its build-out, that same layer corrects hardest.
For now the split is the story. The market is selling the chips and buying the machines that make them.
