Lam Research Corporation LRCX

Key data

The business

Lam Research makes the machines that etch and deposit the microscopic layers of a chip. Two processes are its trade: etch, which carves patterns and holes into the silicon, and deposition, which lays down the thin films between them. In 3D NAND memory, where storage is built by stacking cells hundreds of layers high, every added layer means more etch and more deposition steps, and Lam owns the hardest of them, the high-aspect-ratio etch that drills a clean hole straight through the whole stack. Few rivals can do it at all. The customers are the handful of companies that make memory and leading-edge logic: the big NAND and DRAM makers, plus the foundries. Once a tool is qualified into a customer's process it stays for years, and Lam sells the spares, upgrades and service around a growing installed base. That recurring support revenue is the ballast under a business whose equipment sales swing with the chip cycle.

The numbers

The trajectory is a clean cyclical recovery turning into something bigger.

Revenue climbed through fiscal 2026 and finished at a record $6.72 billion in the June quarter, up 30% from a year earlier, with the reported number again topping consensus. Profit grew faster than sales as the factories ran fuller: gross margin reached about 52% in the final quarter, up from about 50% a year before. The acceleration flagged a week earlier has held, with the September-quarter outlook still pointing higher and no downward revision since.

Fiscal 2022 earnings are shown split-adjusted for the October 2022 ten-for-one split. Step back and the cycle is stark. Fiscal 2024 was the trough, revenue down 14% to $14.9 billion as memory makers idled. Two years later revenue is $23.2 billion, up 56%, net income is up 90% to $7.3 billion, and diluted earnings nearly doubled from $2.90 to $5.76.

So 53 times trailing earnings is the peak-multiple-on-recovering-earnings trick: a huge number against profit that just went vertical. The market is not paying 53x for today. It is paying 22x a fiscal 2029 where consensus has revenue doubling again and earnings more than doubling to $13.66. The variant this memo holds is that the whole multiple rests on the durability of one thing, memory capital spending, and the print that settles it is whether NAND and DRAM makers keep raising capex through calendar 2027 rather than pausing.

Management

No insider has bought a share in a year; 55 sales totaled $106 million. Chief executive Tim Archer sold $11.7 million in July, and finance chief Doug Bettinger $20.5 million across two March sales; plan status is not disclosed, so read them as routine compensation-driven selling, not a signal either way. The capital return is the real tell. Lam repurchased $3.85 billion of stock in fiscal 2026 and $3.42 billion the year before, paid long-term debt down from $4.5 billion to $3.7 billion, and holds $5.6 billion of cash, so it now carries more cash than debt. Research spending held near 10% of sales, which is what keeps the etch lead. The one caution: buying back stock at 53x earnings near a cyclical peak is a poorer use of cash than it was at the trough.

How it fails or surprises you

The memory cycle rolls over (downside). Lam's fortunes track NAND and DRAM capital spending, which is famously cyclical. Fiscal 2024 showed what a pause does: revenue fell 14% and net income dropped to $3.8 billion. If customers finish this build-out and wait, the earnings inside that 53x collapse and the multiple becomes real. Watch customer capex guidance and Lam's own bookings.

China and export controls (downside). China has been a large slice of Lam's revenue, and the 10-K explicitly flags export controls, tariffs and trade disputes as able to inhibit sales. A tightening of rules on advanced etch and deposition tools to Chinese customers would hit revenue and the growth the multiple assumes at once. This is the fact the bullish read explains least well.

HBM and new transistors lift tool intensity (right tail). AI memory needs deep through-silicon etch, gate-all-around logic and backside power add process steps, and each new NAND layer adds more. If tools-per-wafer keeps stepping up, the $46.4 billion fiscal 2029 line is reachable and 22x is cheap. The tell is etch and deposition revenue and installed-base service growing faster than wafer starts.

Closing thoughts

One number decides it: customer capital spending through calendar 2027. The edge is not that memory is cyclical, everyone knows that; it is a judgment on how long this particular up-cycle runs. If the memory makers keep raising their capex budgets, the fiscal 2029 earnings of $13.66 are reachable and today's 22x forward is the right frame. If they pause, the earnings halve and 53x trailing is what you actually paid. An ambiguous quarter, flat capex with cautious language, leaves you holding a peak multiple with no confirmation. The left tail is a memory glut and a China clampdown landing together; the right tail is a structural step-up in tool intensity that makes the cycle shallower than it used to be. The downside is the one to fear, because it is the one you cannot see coming.

The bet is still that the world keeps building more memory and AI chips and keeps buying Lam's machines to etch and layer them. What breaks it is a capital-spending pause at the memory makers, and the pair of numbers that tells you first is customer capex guidance against Lam's own bookings: when those roll before reported revenue does, the cycle has turned, whatever the trailing earnings still say.