LRCX

Lam's etch chambers are truck-bed-sized boxes that occupy a corner of a cleanroom the size of a football field. Inside, plasma — an ionized gas — is precisely controlled to ablate layers of material with atomic precision. Imagine a dentist's drill that removes material one atom at a time, guided by patterns defined to within nanometers. An etch chamber does that millions of times per second, across an entire 300mm wafer. A single fab has dozens of these tools. TSMC alone operates hundreds.

The deposition tools are equally large. Where etch removes, deposition adds — spraying or sputtering atoms onto the wafer to build up the conductive and insulating layers that form transistors. Together, etch and deposition touch nearly every layer of a modern semiconductor. Revenue runs approximately $17-18 billion annually.

Lam holds roughly 45% of the global etch market and 25%+ of deposition. The company manufactures these tools and ships them to every major fab: TSMC, Samsung, Intel, SK Hynix, Micron. Each Exelan or Omega system takes 9-12 months to build and validate. Installation at the customer fab requires additional months of process qualification before the tool enters production. The tools consume specialty gases continuously (argon, nitrogen, chlorine, fluorine compounds) from suppliers like Linde and Air Liquide.

Every advanced AI chip — NVIDIA's H100 and B200, AMD's MI300, TSMC's leading-edge production at 3nm and below — passes through Lam's etch tools. HBM stacking, the only way to deliver the memory bandwidth AI training demands, is etch-intensive: each die stacks 12+ memory layers, each requiring dozens of etch steps. This is the mechanical link between AI demand and Lam's revenue.