Applied Materials opens a $5 billion center to help build the next generation of AI chips - CTech

AMAT's $5B bet isn't a factory—it's a vertical-integration declaration of war. As GAA transistors hit volume production and backside power delivery moves from roadmap to wafer, etch depth control and atomic-level deposition uniformity become the binding constraints. By internalizing process development, AMAT is redefining itself from equipment vendor to process architect. The ripple effects are immediate. Lam Research's moat in 3D NAND etching erodes; Tokyo Electron's pricing leverage in track-and-develop compresses. More critically, TSMC (Taiwan, China) and Samsung at 2nm face a de facto lock-in: once process windows are calibrated on AMAT tools, switching costs escalate exponentially. On compliance, sustained BIS export restrictions structurally compress AMAT's China-revenue share. Capacity gets reallocated toward U.S. domestic fabs (Intel 18A) and Taiwan, China. Extended depreciation cycles and lower per-tool utilization will pressure gross margins—this is a certainty, not a risk. Over 12–24 months, AI inference chips will demand less advanced-node intensity than training silicon, creating a front-loaded, back-light equipment order pulse. AMAT's real wager isn't AI demand—it's who defines the next process standard. Win, and it becomes the next ASML. Lose, and $5B becomes a depreciation black hole.