Key Metrics Cheat Sheet
Equipment Company Metrics
| Metric | What It Means | Good/Bad |
|---|---|---|
| Book-to-Bill | New orders / revenue | >1 = growing backlog |
| Utilization | Fab capacity in use | >85% = tight supply |
| Lead Times | Weeks to deliver tools | Longer = strong demand |
| Gross Margin | Pricing power | >45% = differentiation |
Hyperscaler Metrics
| Metric | Why It Matters |
|---|---|
| Capex as % Revenue | Investment intensity (typically 15-25% for hyperscalers) |
| Cloud Revenue Growth | Demand signal for AI infrastructure |
| AI Revenue Disclosure | Emerging metric (Microsoft, Google now breaking out) |
Foundry Metrics
| Metric | Signal |
|---|---|
| Utilization Rate | Pricing power; <70% = price cuts coming |
| NRE (Non-Recurring Engineering) | Design activity; rising = new products |
| Advanced Node % | 5nm/3nm mix; higher = more equipment intensive |
Memory Metrics
| Metric | Interpretation |
|---|---|
| ASP (Avg Selling Price) | Rising = tight supply |
| Inventory Weeks | <4 weeks = shortage; >8 weeks = glut |
| HBM Capacity | Critical for AI training; constrained through 2025 |
Quick Reference: Public Companies to Watch
Equipment (WFE):
- ASML (Lithography) — Monopoly on EUV
- LAM Research (LRCX) — Etch leader
- Applied Materials (AMAT) — Broadest portfolio
- KLA — Metrology/process control
Hyperscalers:
- Google/Alphabet (GOOGL)
- Microsoft (MSFT)
- Amazon (AMZN)
- Meta (META)
AI Chips:
- NVIDIA (NVDA) — 80%+ training share
- AMD — Rising with MI300
- Broadcom, Marvell — Custom AI chips
Foundries:
- TSMC (TSM) — 60%+ advanced logic
- Samsung — #2
- Intel (INTC) — Trying for #3