AI supply-chain thesis — mapping bottlenecks, focus companies, and supply-chain exposure for investors.
**Bottleneck theme:** Outliers
**Focus:** $AEHR — AEHR TEST SYSTEMS
AEHR is a tiny pure-play in wafer-level burn-in (WLBI) — the high-volume reliability-stress test step that catches infant-mortality failures on power semiconductors and increasingly on AI accelerators and HBM stacks before final assembly. Burn-in at the wafer rather than packaged-part level is the only economically viable approach for high-cost, high-power devices: SiC power MOSFETs (every EV inverter), high-voltage GaN, and now leading-edge AI accelerators where the cost of a packaged-part failure is enormous. AEHR's FOX-NP, FOX-XP, and Sonoma platforms are the only volume-qualified tools at scale.
The historical bull case has been EV-driven SiC burn-in (Onsemi, Wolfspeed, ST). The new and arguably bigger bull case is AI: as accelerator dies grow ($1k-$10k+ packaged-part value), the economic case for WLBI on logic and memory stacks strengthens. AEHR has cited engagements with leading AI silicon vendors and HBM customers; conversion of these to volume tool orders is the catalyst path. Risks: tiny revenue base means single-customer push-outs cause big-percentage misses, EV-cycle weakness has compressed near-term bookings, and AI WLBI adoption is still nascent (could go faster, slower, or to a competitor). High-beta exposure to a structural reliability-test trend.
Memory Supercycle — Investment-thesis bucket from bottlenecks.app: Memory Supercycle
Power Semiconductors — Discrete power devices — silicon MOSFETs and IGBTs plus wide-bandgap GaN and SiC transistors and diodes — used to switch and convert electrical power in EVs and chargers, solar/wind inverters, datacenter PSUs, industrial drives, appliances, and consumer adapters.
