Our pitch: burn-in catches bad die at the wafer, saves 10-100x scrap cost after it's stacked into a CoWoS package with HBM. Most AI chips aren't burned in yet - we estimate 5-20% of ASICs, about half of AI accelerators see production burn-in today. Direction is toward more, okay? Sonoma handles 2,000 watts, a contract manufacturer adding 20+ systems/month. Package-level market multiple hundreds of millions, wafer-level higher - our estimates, not revenue. Whether penetration rises with packaging complexity... open question. We'll see.
← stream@AEHR solid