They can't distinguish HBM from standard memory anymore because the customer uses the tool for both.
Thesis: Veeco is not a cyclical equipment play; it is a structural bottleneck. They own the Laser Annealing (LSA) step for Gate-All-Around and HBM. Margins are expanding (43% -> 44%) because customers have no alternative at advanced nodes. The 'capacity ceiling' is actually a pricing power floor. The street is mispricing a secular AI infrastructure compounder as a trailing-edge China cyclical.
Verdict: LONG — Conviction: HIGH
Catalyst: Qualification of the second HBM customer (moving from demo to eval) or the conversion of the first Nanosecond Annealing (NSA) evaluation at a Tier 1 logic customer.
Key Risk: China revenue (30% of mix) rolling over due to export controls or subsidy shifts before the non-China AI revenue ramps fully to replace it.
The Tell: When asked about HBM revenue sizing, the CEO admitted they 'can't now know' the split because the customer is using the LSA tool for 'most advanced generic memory' as well. This reveals the tool has graduated from a niche HBM application to a standard process of record for advanced DRAM.
Friction Level: MODERATE_FRICTION — Bulls see 'sold out' as pricing power and backlog visibility. Bears see it as a hard revenue cap that limits upside regardless of demand.
Report not found
The report data is no longer available. Please return to the archive.