RESEARCHAMD
AMD: right franchise, wrong point in the capex cycle
No argument with the business. The entry is the problem: the multiple already discounts two more years of a capex cycle that has historically run three.
Evidence · 3 cards
The economics turn on utilisation. Fixed cost per wafer is set at the capacity decision; every point of utilisation above the planning assumption drops through at close to full incremental margin, and every point below does the reverse.
Two procurement disclosures make the case. Both of the largest buyers have now confirmed multi-vendor accelerator strategies in their own filings, and neither framed it as a cost decision.
The position: no call. This is a name I want to own and not at this level, so there is nothing locked here and nothing for the record to grade. If the entry improves I will say so in a piece that does carry a call.
Export controls are the risk that is not hedged here. A restriction round that removes a major end market would overwhelm the company-specific argument entirely, and there is no position size that makes that survivable.
Discussion · 0
No comments yet.