AMD is being re-rated on a mix-shift story, but the key market mechanism is margin durability, not just TAM expansion. If CPUs truly take a larger share of AI buildouts, AMD gets higher attach rates in data-center budgets and better operating leverage; the relative loser is NVDA at the margin, while INTC faces another layer of share pressure in the highest-value socket.

The second-order effect is on hyperscaler economics: lower-power CPU-heavy architectures can improve inference ROI for GOOGL and META, which may extend AI capex duration even if GPU intensity moderates. That is bullish for the semiconductor ecosystem broadly, but it also means the stock is pricing a multi-year narrative before the next few quarters can validate it.

Contrarian view: the consensus may be underestimating substitution into custom silicon and in-house CPU designs, which could cap AMD’s pricing power even if unit growth stays strong. The stock is vulnerable to multiple compression if AI spend pauses, because a premium valuation leaves little room for any guide miss or margin disappointment. Falsify the thesis if the next two earnings cycles fail to show durable server CPU mix gains or if data-center gross margin stops expanding.