CNBC Daily Open: Trump revives his tariff playbook with Canada

This is less about a single product launch than about hyperscalers buying optionality. AMD’s signal matters because it reduces the market’s assumption that AI silicon is a pure winner-take-all oligopoly; once one large customer validates an alternative rack architecture, the pricing conversation with NVDA shifts even if shipment share barely moves at first. Over the next 1-3 months, the key market mechanism is not unit volume but gross-margin dispersion: any evidence that buyers can dual-source at rack level should cap NVDA multiple expansion and give AMD a higher strategic valuation floor.

MSFT and GOOGL are the quieter winners if this turns into a broader procurement trend. Cheaper, more fungible accelerators let them preserve AI investment intensity while improving cloud economics, which is a margin story more than a revenue story. The downside for NVDA only becomes material if the next two earnings cycles show no loss of pricing power and no slowdown in backlog conversion; absent that, this remains a competitive overhang rather than a fundamental break.

The contrarian view is that the market may still be underestimating how quickly hyperscalers will use in-house or alternative silicon to improve bargaining leverage, even without fully replacing NVDA. That argues for lower terminal margins in merchant AI hardware over 6-18 months, but not necessarily lower total AI capex. The falsifier is simple: if NVDA continues to print outsized data-center margin expansion and AMD’s rack-scale rollout slips, the bear case on NVDA is too early and should be covered.