The macro read-through is that AI is still acting more like a demand engine than a productivity story, which is supportive for semiconductor capex now but not benign for rates or multiples. That favors the infrastructure layer over application-layer software: near-term budgets keep flowing into compute, while the later-stage payoff needed to justify a broad re-rating has not shown up yet. In other words, NVDA can keep winning even if the broader market eventually has to live with stickier inflation and a higher-for-longer discount rate.

META’s real risk is not the headline settlement cost; it is the embedded operating tax from product constraints that can quietly reduce engagement, ad load, and targeting efficiency over several quarters. The second-order effect is that ad dollars may redistribute toward other large inventory holders and measurement-sensitive platforms if Meta’s youth product is throttled, but the bigger impact is multiple compression if investors conclude the company is trading growth for compliance. This is a months-long earnings/engagement story, not a one-day legal event.

Contrarian view: consensus is likely underpricing the inflationary side of AI capex. If AI is contributing to GDP before productivity shows up, it is acting as a stimulus layer, not a deflationary one, which argues against fading semis too early but also against assuming the whole tech complex deserves multiple expansion. MCO is essentially a monitor here rather than a direct trade; its commentary matters more as a signal for the macro regime than for company fundamentals.