Analysis
Three companies beat Wall Street's numbers this week and got punished anyway. Marvell delivered second-quarter revenue up 37% year over year to $2.74 billion, guided to roughly 50% growth for the current quarter, and raised its full-year targets -- then fell 6% after running up 184% year to date heading into the print, 24/7 Wall St. reported. Rubrik beat revenue estimates by $31 million and posted 22 cents in EPS against a 4-cent consensus, then fell as much as 9.3% after hours once investors focused on guidance that came in below expectations. SentinelOne beat on revenue too, and still fell roughly 4% in premarket trading after forecasting third-quarter and full-year profit below Street numbers.
I read this as the clearest signal yet that the market has stopped rewarding "beat and raise" on its own for anything priced adjacent to the AI trade -- it now wants beats large enough to justify valuations that already assume years of hypergrowth. Marvell trades near a 60x forward multiple after its 184% run; a quarter that would have been a blowout at a normal multiple reads as merely fine against that bar, and "merely fine" gets sold. That is a market pricing perfection, not growth, and perfection is a much less forgiving standard to clear quarter after quarter.
The AI-bubble-fear framing
All three of these companies sit adjacent to, not squarely inside, the core AI infrastructure trade -- Marvell in custom silicon, Rubrik in data protection increasingly marketed around AI-driven threats, SentinelOne in AI-powered endpoint security. That adjacency is exactly why they're getting hit while Nvidia and CoreWeave, sitting at the center of the trade, rallied hard on their own results the same week. Investors appear to be drawing a sharper line than they have all year between core AI infrastructure beneficiaries and the wider penumbra of companies telling an AI growth story without Nvidia's direct exposure to the capex cycle.
Room for disagreement
The counterargument is that this isn't fear, it's normal post-earnings profit-taking after outsized pre-print run-ups -- Marvell's 184% year-to-date gain alone explains a 6% pullback on an in-line-but-not-spectacular guide without needing a bubble narrative at all. Rubrik's guidance cut and SentinelOne's soft outlook are company-specific execution questions, not evidence of a sector-wide repricing; bundling three unrelated earnings reactions into one "AI capex skepticism" story risks over-reading noise as signal. It's also worth noting all three stocks remain up meaningfully for the year even after this week's declines -- this is a valuation reset at the margin, not a crash.
Where I land: neither story alone explains the pattern across three unrelated companies in the same week. Some of this is genuine multiple compression on AI-adjacent names that ran too far, and some of it is company-specific. The next earnings cycle, not this one, tells you which explanation was right.
