Palantir director Alexander Moore sells over $2.1m in stock

The market is starting to separate AI monetization from AI exposure. That is usually the point where the cheapest cash flows in the chain get hit first: semis and the large-cap platforms have the most crowded ownership, the highest embedded growth expectations, and the most obvious need to justify incremental capex with near-term revenue. The important signal is not one director sale; it is that buyers are demanding proof of deployment, data control, and workflow ROI instead of paying up for generic AI adjacency.

That shift favors names with a defendable implementation moat and hurts names whose AI story is still mostly spend-led. PLTR can still outperform if it keeps converting regulated/private-cloud use cases into repeatable bookings, because that is the part of the stack that benefits when customers get more skeptical about frontier-model economics. RXT is a plausible secondary winner only if its partnership pipeline turns into recurring revenue; otherwise it remains a low-quality way to express the theme and is more likely to trade on liquidity than fundamentals.

The contrarian view is that this is less an AI-bubble break and more a repricing of duration: the market is punishing timelines, not the category itself. If hyperscaler capex guides stay intact and next-quarter enterprise budgets do not roll over, the rotation can reverse quickly. What would falsify the bearish infrastructure view is any combination of re-accelerating cloud spend, stronger AI monetization commentary from META, or a clear inflection in PLTR bookings that proves the deployment layer is scaling faster than expected.