Jim Cramer explains why no AI stock seems safe from the market carnage

Uncertainty over data-center construction and regulatory hurdles drove a sharp two-session sell-off across AI-linked stocks: SOXX fell 3% Wednesday and 10% Thursday, while GE Vernova dropped 7% and Sandisk 22% Thursday. Jim Cramer said valuations depend on whether data-center build-out slows materially, despite Nvidia trading at about 12 times next year's earnings with expected growth of 70%; he also questioned AMD's valuation after its shares rose from $188 in March to $404. Micron's take-or-pay contracts offer revenue visibility, but Cramer raised concerns about customers' ability to pay in severe distress.

Analysis

The key distinction is whether permitting friction reduces ultimate compute demand or merely shifts its timing and location. A delay can still hurt near-term earnings: customers may defer orders, leave suppliers with underused capacity, and force estimates lower before long-run demand is disproved. That makes forward P/E a poor downside anchor for NVDA and AMD if the earnings denominator is built on uninterrupted deployment. MU’s take-or-pay terms may soften volume risk, but do not eliminate counterparty-credit risk; verify contract counterparties, cancellation provisions, and customer concentration before treating them as a floor.

The sell-off may be extrapolating local approval constraints into a broad structural cap on AI spending. If projects are delayed rather than cancelled, utilization and efficiency upgrades can absorb some spend, while grid and power constraints may shift the mix and timing of supplier orders rather than erase them. Conversely, GEV and other infrastructure-linked exposure can be hit even if chip demand holds up, because project timing affects order conversion. SNDK’s sharper move signals elevated sensitivity, but the article alone does not establish whether fundamentals changed.

Near term, positioning and estimate revisions can dominate valuation; over 1–3 months, hyperscaler capex plans, permitting/interconnection decisions, and supplier order commentary matter more. Over 6–18 months, actual project cancellations, utilization, and contract performance determine whether this is a timing shock or a lower-growth regime. The contrarian risk is that a genuine build-out slowdown is underpriced because investors anchor on long-run AI demand; the opposing risk is treating delays as cancellations. The CNBC commentary is not independent evidence of either outcome.

Key Decisions for Investors

• Avoid adding broad AI beta solely because forward multiples look low. Reassess NVDA and AMD exposure against changes in customer capex plans and forward earnings estimates; a valuation-based long thesis is falsified by sustained estimate cuts or explicit project cancellations.
• For a tactical hedge, consider a defined-risk SOXX put spread only if the ETF fails to stabilize and permitting or capex news worsens; avoid an unhedged short after a sharp sell-off. Exit the hedge if orders and customer spending remain intact and the ETF recovers its breakdown level.
• Keep MU contract visibility conditional: seek disclosure on counterparties, enforceability, and receivables before relying on take-or-pay protection. Deteriorating customer credit or evidence of renegotiation would strengthen the downside case; intact collections and firm deliveries would weaken it.
• Treat SNDK and GEV as watchlist names, not automatic shorts. Compare subsequent guidance, order conversion, and estimate revisions with the price declines; absent fundamental deterioration, further weakness may reflect positioning rather than a durable change in demand.