He Predicted the Dot-com Bubble Burst. Now He's Saying SpaceX Could be a Fresh Warning Sign.

Jeremy Grantham (GMO) warns the market may be approaching the end of an AI-driven bubble, citing extreme valuation levels and the huge Space Exploration Technologies (SPCX) IPO that raised about $75B (about $86B including overallotment) as a potential late-cycle signal. While he stresses timing is uncertain, the article frames AI-related exposure as a key risk and suggests reducing AI concentration or rotating toward more resilient sectors/value funds rather than panic-selling.

Analysis

This is mainly a positioning and sentiment warning, not a near-term fundamental break. The market mechanism matters: when a handful of AI beneficiaries become crowded ownership, marginal bad news tends to hit multiples first, then only later the earnings estimates. That means the first casualties are likely the highest-duration names and any adjacent “story stock” financing vehicles, not the profitable franchise leaders with real free cash flow.

The second-order effect is a rotation, not necessarily a crash. If investors de-risk from AI beta, capital should migrate toward sectors with lower multiple sensitivity and steadier cash yield such as staples, utilities, and value indices. For semis, the vulnerable segment is the group priced on 2027+ demand assumptions; if capex commentary from hyperscalers turns even slightly more disciplined over the next 1-2 quarters, the de-rating could be sharp. By contrast, the broad market may absorb an AI unwind better than consensus expects if earnings breadth holds outside the theme.

The contrarian point is timing: bubble calls are usually early, and the market can keep rewarding the same narrow leadership until a real catalyst appears. The clean falsifier is continued upward revisions to AI infrastructure spending and no slowdown in order growth through the next two earnings cycles. Until then, this reads more like an invitation to reduce exposure than an all-in short.

Key Decisions for Investors

• Trim NVDA exposure into strength over the next 1-2 weeks; the risk/reward is unfavorable if the stock is already pricing another year of perfect capex demand. Falsify the caution if data-center order commentary re-accelerates in the next earnings print.
• Pair trade: short a basket of high-multiple AI infrastructure names vs long XLU or XLP for 1-3 months. The thesis is multiple compression, not earnings collapse; expect defensives to outperform if sentiment sours even modestly.
• Use QQQ put spreads or a small Nasdaq hedge into the next major mega-cap earnings window. This is a hedge against crowded positioning, with best payoff if guidance disappoints rather than on the headline itself.
• Avoid chasing speculative private-market/IPO proxies like SPCX until there is public-market monetization evidence. These are late-cycle sentiment vehicles; if they keep pricing aggressively, it would weaken the bearish read.