The AI Trade Has Three New Problems

SpaceX/SPCX is already roughly 21% below its June 18 level, leaving the average post-IPO buyer at a double-digit loss after a 19.2% first-day pop. The article is constructive on AI/data-center earnings, citing FactSet's 59.6% expected Information Technology EPS growth in Q2 and a 26.5% projected 12-month gain, but it flags rising policy risk from proposed AI taxes and a possible Fed hike. Minneapolis Fed President Kashkari became the first voting FOMC member to publicly call for a rate hike by year-end, reflecting a more hawkish and uncertain policy backdrop.

Analysis

The real signal here is not that AI is still strong; it’s that the market is transitioning from a pure multiple-expansion trade to one increasingly constrained by cost of capital and policy risk. If rates stop falling—or worse, move up—the long-duration beneficiaries of AI capex face a double hit: financing costs rise while investors re-rate backlog-driven growth against a higher discount rate. That makes the AI winners more path-dependent over the next 1-3 months than the bullish earnings revisions alone suggest.

Within semis, the second-order winner/loser split is sharpening. Names with near-term cash generation and direct demand visibility should continue to out-execute the broader AI basket, while more levered or narrative-heavy plays become vulnerable to any disappointment in data-center spend cadence. The fact that EPS revisions are broadening across the group is supportive, but it also increases crowding risk: when estimate revisions are already widely shared, the market usually starts paying less for each additional beat.

The political overhang is being underestimated because investors are treating it as a 2026+ issue rather than a live discount-rate input. Even if the most aggressive proposals never pass, they can still raise the implied tax/regulatory take rate on mega-cap AI cash flows and compress terminal value assumptions. That matters most for the highest-duration AI compounders and least for shorter-cycle suppliers with diversified end markets.

The Fed angle is the cleaner near-term catalyst. If incoming inflation prints validate the hawkish turn, the market will likely rotate away from crowded growth and into balance-sheet strength, with semis that have real earnings today outperforming pure AI infrastructure proxies. Conversely, a soft inflation streak would quickly unwind the hike narrative and re-ignite the AI bid, so the next 2-3 CPI/PCE prints are the key binary window.

Key Decisions for Investors

• Stay long NVDA on weakness but reduce gross exposure via covered calls into the next 2 inflation prints; thesis is intact on earnings momentum, but upside is more likely to be multiple-expansion-limited if rates back up.
• Pair long INTC / short a basket of high-duration AI infrastructure names for 4-8 weeks; INTC benefits from estimate reset and mean reversion, while the short leg is exposed to policy-rate and tax-risk compression.
• Use SNDK as a relative-value expression versus NVDA: long SNDK if the market starts rewarding less-crowded earnings acceleration over consensus-owned AI leadership; target 10-15% upside with tighter stop if AI capex data softens.
• If the next CPI/PCE comes hot, add a tactical short in QQQ via put spreads expiring 30-45 days out; risk/reward favors hedging because the market is underpricing the chance of a hawkish repricing.
• Avoid adding fresh long exposure to the most politically visible AI megacaps until after the summer inflation window; policy headlines are unlikely to matter on passage, but they can still cap valuation multiples immediately.