Marvell Leads AI Stocks Lower After Earnings That Narrowly Topped Estimates

Marvell shares fell nearly 6% after the company reported Q3 revenue of $2.74B and adjusted EPS of $0.94, narrowly beating consensus ($2.71B and $0.93). The midpoint of its Q3 outlook also slightly topped estimates ($3.15B revenue and $1.10 adjusted EPS), but investors wanted more upside from the Google custom chip ramp. While management said AI demand remains “exceptionally robust” and lifted revenue forecasts, Jefferies called the lack of additional upside from the Google deal “somewhat disappointing,” contributing to the AI stock selloff.

Analysis

The key read-through is not that AI demand weakened; it is that the market is no longer rewarding merely confirming demand. That matters because semis tied to incremental AI customization are now trading on the size of the second derivative, not the base trend, so any “good but not great” guide can trigger multiple compression across the basket. In the near term, MRVL is the most vulnerable because its premium was built on the idea that custom silicon would be a clean acceleration story; this print suggests the monetization curve may be lumpier than consensus expected.

Second-order, the relative winner may be GOOGL rather than the chip vendors: if custom silicon lowers inference cost, the economic gain accrues upstream to the platform owner even when the supplier’s headline revenue beats are modest. NVDA is less exposed on fundamentals than on positioning; it remains the liquid hedge for any AI selloff, but its earnings power is still tied to the broadest capex pool rather than a single customer. AMD, INTC, MU, and SNDK are likely to trade as sympathy shorts if factor flows continue, even though the fundamental linkage is weaker and mostly through sentiment and basket de-risking.

The catalyst path is two-stage: over days, this is mostly a momentum/factor unwind; over 1-3 months, the market will test whether hyperscaler capex commentary and next-quarter guides re-accelerate enough to justify current multiples. The contrarian view is that the move may be partially overdone: MRVL’s stock had already priced a lot of perfection, so a reset in expectations could be healthy if the back-half ramp actually shows up. What would falsify the bearish read is any evidence that custom chip revenue inflects sharply next quarter or that GOOGL/NVDA capex commentary re-accelerates enough to lift the entire AI semi complex.

Key Decisions for Investors

• Tactically short MRVL into post-earnings strength only if the stock fails to reclaim the first-day breakdown level; 1-3 week horizon, looking for a 5-8% follow-through lower, with a tight stop on a decisive move back above the gap.
• Pair trade: long NVDA / short MRVL for 1-3 months. Thesis: the market is punishing the most narrative-sensitive custom-chip name, while NVDA remains the cleanest expression of broad AI capex; target relative outperformance if the AI basket stays choppy.
• Relative-value long GOOGL / short MRVL over 1-3 months. If custom silicon is economically real, GOOGL captures the margin benefit while MRVL bears the execution risk; this works best if the market starts rewarding platform economics over supplier optics.
• Avoid chasing AMD, INTC, MU, and SNDK on this headline alone; use them as sentiment hedges only if the next hyperscaler commentary misses, because the fundamental signal here is weak and the main effect is basket de-risking.
• Set an alert for the next hyperscaler capex revisions and MRVL’s next guide. If MRVL shows a sharp back-half revenue inflection or GOOGL raises AI infrastructure spend again, cover shorts and rotate back into the AI semi basket.