ARM, Marvell, or Vertiv: Which AI Infrastructure Stock Crushed It in April?

AI infrastructure stocks posted a wide April dispersion, led by Marvell with a 67% gain, followed by Arm at 39% and Vertiv at 31%. Marvell's rally was driven by custom silicon design-win-to-volume conversion, while Arm's royalty model and Vertiv's data center power/thermal gear also benefited from hyperscaler AI capex. Vertiv reported Q1 revenue of $2.65B, up 30.1% YoY, and raised full-year guidance to $13.5B-$14B in net sales and $6.30-$6.40 in adjusted EPS.

Analysis

The key takeaway is that the AI infrastructure trade is fragmenting into three distinct monetization speeds, and the market is paying up most for the fastest conversion from capex intent to recognized revenue. That favors custom silicon designers over IP licensors and hardware enablers over both when deployment is still in the early ramp phase. In other words, the current leader is not necessarily the best secular compounder; it is the name with the shortest path from hyperscaler budget to P&L.

The second-order implication is that the next leg of the trade may be determined less by AI demand direction and more by supply-chain execution risk. If hyperscalers are still shifting designs, the winners are exposed to program concentration and schedule slippage, while the losers can look cheaper on a forward basis precisely when their royalty or equipment revenue is about to accelerate. That creates a likely rotation window over the next 1-3 months: momentum can remain with the most crowded name, but the earnings quality may actually improve further down the stack as volume broadens.

The market may also be underestimating how much of the recent move is already a positioning event rather than a fundamental re-rate. When sentiment reaches extreme bullishness and then rolls over late in the month, the risk is not that AI spending disappears, but that the stock needs a digestion period even if the business stays strong. The more fragile setup is the name with the largest short-term move and the highest dependence on a small number of design-win headlines; the most durable setup is the one with recurring royalty economics and broader customer penetration.

For the broader AI complex, this is mildly negative for direct compute incumbents if custom silicon continues to gain share, because it implies hyperscalers are increasingly optimizing away from merchant GPU dependence. That does not break the AI spend thesis, but it does change where the economic surplus accrues. The beneficiaries are the picks-and-shovels names that scale with deployment density and the IP owner that monetizes every alternative architecture, not just the headline winner of the month.

Key Decisions for Investors

• Trim MRVL into strength and avoid chasing after the 67% monthly move; use any 5-10% pullback over the next 2-4 weeks to re-enter only if hyperscaler commentary confirms new design-win conversion, with a tight stop if sentiment turns while volume estimates flatten.
• Long ARM vs short a basket of high-multiple AI infrastructure momentum names for a 1-3 month mean-reversion trade; ARM has more durable royalty compounding and less headline risk, while the basket remains vulnerable to digestion after outsized price action.
• Buy VRT on pullbacks as a slower-but-lower-volatility beneficiary of AI deployment density; favor call spreads 2-4 months out rather than outright stock if you want exposure without paying full momentum premium.
• Pair long ARM / short NVDA for a tactical hedge against merchant GPU share pressure from custom silicon adoption over the next 1-2 quarters; the thesis works best if hyperscaler capex remains firm but architecture mix continues shifting away from off-the-shelf accelerators.
• If staying long MRVL, finance it with short-dated calls sold against the position or a call spread to reduce gamma risk; the near-term risk/reward is unfavorable after the run, but the medium-term upside remains intact if TPU-related volume conversion is validated in May/June.