The 'Next Nvidia' Trade? Why Investors Are Suddenly Watching Advanced Micro Devices, Arm Holdings, and Marvell Technology

The article argues that Arm, Marvell, and AMD are positioned to benefit from the AI inference cycle, with projected earnings growth outpacing Nvidia in several periods. AMD says the server CPU market could grow 35% annually through 2030 to more than $120 billion, while Arm sees revenue rising to $25 billion over five years from $4.9 billion last fiscal year. The piece is constructive on these stocks but is largely an opinion-driven comparison rather than a fresh catalyst.

Analysis

The market is starting to price a second-order AI spend cycle: not just more accelerator capex, but more silicon around the accelerators. That is structurally better for ARM and AMD than for NVDA because inference workloads are more power- and cost-sensitive, which shifts wallet share toward CPUs, custom ASICs, and lower-power architectures. In other words, the AI winner set broadens, but the marginal dollar of growth likely migrates to the companies that sit closer to system design and custom silicon economics.

The biggest underappreciated beneficiary is the value chain around hyperscaler differentiation. As the largest buyers increasingly seek to reduce dependency on a single GPU vendor, they will dual-source inference stacks and pull in more custom designs, which helps MRVL and ARM while pressuring legacy x86 share. INTC is the structural loser here: even if it participates in some enterprise server refresh, its share-loss risk rises if AI-driven capex extends CPU demand but AMD captures the incremental socket wins.

Timing matters. The next 1-2 quarters are mostly about sentiment and order visibility; the real fundamental inflection comes over 12-24 months as new inference silicon ramps and customer qualification broadens. The risk to the bullish thesis is that NVDA remains too strong in software and platform lock-in, and that inference efficiency gains reduce total silicon content per workload, limiting the upside to the broader ecosystem.

Consensus may be over-extrapolating earnings growth into durable multiple expansion. These names can grow faster than NVDA in percentage terms from smaller bases, but they are also more execution-sensitive and less monetized in software. The trade is less about replacing NVDA and more about owning the picks-and-shovels around AI infrastructure diversification before that diversification becomes fully priced in.

Key Decisions for Investors

• Go long AMD vs INTC in a 3-6 month pair trade: AMD should keep taking server socket share while INTC faces negative mix and slower AI-related monetization; target 15-20% relative outperformance, stop if Intel-guided share stabilization becomes credible.
• Buy MRVL on a 6-12 month horizon as the cleanest custom-silicon beta; use pullbacks to add, with upside tied to hyperscaler ASIC ramps and downside limited by already-expected AI order momentum.
• Use a barbell long AMD / long NVDA, underweight INTC: NVDA remains the platform anchor, but AMD offers higher near-term earnings torque; this captures AI spend without relying on a single winner outcome.
• If options liquidity is adequate, buy 6-12 month call spreads in ARM to express continued architecture leverage while capping premium decay; favorable if inference and edge power-efficiency narratives persist.
• Avoid chasing near-term multiple expansion in the group after momentum-driven rallies; fade spikes that occur without evidence of design-win conversions, because the key catalyst is revenue realization over the next several quarters, not headline TAM expansion.