Not Micron. Not Alphabet. Here's My Top Artificial Intelligence (AI) Stock Pick for the Next 3 Years.

The article argues Nvidia is undervalued at 28x earnings despite management guiding for approximately 70% revenue growth next year. It cites Nvidia's GPU leadership, upcoming architecture launch, and projected global data-center capital expenditures of $3 trillion to $4 trillion by 2030 as catalysts for further upside. The view is highly bullish but represents opinion-driven commentary rather than a new company disclosure.

Analysis

The relevant question is not whether NVDA screens inexpensive on a single-year P/E, but whether the market is correctly discounting the post-ramp earnings curve. A valuation re-rating requires evidence that accelerator demand remains capacity-constrained after the next architecture transition and that gross margin holds despite a larger mix of systems, networking, and customer-specific configurations. The article's growth and valuation assertions are promotional rather than a differentiated catalyst; absent a new estimate revision, this is unlikely to move the stock over days.

Over the next 1-3 months, the decisive data are hyperscaler capex guidance, lead times, Blackwell/next-generation shipment mix, and NVDA's implied gross-margin trajectory. If cloud customers shift more workloads to internally designed ASICs, the first impact is likely lower incremental GPU demand and pricing power rather than an immediate revenue cliff; AVGO could capture part of that substitution through custom silicon and networking, while AMD remains the higher-beta share-gain alternative. MU is a second-order beneficiary only if high-bandwidth-memory content and pricing remain tight, but it carries more memory-cycle downside than NVDA.

Consensus may underappreciate that an apparently modest multiple can still compress sharply if earnings estimates peak: AI infrastructure stocks are priced on sustained upward revisions, not on trailing earnings. Conversely, NVDA's moat is better tested by software attachment, networking share, and customer concentration than by headline GPU comparisons. A durable long thesis is falsified by a material reduction in hyperscaler capex plans, sequential data-center revenue deceleration beyond a product-transition quarter, or gross-margin guidance falling below expectations without offsetting volume upside.

For the 6-18 month horizon, the cleanest opportunity is to own the suppliers with verified bottlenecks rather than extrapolate an industry-wide capex total. Monitor HBM supply commitments, advanced packaging availability, and custom-chip program wins: easing constraints can lift unit volumes but also exposes NVDA and MU to price normalization, while successful ASIC deployments would disproportionately support AVGO.

Key Decisions for Investors

• Do not chase NVDA solely on this article. Add only after the next earnings release confirms data-center revenue upside and stable/improving gross-margin guidance; use a 3-6 month horizon and reduce if forward revenue estimates stop rising or hyperscaler capex guidance weakens.
• Consider a 6-12 month relative-value pair: long AVGO / short AMD in equal beta-adjusted dollars. The trade expresses custom-silicon and networking monetization versus a more execution-dependent GPU share-gain story; exit on material AMD AI revenue upside or evidence that hyperscalers defer ASIC programs.
• Maintain MU as a tactical, smaller-sized AI-memory exposure rather than a direct NVDA substitute. Add only if management confirms HBM pricing and supply visibility through the next two quarters; protect against a conventional DRAM/NAND downturn with a stop tied to weaker memory pricing or inventory-build commentary.
• Set an event alert around major cloud-provider earnings: aggregate capex guidance and disclosed AI capacity constraints are the near-term read-through. A broad capex increase without incremental accelerator supply would favor NVDA/MU; capex growth shifting toward internally designed chips favors AVGO.