71.6% of Nvidia's Portfolio Is Invested in These 2 Top Artificial Intelligence (AI) Stocks

The article highlights Nvidia’s AI-related investments in Intel and CoreWeave, which together made up 71.6% of its portfolio as of March 31. Intel’s revenue rose 7% to $13.6B and adjusted EPS climbed 123% to $0.29, but the stock has already surged nearly 500% over 12 months and looks expensive at 151.5x forward earnings. CoreWeave posted $2.1B in revenue, up 111.6%, but remains unprofitable with a $99.4B backlog and high customer concentration, including Microsoft at 67% of revenue.

Analysis

The market is starting to price AI infrastructure as a secular utility, but the setup is bifurcating: asset-light demand capture is being rewarded, while capital-intensive capacity is being treated like a quasi-financing trade. That matters because the second-order winners are not just the headline AI names, but also the component suppliers and software layers that monetize every incremental rack deployment without taking utilization risk. In that frame, the relative opportunity is better in picks-and-shovels exposure to the buildout than in paying up for the operators shouldering balance-sheet risk.

The biggest hidden risk is that AI capex remains front-loaded while revenue monetization lags by multiple quarters, especially if enterprise agent adoption proves less linear than investors assume. If utilization doesn’t inflect fast enough, margins in the infrastructure layer can compress abruptly even when top-line growth remains strong, which is why the market is likely to punish any sign of backlog conversion slowing or customer concentration worsening. For the CPU side, the bullish case depends on AI agents creating sustained orchestration demand; if that workflow shift is delayed, the current multiple leaves very little room for disappointment.

Consensus appears to be underestimating how much of the upside is already embedded in the most obvious beneficiaries and overestimating the durability of “infinite AI demand” as a near-term equity narrative. The cleaner expression is to own the enablers of scale where competitive position is improving but valuation is still reasonable, and fade the names where expectations have run ahead of operating proof. The most interesting trade is not outright bearish AI, but a rotation within AI from expensive optionality toward cheaper, higher-throughput beneficiaries with less execution risk.

Key Decisions for Investors

• Prefer long AMD vs. INTC on a 3-6 month horizon: AMD benefits from continued server CPU share gains while INTC already discounts a multi-year turnaround; risk/reward is better on the relative winner than on paying peak multiple for the laggard.
• Avoid chasing NVDA's ecosystem beneficiaries indiscriminately; use it as a quality filter and focus long exposure on suppliers with recurring demand and less capex intensity rather than direct AI operators with utilization risk.
• Consider a tactical long CRWV only on a 10-15% post-event pullback, sized small: upside is tied to backlog conversion, but the downside is asymmetric if enterprise demand or a large customer slows; use a tight risk budget.
• Pair trade: long AMD / short INTC for 6-12 months if valuation spread stays wide; thesis is that share gains can continue even if AI compute demand slows, while INTC's multiple leaves limited room for operational stumbles.
• If expressing a broader AI capex view, prefer a basket of non-customer-concentrated infrastructure names over CRWV outright; the cleaner risk/reward is in diversified demand exposure rather than single-account dependence.