3 Artificial Intelligence (AI) Stocks Warren Buffett Might Buy If He Were a Tech Investor

The article argues that Nvidia, Microsoft, and ASML would be Buffett-style AI holdings because of their moats, profitability, and durable demand exposure. It cites Nvidia's $216 billion in trailing-12-month revenue and $120 billion in net income, Microsoft’s more than $305 billion in revenue over the last four quarters with about a 40% margin, and ASML’s 16% sales growth last year and 30% margin. Overall it is a bullish stock-picking piece, but it is commentary rather than new company-specific news.

Analysis

This reads less like a stock-picking note and more like a hierarchy of AI toll collectors. The real second-order insight is that capital spending on AI is increasingly shifting from “model winners” to the picks-and-shovels stack: NVDA monetizes compute intensity, MSFT monetizes distribution and workflow lock-in, and ASML monetizes semiconductor complexity. That creates a durable funnel where hyperscaler capex can be strong even if end-demand for AI applications remains uneven, because each incremental AI workload still forces more spending across the chain.

The market is likely to keep assigning a premium to businesses with both moat and visible earnings conversion, but the dispersion matters. NVDA is the most exposed to a future multiple reset if capex growth decelerates even modestly, because its earnings base is so powerful that any growth disappointment shows up immediately in forward P/E compression. MSFT is the most defensible “compounder” because AI is being layered into an existing enterprise spend line rather than sold as a standalone new category; that makes adoption easier to finance and slower to churn. ASML is the cleanest long-duration industrial scarcity trade, but its catalyst cadence is lumpy, so the stock may lag on near-term headlines even while fundamentals compound underneath.

The contrarian point is that the article implicitly treats AI infrastructure demand as quasi-linear, but the path will likely be cyclical: every wave of GPU and lithography ordering eventually runs into digestion periods. The near-term risk is not that these businesses are bad, but that expectations have moved faster than realized supply-chain throughput and customer ROI. If enterprise AI monetization stalls over the next 2-4 quarters, the market will punish the most expensive parts of the stack first, even if the long-term thesis remains intact.

Key Decisions for Investors

• Long MSFT / short NVDA as a 6-12 month relative-value trade: MSFT has lower execution risk and better downside capture if AI capex growth normalizes, while NVDA is more vulnerable to multiple compression after any guide-down; target 15-20% relative outperformance if the market shifts from scarcity premium to durability premium.
• Buy ASML on 3-6 month weakness rather than strength: use any order-intake or macro-driven pullback to build a position, because the stock’s best risk/reward is when the market discounts the next capex digestion phase; upside should come from multi-quarter scarcity re-rating, not headline momentum.
• Sell out-of-the-money NVDA calls against a core long into strength: the business remains structurally strong, but implied expectations are elevated; covered-call structures can monetize the probability of a near-term valuation pause while preserving most of the long-term AI upside.
• Pair long MSFT / long AAPL with a tilt toward MSFT over 12 months: if AI upsells materially reaccelerate enterprise software budgets, MSFT should widen the quality premium versus AAPL, which remains more exposed to a slower innovation cadence and less direct AI monetization.
• Monitor for a broader AI capex air pocket in 1-2 quarters: if hyperscaler spend growth rolls over, fade the entire infrastructure basket temporarily and rotate into cash-generative software; the first-order winners can become the first-order de-raters when order growth normalizes.