Billionaire Investor David Tepper's Top 5 Stocks

This basket is signaling that AI alpha is migrating from “who has the best model” to “who controls bottlenecks and monetization.” The cleanest winners are the toll collectors and the platforms with non-AI cash engines that can absorb capex without destroying FCF; the more fragile names are the ones whose economics depend on every incremental compute dollar being perfectly efficient. That argues for owning the constraint owners over the pure exposure names, because the former can reprice scarce supply while the latter are still competing on product cycles.

Near term, the market will probably keep rewarding anything tied to confirmed demand, but the first reversal catalyst is always capex discipline. If hyperscaler spending growth slows even modestly over the next 1-3 quarters, accelerator multiples can compress faster than memory or foundry because the stock market discounts “peak build” long before reported revenue rolls over. The real check is whether orders remain pre-committed; if lead times shorten or pricing softens, the scarcity premium disappears quickly.

The contrarian read is that consensus is still underweight the application layer’s operating leverage. UBER looks more asymmetric than the market gives it credit for because autonomous adoption can reduce unit cost without requiring it to be the capital spender in the stack. By contrast, AMZN and GOOG may continue to execute but still trade with a capex overhang until the market trusts that incremental spend is translating into durable margin expansion rather than just bigger infrastructure.

Over 6-18 months, the biggest risk to the bullish AI trade is supply normalization: once HBM and advanced-node capacity catch up, the margin transfer from customers to suppliers slows. That makes the best longs the names with recurring demand, pricing power, and diversified monetization, not the ones that rely on perpetual shortage.