Billionaire Ray Dalio Just Warned of an AI Bubble, but His Top Stock Holdings Are in These 6 AI Names

Ray Dalio warned that AI stocks may be in a bubble, but Bridgewater remained heavily exposed to the group at the end of Q1, with its top six holdings all tied to AI. He increased stakes in Amazon to 4.1% of portfolio value, while also adding 50% to Alphabet, 10% to Microsoft, 20% to Nvidia, more than 50% to Broadcom, and 65% to Micron. The article is primarily a positioning and valuation commentary piece rather than new company-specific fundamentals.

Analysis

The key market signal is not that AI is expensive; it is that the capex cycle is still being funded by balance-sheet giants with durable cash flows. That means the near-term winners are the “picks and shovels” suppliers that sit inside hyperscaler budgets, while the first-order loser is any marginal AI startup or software vendor forced to compete against incumbents that can subsidize inference costs for years. The second-order effect is a widening gap between compute demand and monetization: spend can accelerate faster than revenue conversion, which eventually pressures procurement discipline across the stack.

The most interesting part is the divergence between infrastructure enthusiasm and eventual spend scrutiny. If cloud leaders keep adding capacity while utilization lags, the market will eventually force a reset from “build for scarcity” to “prove ROI,” which is when semis with the most levered expectations tend to de-rate fastest. Memory is the cleaner near-term trade because HBM remains structurally tight, but that also makes MU the most cyclical of the group once inventory builds normalize.

The contrarian setup is that consensus is treating AI as one trade, but it is really three different ones: cloud monetization, custom silicon, and memory bottleneck relief. The first two are multi-year compounding stories; the third is a classic shortage supercycle that can unwind faster than investors expect. A sentiment-driven pullback in NVDA or AVGO could be an attractive entry point, but only if hyperscaler capex commentary stays intact over the next 1-2 quarters.

A subtle read-through is that the market is underpricing the optionality in Amazon and Alphabet relative to their self-funded AI deployment. They can absorb a capex miss without impairing capital returns, whereas vendors upstream rely on continued order growth to justify valuation. That asymmetry argues for owning the beneficiaries with internal financing power and being more selective on the highest-multiple hardware names.

Key Decisions for Investors

• Go long GOOGL and AMZN on any 5-8% pullback over the next 2-6 weeks; prefer these over pure-play semis because they can self-fund AI capex and still defend FCF, giving better downside protection if the spend cycle slows.
• Maintain a tactical long NVDA / short basket of AI-adjacent software names for 1-3 months; the thesis is that infrastructure demand remains real while software monetization is still being forced through a narrower set of winners.
• Use AVGO as a relative-value long versus NVDA on 3-6 month horizon; AVGO has more diversified AI exposure through custom chips and networking, which should hold up better if GPU growth decelerates.
• Treat MU as a momentum trade, not a core compounder: buy only on inventory-driven dips and use trailing stops, since memory has the best near-term pricing power but the sharpest reversal risk over 6-12 months.
• If hyperscaler capex commentary weakens, rotate out of NVDA first and into MSFT/GOOGL; MSFT/GOOGL have stronger platform monetization and less valuation sensitivity to a 1-quarter slowdown in hardware orders.