Billionaire Dan Loeb of Third Point Dumped Nvidia, Broadcom, and Meta, but Increased His Stake in This Virtual Monopoly by Almost 500%

Third Point’s Dan Loeb trimmed AI bellwethers—selling 190,000 Nvidia shares, 50,000 Broadcom shares, and 90,000 Meta shares in Q2—while adding 850,000 shares of Alphabet (GOOGL), nearly 6x (+489%) his position. The article ties Alphabet’s rebound appeal to AI-driven growth at Google Cloud, with Google Cloud revenue up 82% Y/Y to $24.8B and operating margin at 36% (+15pp Y/Y), plus Q2 Alphabet revenue up 24% to $119.8B (+$2.8B beat). Overall, the Form 13F flow is modestly constructive for Alphabet given the strong fundamentals, though the selling in other AI names suggests rotation/profit-taking rather than a broad AI selloff.

Analysis

The signal here is less about one manager’s taste and more about factor rotation: the market has crowded into “AI picks-and-shovels,” while the monetization layer is still under-owned. That creates a near-term setup where semis can keep compounding fundamentals, but multiples are vulnerable to any deceleration in hyperscaler capex or a shift from “build now” to “harvest now.” GOOG/GOOGL is the cleaner beneficiary because it can fund AI spend internally and monetize it across search, YouTube, and cloud without needing external capital.

Second-order, Alphabet’s strength pressures the rest of the ad stack and AI-adjacent platforms: if AI improves query quality and ad ROI, budget share can migrate away from weaker ad franchises and toward the platform with the broadest distribution. META is not broken, but it becomes more of a competitive check than a direct AI winner unless engagement gains clearly outrun incremental inference costs. NVDA/AVGO remain structurally strong, yet their stocks are most exposed to “good news fatigue” because expectations already discount several years of exceptional demand.

Contrarian view: the consensus is still underappreciating duration. A 13F is stale and often reflects profit-taking rather than a true bearish thesis, so the immediate signal for NVDA/AVGO may be overstated. What would falsify the long-GOOG case is any evidence that AI search cannibalization overwhelms ad pricing power or that cloud growth re-accelerates only with margin destruction, not volume; absent that, the more likely path is multiple expansion in the cash-rich monopolist and range-bound action in the crowded infrastructure names over the next 1-3 months.

Key Decisions for Investors

• Long GOOGL / short NVDA as a 1-3 month relative-value rotation trade: buy the monetization layer and fade crowded AI infrastructure; thesis works if capex commentary cools and GOOGL keeps showing cloud margin leverage.
• Long GOOG on post-13F dips rather than chasing strength; target a 6-18 month hold where cash flow can self-fund AI optionality. Falsify if cloud growth slows materially or search monetization weakens.
• Reduce overweight exposure to NVDA and AVGO on strength or use call spreads instead of outright longs; the risk/reward is less attractive if the market starts pricing a capex plateau before fundamentals roll over.
• For those already long META, hedge with a partial short GOOG/GOOGL basket only if ad auction data shows share loss; otherwise META looks more like a competitive laggard than a direct short.
• Watch the next hyperscaler capex updates as the key catalyst. If 2-3 of the major buyers signal a normalization in spend, expect a 10-15% multiple compression risk in semis even if earnings remain solid.