Dan Loeb’s Third Point expands AI bets with new stakes in Hut 8 and chip names

Third Point added new positions in ASML, Lam Research, KLA, Broadcom, the VanEck Semiconductor ETF, TransDigm, and Hut 8 in the first quarter, signaling continued emphasis on the AI trade. The fund also cut Taiwan Semiconductor by 35% and CRH by 27%, while Amazon remained its largest equity holding despite a 10% reduction. The filing reinforces Dan Loeb's increasingly bullish stance on AI beneficiaries and infrastructure themes.

Analysis

The key signal is not the individual adds but the portfolio construction: this is a late-cycle AI barbell, with direct semicap exposure on one side and infrastructure/power beneficiaries on the other. That tends to work when the market is still rewarding “picks and shovels,” but the second-order risk is factor crowding — if hyperscaler capex slows even modestly, the multiple compression hits equipment names first, while the more cash-generative names with service/installed-base exposure should hold up better.

The small size of the new positions matters. It suggests optionality rather than high-conviction replacement of the core book, which means the flow impact is more important for sentiment than fundamentals. In that context, HUT is the most asymmetrical expression here: it is really a levered proxy on power scarcity, data-center adjacency, and crypto beta, so it can outperform in momentum markets but is also the first name to de-rate if the market rotates from “AI power demand” to “show me realized earnings.”

CRH and AMZN reductions are a useful tell. Cutting a building-materials name alongside adding AI beneficiaries implies an implicit bet that capital is shifting from old-economy capex to digital infrastructure capex; that is bearish for broad construction/industrial cyclicals relative to semis and grid-related assets. Meanwhile AMZN’s trim does not read as a bearish fundamental call so much as a relative-value rotation away from a name where AI benefit is already crowded and partially embedded in valuation.

The contrarian angle is that the semicap trade may be too consensus, but the less-consensus expression is the infrastructure tail: power, interconnect, and data-center suppliers are still under-owned versus the obvious chip winners. The market may be overestimating how linear the AI capex curve is over the next 6-12 months and underestimating the lag in monetization, which could create a window where equipment order data stays strong but end-market spending broadens slowly. That favors being selective on upstream semis and more aggressive on names tied to bottlenecks rather than pure AI narrative exposure.

Key Decisions for Investors

• Long AVGO / short SOXX for 1-3 months: express preference for diversified semis with software/service content over pure equipment beta; if AI capex decelerates, AVGO should hold up better on earnings quality while the ETF compresses faster.
• Buy HUT call spreads 3-6 months out: use a defined-risk structure to capture momentum from AI power-demand narrative, but cap upside because the name can retrace sharply if crypto beta or risk appetite fades.
• Pair long TDG vs short CRH over 2-4 months: TDG benefits from aerospace aftermarket pricing power and secular fleet utilization, while CRH is more exposed to slowing non-resi construction and a potential rotation out of cyclicals.
• Initiate a tactical long LRCX/KLAC basket on a 6-8 week horizon only after confirming capex commentary from hyperscalers: the setup is favorable, but entry should be delayed until order visibility improves to avoid paying peak sentiment multiples.
• Fade AMZN strength with a small put spread into the next earnings cycle: the trim suggests less incremental conviction at current positioning, and the stock is vulnerable if AI spend remains capital-intensive without immediate margin expansion.