This Billionaire Dumped Cloud Stocks for These New AI Stocks. Should Investors Follow Suit?

TSMC and ASML are highlighted as key beneficiaries of the AI infrastructure boom, with TSMC benefiting from its near-monopoly in advanced foundry and packaging work and ASML from its monopoly in EUV lithography. The article notes TSMC's pricing power, multiyear price hikes, and new demand from GPUs, ASICs, CPUs, and HBM-related production. ASML is also positioned to gain as foundries and memory makers expand capacity and eventually adopt high-NA EUV tools.

Analysis

The key second-order read is that the AI capex cycle is broadening from “model builders” to “tooling bottlenecks,” which should prolong pricing power for the semiconductor equipment stack even if hyperscaler spending gets choppier. TSMC and ASML are effectively toll collectors on incremental AI capacity: one monetizes wafer starts and packaging complexity, the other monetizes node shrink and tool replacement. That shifts the debate from whether AI demand exists to whether supply-chain capacity can clear without margin leakage — a much better setup for the enablers than for the users.

TSMC’s real edge is not just manufacturing scale; it is the ability to arbitrage uncertainty across GPU, ASIC, and high-performance CPU demand while keeping utilization high. If agentic AI drives more inference at the edge and inside enterprise workflows, the mix likely tilts toward more diversified logic demand rather than a single-winner accelerator market, which further insulates TSM from technology selection risk. The bigger catalyst over the next 6-18 months is packaging intensity and lead-time normalization: any delay in CoWoS expansion would propagate scarcity pricing through the whole AI hardware stack.

ASML’s setup is more convex but timing-sensitive. The market tends to underappreciate that memory capex can be a meaningful offset to logic-cycle volatility, and HBM growth alone can support a longer equipment upcycle than a pure foundry-driven cycle would imply. The near-term risk is less order cancellation than customer pushout on high-NA adoption; if leading-edge yields improve slower than expected, ASML could see a perception gap between secular demand and near-term revenue conversion.

Contrarian angle: the market may be too comfortable extrapolating AI infrastructure growth in a straight line, while the real bottleneck is progressively shifting from demand to tool qualification, export controls, and customer concentration. That creates episodic pullbacks even in secular winners. The more durable edge is to own the picks-and-shovels with the clearest monopoly characteristics, while fading overowned “AI beneficiaries” whose earnings depend on faster monetization than the infrastructure build-out can currently support.

Key Decisions for Investors

• Add TSM on 2-3% pullbacks over the next 2-6 weeks; risk/reward favors buying weakness because packaging and advanced-node scarcity should keep pricing power intact through the next 2-4 quarters.
• Initiate a TSM/AMD relative-value long TSM, short AMD pair for a 3-6 month horizon; the trade benefits if AI demand broadens but supply-chain toll collectors capture more of the economics than chip designers.
• Buy ASML on any post-earnings compression or order-timing miss; use a 6-12 month window and expect the market to rerate once HBM and high-end logic capex translate into visible tool demand.
• Prefer ASML over ORCL on a 6-9 month basis; the former has secular scarcity pricing and longer-duration optionality, while the latter is more exposed to sentiment-driven AI monetization disappointment.
• For more aggressive positioning, use ASML call spreads 9-15 months out to express high-conviction upside with defined risk, targeting a reacceleration trade if high-NA adoption broadens beyond early adopters.