This fund manager bought Nvidia and SK Hynix and sold software before others. His simple message on AI: ‘Follow the money.

Stephen Yiu, lead manager of the nearly $4.5B Blue Whale Growth Fund, says investors should “follow the money” in AI, citing his recent purchases of Nvidia and SK Hynix. The note is more positioning/strategy than a new fundamental catalyst, so likely limited near-term impact beyond sentiment for AI-related names.

Analysis

This is less a stock-specific note than a capital-allocation signal: AI budgets are still flowing first to infrastructure, not applications. That keeps the near-term earnings delta concentrated in NVDA and adjacent picks-and-shovels, while software faces a longer monetization lag and more scrutiny on pricing power. In the next 1-3 months, the key mechanism is revision breadth: if hyperscaler capex stays elevated, the market can keep re-rating the supply chain even without a big multiple expansion.

The second-order winner set is broader than just the GPU vendor: memory, advanced packaging, and networking can retain pricing power because bottlenecks migrate as core compute supply improves. The vulnerable group is software with AI narratives but weak usage-based revenue, where investors may start asking for proof of payback rather than product launches. Over 6-18 months, the main risk to the whole trade is capex digestion; if incremental AI revenue does not show up, spend growth can slow abruptly.

Contrarianly, the consensus may already understand that infrastructure wins the first inning, so the cleaner edge is avoiding overpaying for narrative-heavy software rather than blindly chasing NVDA at any price. The key falsifier is a step-down in hyperscaler capex guidance or a pattern of NVDA gross-margin / backlog disappointment, which would signal supply is catching demand and the ROI debate is turning. Until then, the market still rewards the part of the chain that captures dollars first, not the layer that promises future productivity gains.

Key Decisions for Investors

• Buy NVDA on pullbacks over the next 1-3 months via 3-6 month call spreads; the thesis is continued estimate revisions from AI capex, with risk defined by any slowdown in hyperscaler spending or margin compression.
• Run a pair trade: long NVDA / short an AI-software basket such as IGV for the next 1-2 quarters. This expresses the view that infrastructure monetization is more immediate than application-layer monetization; cover if enterprise AI usage metrics accelerate meaningfully.
• Add a smaller satellite long in memory/semicap beneficiaries such as MU or SMH if HBM and packaging bottlenecks remain tight. This is a second-order trade, so size modestly and use it only if channel checks confirm pricing discipline.
• Set a tactical alert on the next hyperscaler capex updates and NVDA margin commentary. If capex growth flattens or guidance comes in softer, reduce exposure or hedge immediately; that is the most direct falsifier of the follow-the-money thesis.