Is Billionaire Michael Platt Souring on the Magnificent Seven? He Just Closed Positions in 4 AI Titans in Favor of a Company That May Be a Surefire Winner in the AI Infrastructure Spending Boom.

Michael Platt of BlueCrest sold all of his Nvidia, Microsoft, Meta, and Amazon holdings in Q1 while increasing Taiwan Semiconductor Manufacturing by more than 430% to 35,824 shares, now 0.7% of his portfolio. The article argues TSMC may be the clearest beneficiary of the current AI infrastructure spending phase, as chip demand for Nvidia, Broadcom, AMD, and others flows through its foundry business. The piece is largely portfolio commentary rather than company-specific news, so near-term market impact is limited.

Analysis

The market is still treating AI as one trade, but this filing split exposes the second phase: capital is rotating from the “story” layer to the picks-and-shovels layer. If hyperscalers remain disciplined on capex, the chip-design names can de-rate while the foundry and supply-chain bottleneck retains pricing power, which is why TSM should outperform on earnings durability even if the broader AI basket stalls. The key second-order effect is that every incremental GPU/ASIC deployment increases utilization and negotiation leverage for the most constrained manufacturing node, not just the designer with the loudest narrative.

The consensus mistake is to assume selling the large platform names is a bearish AI signal. It is more likely a preference for cleaner exposure with less software margin compression risk and less multiple sensitivity to slowing ad/cloud growth. The move out of AMZN, MSFT, META, and NVDA also hints at a view that near-term upside is increasingly capital intensity-driven rather than demand-acceleration-driven, which tends to favor suppliers with hard bottlenecks and visible order books over end-market platforms.

TSM is not risk-free: the main failure mode is a capex air pocket if hyperscalers freeze budgets for 1-2 quarters, or if geopolitics reprice the equity risk premium abruptly. But the more likely medium-term reversal is that AI inference demand broadens faster than expected, forcing a second wave of advanced packaging and wafer starts, which would amplify TSM’s operating leverage over the next 6-12 months. In that scenario, the current rotation is early, not late, and the best entry is on any consolidation in the foundry space rather than chasing the platform names into earnings.

For the rest of the basket, AVGO and AMD are the cleanest secondary beneficiaries because they monetize the buildout through design wins and custom silicon, but they carry higher product-cycle risk than TSM. NFLX and AAPL remain largely insulated from this specific capital reallocation, making them useful as financing sources in relative-value structures if AI enthusiasm fades and factor crowding unwinds.

Key Decisions for Investors

• Overweight TSM vs. NVDA for a 3-6 month horizon; use a 1:1 pair trade to isolate the supply-chain bottleneck thesis and reduce exposure to multiple compression in AI leaders.
• Add AVGO on pullbacks as a higher-beta AI infrastructure winner; target a 6-9 month window with asymmetric upside if custom silicon adoption accelerates, but size below TSM due to cycle risk.
• Short a basket of the most crowded platform names on strength (MSFT/META/AMZN) against long TSM; the trade benefits if capex growth persists but software/cloud multiples compress as investors discriminate between beneficiaries.
• Sell downside protection on TSM only after a post-earnings volatility reset; implied volatility should stay supported, and premium capture is favorable if geopolitical headlines are stable.
• Avoid chasing NVDA here; if hyperscaler capex slows even modestly, the stock’s duration risk is higher than the foundry’s because expectations already embed near-perfect demand continuation.