Peter Thiel dumped Nvidia and invested $45 million into Microsoft and Apple—sending a strong signal about who will win the AI race

Thiel Macro sold all 537,742 Nvidia shares in Q3 2025, cutting U.S. equity exposure from $212M to $74M (a >50% reduction) and reallocating roughly $45M into Apple and Microsoft. The trades reflect a cautious view on an "extremely bubbly" AI market and concerns about Nvidia's circular investment-deal arrangements with startups and cloud providers. Apple and Microsoft are positioned as more diversified, durable paths to commercialize AI via consumer platforms and services. The repositioning could dent Nvidia sentiment while modestly supporting platform/consumer tech names.

Analysis

The trade rotation away from concentrated infrastructure exposure toward platform/distribution signals a re-price of risk from a hardware-driven, volume/capex story to a monetization-and-distribution story. Expect the hardware cohort to face margin compression if enterprise/cloud capex growth slows; my base case is a 6–12 month window in which spot pricing for used datacenter accelerators falls 20–40% and new-ASP growth stalls, pressuring revenue recognition for pure-play suppliers.

Second-order supply-chain impacts will surface quickly: memory (HBM) and substrate vendors see lumpy, high-variance order books that can amplify earnings volatility by +/-30% quarter-to-quarter, and the expanding secondary GPU market will blunt OEMs’ replacement cycles. Startups that used vendor financing or off-take-backed capex face covenant and liquidity squeezes within 3–9 months, which will cascade into renegotiated commitments and potential markdowns for chip vendors that relied on those deals to smooth demand.

Key catalysts that can reverse the move are asymmetric. Near-term: cloud providers’ next-quarter capex guidance and major earnings calls (days–weeks) can reflate or puncture hardware demand expectations. Medium-term (3–12 months): a demonstrable jump in model scale or a new training paradigm that requires ~5–10x more FLOPs would reflate GPU demand; conversely, regulatory scrutiny of circular financing or accelerated in-house accelerator development at hyperscalers would structurally compress margins for incumbents.

From a portfolio perspective, lean toward platform owners with distribution moats and recurring monetization (software + services) while keeping convex, defined-risk hedges against a hardware melt-up or crash. Size trades to 0.5–2% NAV lines with explicit stop-losses and prefer option structures or spreads that cap downside while leaving upside asymmetry if the market re-rates relative franchising power over the next 6–18 months.

Key Decisions for Investors

• Relative-value pair (size 1.5% NAV): Long MSFT 12-month call spread (buy 10% ITM, sell 30% OTM) funded by selling a 12-month NVDA call spread (sell 10% ITM, buy 30% OTM). Rationale: capture platform monetization > hardware re-rating; target 25–50% return if MSFT outperforms NVDA by ~15% in 12 months. Max loss = net premium (~1.5% NAV); unwind if NVDA vs MSFT dispersion moves >25% intraday.
• Covered-income AAPL leg (size 1–2% NAV): Buy AAPL equity and sell 3-month covered calls (at-the-money) to lower basis and generate 4–8% annualized carry. Target total return 8–15% over 6–12 months; place stop-loss at -10% absolute on the equity leg and roll calls if implied vol spikes >40%.
• Asymmetric NVDA downside hedge (size 0.5% NAV): Buy 9–12 month NVDA puts 20–25% OTM financed by selling near-term (3-month) puts 10% OTM to lower cost. This structure pays off ~3–5x if NVDA drops >30% while limiting carry cost; cap allocation to protect portfolio rather than speculate.
• Opportunistic short-duration trade (size 0.5–1% NAV): Sell NVDA-dedicated capex/resource weakness by shorting listed secondary beneficiaries (or buying CDS on smaller suppliers) around next-quarter capex-guide misses; target event window 0–90 days, aim for 10–30% short-term P&L if guidance disappoints. Strict cut if cloud providers signal firm multi-year commitments.