Palantir Billionaire Peter Thiel Sells Nvidia and Buys an AI Stock Up 476,900% Since Its IPO

Thiel Macro sold its entire Nvidia stake in Q3 and reinvested the proceeds into Microsoft, reflecting a shift from a GPU-focused AI play to a software/cloud AI exposure. Nvidia still controls an estimated 70–90% share of AI accelerators today and >80% revenue share in AI accelerators, with Wall Street projecting ~37% annual EPS growth over the next three years and a current valuation near 44x earnings; the AI-accelerator market is forecast to grow ~29% annually through 2033. Microsoft, the largest enterprise software vendor and No.2 public cloud, is rolling out Microsoft 365 Copilot (90% of Fortune 500 adoption cited), faces 28% cloud sales growth amid capacity constraints, plans to double data-center footprint in two years, and has a consensus ~14% annual EPS growth with a current multiple of ~34x and PEG ~2.4 (below its 3- and 5-year averages).

Analysis

Market structure: Nvidia remains the incumbent winner in AI accelerators with an implied 70–90% share and a consensus earnings CAGR ~37% next 3 years, which sustains pricing power for GPUs vs. bespoke hyperscaler chips because of CUDA lock‑in and developer onboarding costs. Microsoft is the primary beneficiary among software/cloud players — its Microsoft 365 Copilot monetization and capacity doubling over 24 months point to share gains in enterprise AI and higher cloud ASPs, justifying a premium versus traditional software peers.

Risk assessment: Key tail risks are rapid hyperscaler migration to custom silicon (reducing Nvidia revenue by >20% within 2–4 years), antitrust action on dominant ecosystems, and supply-chain shocks (power/wafer constraints) that could compress margins. Short term (days–months) moves will be driven by MLPerf releases, hyperscaler capex announcements, and quarterly cloud guidance; long term (years) outcomes hinge on software ecosystem entrenchment and model runtime economics.

Trade implications: Tactical approach favors concentrated, size‑limited exposure: own platform winners (MSFT) via equity/leaps and play Nvidia through option spreads to cap downside while keeping upside; overweight semiconductor equipment/capacity beneficiaries if capex accelerates. Cross-asset: rising tech capex and rate sensitivity argue hedging nominal duration (buy 3–6 month S&P puts or VIX calls) and monitoring USD strength which can compress reported multinational revenue.

Contrarian angles: Consensus underweights the durability of software lock‑ins — CUDA’s ecosystem and pretrained models create a multi‑year switching cost that could preserve Nvidia’s economics even if raw perf gaps narrow. Conversely, markets may underprice a coordinated hyperscaler pivot: if two of {GOOGL, AMZN, META} publicly commit to TPU/Arm fleets within 12–18 months, reprice semiconductor winners/losers quickly.

Key Decisions for Investors

• Establish a 2–3% portfolio long position in MSFT within 30 days (buy shares or 12‑month LEAPS) sizing to target a 12‑month upside of 15–25% if cloud AI revenue growth reaccelerates above 25% YoY; consider selling 5% OTM 3‑month cash‑secured puts to lower basis if implied vol < historical 90‑day.
• Initiate a tactical 1–2% exposure to NVDA via 6–9 month call debit spreads (buy 10–20% OTM, sell 30–40% OTM) to capture further AI demand while capping capital at risk; scale into a 3–5% outright long equity position only if forward P/E falls to ≤35x or consensus 3‑yr EPS CAGR remains ≥30%.
• Enter a relative value pair: long MSFT (2%) / short GOOGL (1.5%) over 6–12 months to capture differential enterprise monetization of AI; trim if Google announces >15% YoY improvement in MLPerf TPU performance or if Alphabet cloud share rises >300bps sequentially.
• Hedge tail risk: allocate 0.5–1% notional to 3–6 month S&P 500 put spread or VIX 30‑60 day call package to protect against a tech drawdown triggered by hyperscaler capex pivot or regulatory action announced within the next 90 days.