Billionaire Philippe Laffont Has Sold Shares of Nvidia in 10 of the Last 11 Quarters -- What Does He Know That Wall Street Doesn't?

Coatue founder Philippe Laffont sold more than 40.6M split-adjusted Nvidia shares, reducing his position from 49,802,020 on 2023-03-31 to 9,203,338 on 2025-12-31 (≈81.5% reduction) after selling in 10 of 11 quarters. The article frames the selling as partly benign profit-taking after NVDA’s >1,200% rise since Q3 2022 but highlights risks that could pressure pricing and margins (customer in-house GPUs, slower AI optimization, and China/trade uncertainty); monitor large-holder flows, Nvidia gross-margin trends, customer chip programs, and China exposure as potential drivers of low-single-digit, stock-specific moves.

Analysis

Nvidia’s leadership in AI compute creates an arms-race dynamic where hardware, software (model stacks/CUDA), and customer lock-in interact — that interlock is the core source of pricing power but also its Achilles’ heel. Over the next 12–36 months the key margin risk is not immediate competitive parity from an external vendor but gradual wallet-share loss as hyperscalers (who control demand volume) substitute cheaper, in-house accelerators for non-core workloads and negotiate deeper software/hardware co-design discounts.

A second-order supply-chain effect: sustained high Nvidia ASPs accelerate capex and fab demand at TSMC and memory vendors, which in turn tightens access and motivates cloud providers to secure alternatives (in-house chips, multi-vendor stacks) faster than previously modeled. That pathway creates a feedback loop where strong revenues today seed future supply diversification that meaningfully reduces durability of Nvidia’s TAM penetration over a 2–5 year horizon.

The near-term catalysts that could re-rate risk/reward are discrete: (1) a product-price or architecture surprise that restores pricing leverage within 0–6 months; (2) public disclosures from top customers about meaningful ASIC migrations over 6–24 months; (3) trade-policy shocks or China restrictions that re-route demand on a 0–12 month timeline. Watch implied volatility, hyperscaler capex cadence, and public customer roadmaps; any of these flip the trade from convex upside to structural deceleration.

Key Decisions for Investors

• Directional, convex long: Buy NVDA Jan-19-2029 LEAPS call spread (long 600C / short 1200C). Size 1–2% portfolio. Rationale: preserves long-term exposure to secular AI while capping cost; target 2x return if NVDA sustains dominance; max loss = net premium. Entry: on pullback ≥10% from current price or immediately funded by selling near-term calls.
• Pair trade (structural hedge): Long NVDA cash 0.75–1.5% portfolio / Short INTC equal notional delta for 6–12 months. Rationale: expresses premium for Nvidia’s AI moat while shorting legacy CPU exposure and valuation dispersion. Risk/reward: asymmetric upside if Nvidia retains share; downside partially financed by proceeds from the Intel short.
• Vol/arbitrage: Sell near-dated NVDA calls (30–45 days) when implied vol > realized vol by 8–10 vol points; buy protective OTM puts (45–60 days) to cap tail risk. Size small (0.25–0.5% portfolio). Rationale: monetize elevated short-term optimism while limiting blowup from product/China shocks.
• Event trigger: If a major hyperscaler publicly announces migration >15% of AI workloads to in‑house silicon or if trade curbs expand, increase hedges (buy NVDA puts 3–6 month tenor) to cover a 20–40% downside over 3–12 months.