TIP800: Navigating an AI-Driven Market w/ François Rochon

What It Covers

François Rochon of Giverny Capital, who has compounded capital at 13.4% annually since 1993, discusses navigating AI disruption across his portfolio, explains why Constellation Software trades cheaply at 18x earnings after a 50% drawdown, analyzes Alphabet and Meta's $280B combined CapEx as defensive spending, and outlines why rationality, humility, and patience define long-term investing success.

Key Questions Answered

• •AI Circular Revenue Risk: Nvidia's self-reinforcing revenue loop — investing in OpenAI, which contracts Oracle, which buys Nvidia chips — creates non-recurring profits that markets incorrectly capitalize at 25x PE over 20+ years. Investors should apply a lower multiple to earnings generated through this circular financing structure, treating them closer to a 3-year contract value than a perpetual income stream.
• •Defensive CapEx Framework: Alphabet's projected $180B and Meta's $100B+ in 2026 CapEx should be evaluated as defensive moat-spending rather than growth investment. Alphabet responded to ChatGPT threatening its search business; Meta rebuilt ad-targeting after Apple's 2022 privacy changes cost ~10% of revenue. Both fund this spending from operating cash flow, eliminating bankruptcy risk that plagued debt-financed railroad and fiber buildouts.
• •Software Moat Durability Test: When evaluating AI disruption risk to software companies, assess whether the software handles money transfers, relies on proprietary client data inaccessible to LLMs, or costs so little (e.g., $5,000–$10,000 annually for a $1B-revenue business) that replacement economics don't justify disruption. Constellation Software's ~1,000 niche vertical market software businesses largely pass this test, supporting Rochon's view that 18x earnings is cheap for a 20% grower.
• •Owner's Earnings Tracking Method: To maintain conviction during drawdowns, track the aggregate earnings growth of all portfolio holdings annually as if they were private businesses. Rochon found a strong 30-year correlation between his portfolio's owner's earnings growth and actual portfolio returns. When Constellation's stock fell 50%, earnings continued growing, signaling the business remained intact regardless of market pricing.
• •Sell Discipline Triggers: Two concrete sell signals emerged from Rochon's 2025 mistakes. For CarMax (held 18 years), permanent margin compression from Carvana and AutoNation's used-car expansion indicated moat erosion, not cyclicality. For Fiserv, net debt exceeding 5x net income after CEO departure created unacceptable downside risk — Rochon's hard threshold — even though the stock traded at just 7–8x earnings at the time of sale.
• •S&P 500 Return Expectations: The S&P's 14.8% annual return from 2015–2025 combined ~9% earnings growth (driven by mega-cap concentration) with PE expansion from ~18x to ~25x historical median. Since long-run S&P earnings growth averages 6–7% annually and mega-caps face size constraints, investors should lower forward return expectations for the next decade, while selectively targeting quality companies now trading 30–50% below recent highs at sub-20x earnings.