How Would You Value an OpenAI IPO? A Reverse DCF Thought Experiment

Charlie MungerInvert, always invert. Figure out what causes failure, then avoid it.

• Compute costs must fall faster than price competition compresses revenue per token. If inference prices race to zero faster than costs do, scale becomes a treadmill, not a moat.
• The consumer subscription must survive commoditization. Hundreds of millions of users today; the question is what happens when comparable assistants are free, bundled, or built into the operating system.
• Governance and capital structure must not surprise. OpenAI's unusual history of nonprofit-to-commercial restructuring is exactly the kind of complexity Munger filed under "too hard" — not because it's bad, but because it resists analysis.
• Key-person and talent risk must stay contained. The lab's output is its researchers; researcher mobility in AI is the highest of any industry in memory.

Warren BuffettThe best business is one that earns more cash each year without requiring more capital to do it.

• OpenAI's reported compute commitments run into the hundreds of billions of dollars across multi-year deals. In Buffett's language: this is a business that currently must spend enormously to stay in place.
• Compare his favorite economics — See's Candies raising prices yearly with almost no new capital — to a frontier lab that must rebuild its core asset (the model) every 12–18 months at rising cost.
• The bull case is that this is a railroad in 1850: brutal capex now, tollbooth later. Buffett owns railroads — bought a century after the capex frenzy, at a single-digit multiple. Timing of entry is the entire difference.

John TempletonBuy at the point of maximum pessimism; sell at the point of maximum optimism.

• An OpenAI IPO would arrive at — almost by construction — the point of maximum optimism: every fund must own it, every index will add it, every headline will anchor on day-one pop percentage.
• Templeton's discipline is not "never buy great companies." It is: the price you pay is set by the crowd's mood, and the mood at a mega-IPO is as good as it ever gets for the seller — which makes it as bad as it ever gets for the buyer.
• His practical move: put it on the watchlist with a pre-written price, and revisit when the lockup expires, the first guidance miss lands, or the narrative wobbles. Optimism is temporary; filings are quarterly.