marco garattoni @marcogarattoni: "Karp vs. the AI Business..."

Italy

Karp vs. the AI Business Model Alex Karp (Palantir CEO) took a shot at OpenAI and Anthropic in a recent CNBC interview, boiled down to one blunt question: "If I can make you a billion dollars, why am I selling you tokens?" Karp's point: the big AI labs sell consumption (tokens, API calls, subscriptions, compute) instead of participating in the value they claim to create. If a consulting firm or PE fund generates value for a client, they take a cut of the results. AI doesn't — the more you use, the more you pay, regardless of the output generated. Why this question matters: It challenges the dominant narrative ("AI is the new electricity") underpinning the eye-watering valuations of frontier labs It raises the data-ownership issue: companies are pouring documents, processes, and know-how into third-party models It comes from an industry insider, not an outside critic The counter-argument (and why it's not that simple): Infrastructure capitalism never sells the final outcome — it sells access. Railroads sold rails, not the industrial revolution. Electricity sells kWh, not a cut of the factory's profits it powers. Demanding a percentage of outcomes would mean turning every enterprise contract into a negotiation, with audits on "how much value did that prompt actually generate" — a nightmare of transactional friction. Also, the value of an AI output isn't predictable ex ante: the same prompt can be worth zero or a billion dollars. The only measurable variable upfront is compute cost. That's why compute gets sold, not outcomes. My take: This isn't (yet) a sustainability problem for the AI business — there's too much liquidity and enterprise demand for the consumption model to collapse short-term. But it's a signal of growing friction between enterprise/government buyers and AI vendors over pricing and data governance. Worth watching whether players like Palantir (which sells "outcomes" via Foundry, not tokens) gain relative ground by leaning into this positioning against the AI hyperscalers. Keep an eye on: "AI as indispensable infrastructure" vs. "AI as metered service" — likely the next competitive fault line in the sector, not just a philosophical debate. Marco @marcogarattoni @RelentlessAlpha $MSFT (Microsoft) $GOOG (Alphabet) $NVDA (NVIDIA Corporation) $TSLA (Tesla Motors, Inc.) $NSDQ100 $PLTR (Palantir Technologies Inc.)

Not investment advice. The author may have financial interests in the mentioned instruments.