Structural Analysis — ChatGPT Pro runs -1,650% margin at full utilization. Claude Max runs -900%. The market sees a bubble. The structural read: the negative margin is not a leak — it is a procurement budget for the three assets that determine who wins the AI era. Full analysis on Business Engineer.
Power users running 14-hour coding sessions generate the real-world agentic workloads that train the next frontier. Benchmarks can’t provide this. Only subscriptions can.
Asset 2: Power User Lock-In
A developer who has wired their entire workflow around Claude Code won’t switch — even if GPT-5.6 benchmarks better. The harness is stickier than the model.
Asset 3: Token Deflation Call Option
Token costs fall ~10x/year. A $200/mo subscription that costs $8K to serve today costs $800 next year and $80 the year after. The price is locked. The cost is collapsing. The spread is the franchise.
The customer never leaves — and never sees a price increase
The AI subscription is the same trade, but faster. Token costs are falling faster than any commodity in history — 10x per year for two years running. A plan that runs -1,650% margin today runs +50% margin in 24 months without changing a single line item.
The Withholding Strategy
If the labs can’t nerf subscriptions (public backlash is asymmetric), they nerf the vintage of frontier capability inside them:
Free → yesterday’s frontier
Subscription → last quarter’s frontier
API → current frontier
Enterprise → customized frontier
Glasswing / Mythos → next frontier (trust-gated, never on subscription)
This is exactly what Anthropic did with Fable 5 vs Mythos 5. The subscription gets abundant capability. The frontier stays scarce — and priced by token, contract, or trust.
OpenAI vs Anthropic: Who’s Betting Harder
The data shows OpenAI running a 75% larger subsidy per power user ($14K vs $8K of equivalent usage at the same $200 price). This means OpenAI is either:
Further from the agentic frontier and needs more training signal
Prioritizing consumer workflow ownership over enterprise discipline
Either way — the labs are running a backwards auction. The more they subsidize, the worse the accounting, and the better the strategic position. Margin and harness density are anti-correlated by design.
The Bottom Line
The real question is not “Are AI subscriptions profitable?”
The real question is: “Who will own the workflows once intelligence becomes too cheap to meter?”
The labs are positioning for that moment. The negative margin is the price of admission.
Gennaro is the creator of FourWeekMBA, which reached about four million business people, comprising C-level executives, investors, analysts, product managers, and aspiring digital entrepreneurs in 2022 alone | He is also Director of Sales for a high-tech scaleup in the AI Industry | In 2012, Gennaro earned an International MBA with emphasis on Corporate Finance and Business Strategy.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
