OpenAI just raised $122 billion at an $852 billion valuation while planning to lose $111 billion over five years.

The Summary

  • OpenAI closed $122 billion in committed capital at $852B valuation, putting it alongside Walmart and Samsung by market cap
  • The company projects $111B in losses through 2029, averaging more annual losses than most comparable companies earn in profit
  • This is a bet that Web4 infrastructure will eventually print money, but nobody knows when or if that inflection point arrives

The Signal

The numbers are wild enough to warrant a pause. OpenAI's own projection shows $111 billion in losses between 2024 and 2029. That's not a typo. For context, Walmart made $22 billion in profit last year. Samsung made $29 billion. Eli Lilly made $21 billion. OpenAI will lose, on average per year, what these century-old companies earn.

The bull case is straightforward: we're watching the buildout of Web4 infrastructure. The companies pouring money into OpenAI believe agents will replace entire categories of human work, and whoever owns the foundation models owns the rails. If you're building the railroad, you don't need to be profitable while laying track. You need capital and conviction that people will ride the train.

The bear case is equally clear: there's no proof the train will come. We're two years into the ChatGPT era and the revenue model is still "charge people $20/month for a chatbot." Enterprise is adopting, sure, but at what margin? The gap between "this is useful" and "this prints money at scale" remains enormous. Announced plans for a "superapp" suggest OpenAI is searching for that business model, not executing one they've already found.

What makes this stranger is the timing. We're past the easy amazement phase. Everyone's used ChatGPT. The tech works. But working and being worth $852 billion are different problems. This raise suggests the money believes Web4 happens fast and winner-take-most. It also suggests OpenAI needs this capital to survive long enough to find out if that's true.

The Implication

Watch what OpenAI builds with this money. If it's all compute and model training, that's one bet. If it's product diversification and the promised superapp, that's a different bet, one that admits foundation models alone won't get them profitable. For anyone building in the agent space, this sets the table: the infrastructure you're building on is funded for five more years of losses. Plan accordingly.

Source: Daring Fireball