The conversation around artificial intelligence has moved from world-domination theories to balance sheets. The shift is ugly. OpenAI’s recent financial disclosures suggest the leading generative AI model is not just expensive; it might be fundamentally broken as a business.
Internal forecasts reviewed by The Information paint a grim picture. OpenAI expects to burn $5 billion in 2024 on $3.7 billion in revenue. That deficit triples by 2026, with a projected loss of $14 billion. By 2028, total losses could reach $44 billion. Profitability is not on the roadmap until 2029, and getting there requires raising another $100 billion just to keep the lights on.
This kills the standard Silicon Valley software narrative. Usually, software benefits from high margins: you code it once, sell it a million times, and costs stay flat. Generative AI works differently. It functions more like manufacturing or heavy industry. Every user query carries a significant compute cost. As the product gets more popular, the company bleeds more cash.
Despite the burn rate, OpenAI recently closed a funding round valuing the company at $157 billion. This number is a bet on magic. Specifically, Artificial General Intelligence will arrive soon and fix the economics. Asset managers identify this as a speculative bubble. If the technology plateaus, and engineers are already seeing diminishing returns, the financial logic collapses.
The company also lacks a defensive moat. Meta is releasing open-source models like Llama that match OpenAI’s performance for free. When a competitor gives away your core product, you lose pricing power. OpenAI is trying to sell a commodity while competitors flood the market. It is nearly impossible to recoup $44 billion when the alternative is free.
The clearest signal of trouble is the pivot to advertising. AdWeek reports OpenAI is exploring ad products, a move widely seen as a last resort. This puts them in a war with Google and Meta, who own that space. It is also an admission that subscriptions alone cannot pay the bills. The utility of the “intelligence” is not valuable enough to users to cover the cost of creating it.
We are watching a market correction in real time. The tech is impressive, but the business is immature. Venture capital currently subsidizes every ChatGPT prompt, shielding users from the true cost. When that money runs out, the ecosystem will contract. The intelligence was never free. We just bought it on credit.
