OpenAI’s financial plan has become much larger. According to The Information, the company now projects a total cash outflow of $115B through 2029, about $80B more than earlier forecasts that had pointed to break-even by then.
The revised forecast shows the scale of the AI buildout behind OpenAI’s products. Costs are rising across model training, inference, infrastructure and compensation, while the company is also counting on a major increase in revenue, especially from ChatGPT.
A much bigger spending curve
The reported spending path is steep. OpenAI expects to go through more than $8B this year, with spending rising to $17B in 2026. By 2027, costs are expected to reach $35B, followed by $47B in 2028.
That 2028 figure is especially notable because, at the start of the year, OpenAI was projecting only $11B in spending for 2028. The new forecast therefore does not just extend an existing trend. It changes the scale of the company’s expected cash needs.
The core driver is computing power. OpenAI needs large amounts of compute to train and run its models, and the company is also pushing to build its own server infrastructure. The plan includes nearly $100B in data centers and custom chips by 2030.
The logic behind that push is straightforward. If OpenAI can reduce its dependence on outside cloud providers, it may be able to lower costs over time. But reaching that point requires heavy upfront investment.
Training and inference are doing the damage
Training foundation models remains one of the largest cost centers. OpenAI expects to spend over $9B on training alone in 2025, about $2B more than it had budgeted. By 2026, that figure could reach $19B.
Training costs are only part of the issue. Once models are deployed, users generate ongoing inference demand. The report says inference costs could top $150B by 2030.
That matters because OpenAI’s products depend on both sides of the equation. Better models require expensive training runs, and popular products require the company to keep serving model outputs at scale. Growth can therefore increase revenue while also increasing the cost of running the service.
OpenAI is also budgeting for talent. The company is setting aside about $20B in extra stock compensation through 2030 as competition for AI researchers and engineers intensifies. The report notes that Meta is reportedly offering nine-figure deals to lure away leading experts.
Revenue expectations are rising too
OpenAI is not only forecasting higher costs. It is also projecting higher revenue. For this year, the company expects total revenue of about $13B, roughly $300M above its earlier 2025 forecast.
By 2030, OpenAI now projects $200B in revenue, about 15 percent higher than previous estimates. That would be a dramatic increase from the current revenue base described in the report.
ChatGPT is central to that plan. The chatbot is expected to bring in almost $10B this year alone, around $2B more than projected at the start of the year. By 2030, OpenAI expects ChatGPT to generate nearly $90B in revenue, a 40 percent increase over earlier projections.
Over the next six years, the company is banking on an extra $70B in revenue from ChatGPT alone. That makes ChatGPT more than a consumer-facing product. In the reported forecast, it is one of the main engines expected to help offset OpenAI’s rising cash needs.
Free users become part of the business model
The report also points to a major revenue role for free ChatGPT users. Between 2026 and 2030, OpenAI aims to bring in about $110B from this group through things like shopping commissions or possibly advertising.
Earlier projections for free ChatGPT users put average revenue per user at $2 in 2026 and $15 in 2030. That is described as a 650 percent jump, with a target user base of two billion weekly active users.
OpenAI currently counts 700 million weekly active users. Reaching the larger target would require the free user base to become far larger, while also generating more revenue per user.
The company expects gross margins in this segment to reach between 80 and 85 percent, putting it in the same league as Meta’s advertising business. That margin goal depends on system costs coming down, or on ad deals with ChatGPT becoming expensive for advertisers.
Investors are still watching closely
The spending forecast carries obvious risk, but investor interest remains. The report says several investors are prepared to buy shares at a valuation of $500B.
That compares with an OpenAI valuation of $300B earlier in August. Rival Anthropic is currently valued at about $180B.
The numbers show how much capital the leading AI companies may need as models, products and infrastructure scale together. OpenAI’s revised forecast suggests that the race is not only about model capability or user adoption. It is also about whether revenue can grow fast enough to keep pace with the cost of building and operating the systems behind it.