OpenAI’s annualized recurring revenue is approaching $70 billion, according to reporting in late September 2026. That number is extraordinary — but it should not be confused with $70 billion of booked annual revenue, and it does not answer the most important question: how expensive will the infrastructure required to sustain that growth become?
$70 billion is a run rate, not a completed year
An annualized revenue run rate extrapolates a recent period of sales into a full-year figure. It can be useful for understanding momentum in a fast-growing company, but it can also make growth look more mature than it is. Reuters reported that OpenAI’s enterprise sales had more than doubled since July and that its annualized revenue run rate had risen more than 70% since the beginning of the third quarter.
That acceleration is evidence of enormous demand. It is not, by itself, evidence that the business has solved the economics of AI.
The hidden side of AI revenue is infrastructure
Every additional user, enterprise deployment and model improvement requires computing resources. At OpenAI’s scale, that means data centers, advanced chips, power generation, networking and long-term cloud commitments.
Oracle has become a critical computing-infrastructure partner for OpenAI. In September, Oracle reported more than $30 billion of additional AI cloud contracts in a quarter and a total revenue backlog of $664 billion. That backlog demonstrates the scale of demand, but it also shows how much capital must be deployed before those contracts can turn into revenue.
Growth can create a financing problem
AI infrastructure has unusual economics. Demand can grow faster than data centers and power plants can be built. Companies therefore need to make very large commitments years before all of the underlying revenue has been realized.
That creates a gap between commercial momentum and physical capacity. Financing bridges the gap, but financing is not free. Higher interest rates, construction delays, power shortages or weaker-than-expected AI demand can all change the economics of projects that looked compelling when they were announced.
Why Oracle matters
Oracle’s role makes it a useful window into the infrastructure side of the AI race. The company is spending aggressively to expand cloud capacity while signing long-duration agreements with major AI customers. If those customers continue scaling, Oracle gains a powerful growth engine. If utilization or financing assumptions disappoint, the same expansion can pressure cash flow.
This is why OpenAI’s revenue and Oracle’s infrastructure commitments should be analyzed together. One side shows demand. The other shows the cost of supplying it.
The real question is not whether AI can grow
The more useful question is whether AI revenue can grow fast enough — and with enough margin — to justify the scale of capital being committed across chips, data centers and electricity.
OpenAI approaching a $70 billion annualized run rate is a major signal. But the infrastructure race surrounding it may be even more consequential. The companies that can finance, build and power AI at scale are becoming part of the product itself.
