First-quarter figures indicated OpenAI was losing roughly $1.22 for every dollar of revenue, with approximately $25 billion in annualised revenue against about $30 billion in annual spending.
The ratio is the relevant data point, more than the absolute values. Losing more than revenue isn't unusual for a company expanding rapidly, but it usually happens when the main cost is customer acquisition, which can be reduced by decision.
Here the main cost is serving the product. Every response consumes paid compute, which means growing users increases losses until unit cost falls enough, and unit cost depends on power, memory and fabrication capacity.
That's why the sector concentrated efficiency announcements in the period, with models consuming fewer tokens for the same task and cost-based routing between models. It isn't technical refinement: it's the only lever available in the short term.
For anyone depending on these services, the practical reading is about future pricing. A company needing to close that gap tends to raise prices, reduce free tier generosity, or push customers toward longer contracts.