OpenAI generated $4.3 billion in revenue in the first six months of 2025, surpassing its entire earnings for 2024 by 16%.
But the growth comes with a price, where massive spending on research and development is driving heavy losses.
According to financial disclosures seen by The Information, the company spent $6.7 billion on R&D in the first half of the year, most of it tied to training models and maintaining the infrastructure behind ChatGPT. That pushed its cash burn to $2.5 billion, with its ability to strike a balance between rapid expansion and steep operating costs in question.
Despite the high outflow, OpenAI appears to be on solid footing for now. It closed June with around $17.5 billion in cash and securities, enough to sustain its aggressive investment strategy while targeting $13 billion in revenue and $8.5 billion in full-year cash burn by December.
The company is also manoeuvring to ensure value for insiders. In August, OpenAI entered early-stage talks for a stock sale that would allow employees to cash out, a deal that could peg its valuation at about $500 billion. That figure would place the firm among the most valuable technology companies worldwide, highlighting investor belief in its dominance in generative AI.
Growth around OpenAI has been strengthened by its deepening ties with chipmaker Nvidia. Earlier this month, Nvidia announced it would commit as much as $100 billion to support the firm, supplying high-performance data-centre chips to bolster training and inference at scale. This shows the growing demand for AI hardware but also the interdependence between model developers and infrastructure providers.
For competitors such as Anthropic and Google DeepMind, the challenge is increasing. OpenAI’s integration into Microsoft Copilot is spreading across industries from healthcare to finance, giving it a commercial reach that few competitors can match.
Its spending levels are also setting new benchmarks for research intensity, with consequences for talent competition, regulation, and industry standards.