OpenAI's Finances: Microsoft Payments, Compute Costs & the AI Bubble?
Share
The Growing Financial Interdependence of OpenAI and Microsoft
Recent leaks of internal documents have shed light on the complex financial relationship between OpenAI and Microsoft, revealing significant revenue sharing agreements and escalating compute costs. As OpenAI gears up for a potential IPO, understanding these financials is crucial for assessing the company's long-term viability and the broader implications for the AI industry. This article dives deep into the leaked data, analyzing the revenue streams, expenses, and potential red flags surrounding OpenAI's financial health.
Microsoft's Revenue Share Payments to OpenAI
According to documents obtained by tech blogger Ed Zitron, Microsoft paid OpenAI $493.8 million in revenue share payments during 2024. This figure jumped to $865.8 million in the first three quarters of 2025. This arrangement stems from Microsoft's substantial $13 billion investment in OpenAI, where the AI startup shares 20% of its revenue with the software giant. However, the picture is more nuanced. Microsoft also shares revenue with OpenAI, deducting its own expenses, including those related to Bing and Azure OpenAI royalties, from the reported revenue share.
Estimating OpenAI's Revenue
Based on the 20% revenue-share statistic, we can estimate OpenAI's revenue. In 2024, the revenue was likely at least $2.5 billion, and in the first three quarters of 2025, it was likely around $4.33 billion. Sam Altman, OpenAI’s CEO, has suggested even higher figures, predicting an annualized revenue run rate of over $20 billion and potentially reaching $100 billion by 2027. These projections, however, remain speculative.
The Rising Cost of Inference
While revenue is important, the leaked documents also highlight OpenAI's significant inference costs. Inference, the process of using a trained AI model to generate responses, is proving to be incredibly expensive. OpenAI reportedly spent roughly $3.8 billion on inference in 2024, a figure that increased to approximately $8.65 billion in the first nine months of 2025. This represents a substantial portion of OpenAI's overall expenses.
Compute Infrastructure: Azure, CoreWeave, Oracle, AWS & Google Cloud
OpenAI has historically relied heavily on Microsoft Azure for compute access. However, the company has diversified its infrastructure, establishing deals with CoreWeave, Oracle, AWS, and Google Cloud to manage these escalating costs. The shift demonstrates a strategic effort to optimize compute resources and potentially negotiate better pricing.
Training vs. Inference Costs & Cash Flow Concerns
A crucial distinction lies between training and inference costs. While OpenAI's training expenses are largely offset by credits from Microsoft as part of their investment, inference costs are primarily cash expenses. This discrepancy raises concerns about OpenAI's cash flow. The leaked data suggests that OpenAI may be spending more on inference than it is earning in revenue, a potentially unsustainable situation.
Implications for the AI Bubble
The leaked financial data has fueled concerns about the “AI bubble.” If a leading AI company like OpenAI is struggling to achieve profitability, it raises questions about the valuations of other AI startups and the long-term sustainability of the industry's rapid growth. The situation warrants careful scrutiny and a more realistic assessment of the AI landscape.
Further Reading & Resources
For more in-depth analysis and related news, explore these resources:
