After more than a year characterized by relentless dealmaking, frenzied speculation, and persistent rumors about an impending initial public offering, the financial spotlight on OpenAI has grown considerably more intense. Newly disclosed materials, reportedly leaked to technology commentator Ed Zitron, offer a more detailed and nuanced view of the company’s inner finances—revealing intricate data about its revenue streams and soaring computational expenses accumulated over the past several fiscal years.
According to Zitron’s report, which surfaced earlier this week, Microsoft received an extraordinary $493.8 million in revenue-share payments from OpenAI during 2024. That figure, based on the internal documents he claims to have examined, climbed sharply the following year, reaching approximately $865.8 million in the first three quarters of 2025 alone. The arrangement between the two companies, as described, reflects a mutually beneficial financial ecosystem arising from their deep strategic collaboration.
OpenAI is said to remit roughly 20% of its total revenue to Microsoft. This stems from a prior agreement forged when Microsoft invested an astonishing $13 billion into OpenAI, solidifying its position as a central partner and infrastructure provider for the AI firm’s exponential growth. While neither company has publicly confirmed the exact percentage associated with this revenue-sharing framework, it has long been the figure cited by individuals close to the partnership, shaping the market’s perception of the financial intricacies underlying their alliance.
However, the situation is complicated by the fact that financial reciprocity reportedly flows in both directions. According to a source familiar with these matters who spoke with TechCrunch, Microsoft also forwards a portion of its own revenue back to OpenAI—approximately 20% of the proceeds derived from its Bing search engine operations and the Azure OpenAI Service. These two Microsoft divisions rely heavily on OpenAI’s models: Bing incorporates OpenAI’s advanced language technology to enhance user search experiences, while the Azure OpenAI Service allows companies and developers to integrate OpenAI’s models into enterprise applications through Microsoft’s cloud infrastructure. This interdependent financial relationship, though symbiotic, introduces considerable complexity when calculating how much money actually changes hands.
The same source explained that the leaked figures actually reference Microsoft’s net revenue share rather than its gross totals. This means the disclosed payments exclude any royalty amounts Microsoft has already paid to OpenAI for usage through Bing or Azure OpenAI. In essence, Microsoft first deducts those royalty payments internally before reporting its resulting net revenue share—a distinction that significantly alters external estimations of the arrangement’s scale. Because Microsoft does not publicly break down revenue from Bing or Azure OpenAI within its official earnings statements, analyzing or verifying the full economic scope of these transactions becomes nearly impossible for outside observers.
Despite this opacity, the leaked documentation still offers a rare, high-resolution view into one of the most closely watched private companies in the global technology market. Beyond providing insight into OpenAI’s revenue, it reveals facets of how much the firm is expending relative to its income—a crucial metric that has fueled speculation about the sustainability of its massive operations. For instance, using the commonly cited 20% revenue-share ratio, analysts can infer that OpenAI’s total revenue reached at least around $2.5 billion in 2024, escalating to approximately $4.33 billion during the first three quarters of 2025. Such estimates, though derived indirectly, suggest that OpenAI’s financial throughput may actually surpass those baseline figures.
Other independent reports, such as those published by The Information, have painted a similar but slightly more expansive portrait. Their data placed OpenAI’s 2024 revenue closer to $4 billion and estimated the first half of 2025 at roughly $4.3 billion. Chief executive Sam Altman has additionally amplified those expectations, stating publicly that OpenAI’s actual revenues are “well more” than previously reported amounts. He noted that the company is positioned to end the current year with an annualized run rate exceeding $20 billion, an extrapolated figure that signals extraordinary growth—though it constitutes a projection rather than formal financial guidance. Altman added that if such momentum continues, OpenAI could conceivably approach an astounding $100 billion in annual revenue by 2027.
Zitron’s analysis also illuminated OpenAI’s steep and accelerating operational costs, especially those tied to inference—the computational process required for deploying trained AI models to produce real-time outputs and responses. According to the leaked numbers, OpenAI may have spent about $3.8 billion on inference during 2024 alone, with that figure soaring to nearly $8.65 billion in just the first nine months of 2025. These escalating costs underscore the staggering financial demand of continuously running AI infrastructure at scale.
Historically, OpenAI has been almost entirely dependent on Microsoft’s Azure platform to obtain the immense computational capacity necessary for both training and inference workloads. However, over time, OpenAI has diversified its cloud partnerships by striking additional agreements with CoreWeave and Oracle, and more recently extending those relationships to Amazon Web Services and Google Cloud. This diversification highlights both the company’s growing bargaining power and the industry’s recognition that OpenAI’s infrastructure needs are colossal.
Earlier investigative accounts estimated the company’s entire compute expenditure at roughly $5.6 billion for 2024, with its cost of revenue in the first half of 2025 hovering around $2.5 billion. A source familiar with internal operations further clarified to TechCrunch that OpenAI’s training costs are largely non-cash—mostly offset through compute credits extended by Microsoft as part of its multibillion-dollar investment package. In contrast, the vast majority of inference expenditures are actual cash outflows, representing true operational spending. Training refers primarily to the massive cloud workloads that teach AI models to understand and generate text, while inference refers to using those models in production—a distinction with enormous financial implications.
Collectively, these figures present a somewhat sobering financial picture. The data imply that OpenAI, while generating exceptional revenue, might still be spending an equal or even greater sum to sustain ongoing inference operations. This imbalance, though common in capital-intensive technology fields, has sparked significant commentary within both financial and technological circles—fueling discussions about whether the current wave of AI investment reflects sustainable innovation or an overheated speculative bubble. If the sector’s most prominent figurehead remains in deficit, what does that signal for the multitude of startups and multinationals racing to capture value in the same domain, often at equally dizzying valuations?
Both OpenAI and Microsoft have declined to comment publicly on the specifics of these disclosures. TechCrunch noted that its inquiries to representatives from each company were either unanswered or met with a ‘no comment’ response. The silence underscores how sensitive and strategically crucial such financial information remains in a landscape dominated by competition, hype, and massive infrastructural costs.
For readers possessing confidential information, TechCrunch encourages secure communications regarding the ongoing investigation into the economics and internal dynamics of the AI industry—an industry where decisions made in corporate boardrooms reverberate across global markets and into people’s everyday technologies. Thus, the unfolding story of OpenAI’s balance between innovation, cost, and profitability continues to serve as both a cautionary tale and a testament to the breathtaking pace at which artificial intelligence reshapes the modern technological economy.
Sourse: https://techcrunch.com/2025/11/14/leaked-documents-shed-light-into-how-much-openai-pays-microsoft/
