OpenAI's 2025 Financials Reveal $13 Billion Revenue Against $21 Billion Operating Loss

Newly disclosed audited financial statements show OpenAI generating $13.07 billion in revenue during 2025, more than triple its 2024 figure of $3.7 billion, yet the company simultaneously recorded an operating loss of $20.92 billion. The documents, obtained by independent journalist Ed Zitron and reviewed by multiple outlets as the company prepares for a potential initial public offering, illustrate the fundamental tension underlying the current AI boom: rapid revenue scaling coupled with even steeper expense growth.

OpenAI's research and development spending nearly doubled year-over-year, rising from $7.81 billion in 2024 to $19.18 billion in 2025—a figure that alone exceeded the company's total annual revenue in both years. Of this R&D total, $10.59 billion was paid to Microsoft for infrastructure support, underscoring the capital-intensive nature of training and maintaining advanced AI models. Additionally, the cost of revenue—expenses associated with generating user interactions—nearly tripled from $2.65 billion to $7.5 billion, reflecting mounting inference costs as millions of users generate prompts. Sales and marketing expenses also surged from $1.11 billion to $5.73 billion as OpenAI competed aggressively to expand its user base.

While the company's net loss appears to reach nearly $39 billion, a substantial portion stems from one-time accounting adjustments tied to OpenAI's transition to for-profit structure and changes in investor valuations. Excluding this adjustment, the underlying net loss approximates $8 billion, which better reflects operational performance. Monthly revenue reached approximately $2 billion by year-end 2025, and OpenAI reported ChatGPT now has over 900 million weekly users, though only around 50 million subscribe to paid plans.

Despite the significant losses, investor confidence remains intact. OpenAI raised $122 billion in March 2026 at an $852 billion valuation. The company has publicly stated it aims to achieve profitability by 2030, a target dependent on reducing training and inference costs. However, challenges are emerging: some enterprise customers are resisting token-based pricing models and demanding clearer returns on investment, while intensifying competition from rivals like Anthropic is creating pricing pressure. The financial disclosures indicate the company has already begun consolidating efforts, including shutting down its Sora video model and refocusing on core products serving developers and business users.