Burning money to fuel growth! Leaked OpenAI documents reveal an “operating loss of $20.9 billion,” suggesting it won’t be profitable until 2030.

As AI giant OpenAI prepares for its first initial public offering (IPO) and plans to submit documents to the U.S. Securities and Exchange Commission (SEC), a leaked audited financial document has finally revealed its astonishing “burning money” pace. The data shows that although OpenAI’s 2025 revenue has surged more than 3 times to exceed $13 billion, the massive research and development (R&D) and compute spending still leaves the company with operating losses of more than $20 billion.
(Background: OpenAI strengthens enterprise customers! Spends $150 million to bring in consulting giants, aiming to train 300,000 certified AI consultants by the end of the year)
(Additional background: Musk suffers another defeat! A U.S. judge dismisses xAI’s lawsuit against OpenAI over trade secrets)

Table of Contents

Toggle

• Revenue surges 3 times, but R&D expenses are “more than what they earn”
• Operating losses exceed $20 billion; target profitability only in 2030
• Facing competitive pressure from Anthropic; Sora shut down to stop the bleeding

As the AI overlord OpenAI actively prepares for its highly anticipated initial public offering (IPO) and gets ready to file with the U.S. Securities and Exchange Commission (SEC), its mysterious financial “black box” has finally been opened. According to the latest report from foreign media Ars Technica on June 16, 2026, independent journalist Ed Zitron obtained a recent audited financial document from OpenAI, revealing that this AI startup—valued at the highest level globally—is in an extreme “burning money to buy growth” phase.

Revenue surges 3 times, but R&D expenses are “more than what they earn”

The leaked document shows that OpenAI has indeed made remarkable progress in commercialization. The company’s total revenue jumped from $3.7 billion in 2024 to more than 3 times that amount, reaching $13.07 billion in 2025; by the end of 2025, its monthly revenue was already nearing the $2 billion mark. In addition, ChatGPT’s weekly active users surpassed 900 million, including about 50 million paying users.

However, the impressive revenue is completely unable to fill the bottomless pit of compute power and model training. The report states that OpenAI’s 2025 “research and development (R&D) expenditure” reached $19.18 billion ($7.81 billion in 2024). This single line item alone far exceeded total annual revenue; within it, payments to Microsoft for R&D and cloud compute costs accounted for a staggering $10.59 billion. Moreover, revenue costs—including inference computation—also surged to $7.5 billion, while sales and marketing expenses climbed to $5.73 billion.

Operating losses exceed $20 billion; target profitability only in 2030

With high costs coming from all sides, OpenAI’s operating loss in 2025 reached $20.92 billion, accounting for 160% of total revenue (compared with a loss of $8.78 billion in 2024, which was 237% of revenue—though the ratio improved, the absolute amount expanded significantly).

As for net loss, the 2025 figure on the books is as high as nearly $39 billion. The document specifically notes that about $30 billion of this was a one-time accounting adjustment made for investor valuation purposes when the company transitioned to a “for-profit structure.” If this non-recurring item is excluded, the actual net loss is about $8 billion. Faced with such a massive funding shortfall, OpenAI has told investors plainly that it expects to achieve true profitability only by 2030 at the earliest.

Facing competitive pressure from Anthropic; Sora shut down to stop the bleeding

In addition to internal cost-management pressure, OpenAI is also facing severe challenges in the external market. The report’s analysis indicates that enterprise customers are starting to be dissatisfied with Token-based pricing models and are strongly demanding clear return on investment (ROI). At the same time, the rising strength of competitor Anthropic may force OpenAI to lower prices, which in the short term will further intensify its losses.

To stop the bleeding and concentrate resources, OpenAI reportedly made a major move in March 2026 by shutting down the once sensational video generation model “Sora,” and cutting multiple side projects, focusing all efforts on strengthening its core code generation and the enterprise user market. Despite the many challenges it faces, thanks to this year’s historic $122 billion funding round completed in March with an $852 billion valuation, OpenAI still has plenty of cash to maintain its dominance in this AI arms race.