OpenAI keeps shattering revenue records. Yet its losses grow even faster. The artificial intelligence company posted $13 billion in revenue for 2025 while burning through more than $20 billion. By early 2026 its annualized run rate climbed above $25 billion. Cash outflow hit $3.7 billion in the first quarter alone. The disconnect raises hard questions ahead of a planned initial public offering.
Leaked financial statements paint a clear picture. Revenue jumped from $3.7 billion in 2024 to $13.07 billion in 2025. Cost of revenue rose to $7.5 billion. Research and development swallowed $19.18 billion. Sales, marketing and administrative expenses added another $7.3 billion. Total costs reached $34 billion. The operating loss landed at $20.92 billion. Twenty-one billion dollars lost. On $13 billion brought in.
But look closer. The expense ratio improved. In 2024 the company spent $2.37 for every dollar earned. That fell to $1.60 in 2025. Growth outpaced the burn. Still, the absolute numbers shock. Fortune first reported the figures after they reached blogger Ed Zitron and the Financial Times.
By February 2026 the annualized revenue figure had climbed further. Sacra estimates put it at $25 billion. Enterprise customers now drive more than 40 percent of that total. They stand on track to match consumer revenue by the end of the year. ChatGPT itself passed one billion monthly active users in May. The product side looks unstoppable.
Yet infrastructure costs threaten to overwhelm. Inference expenses alone reached $8.4 billion in 2025 and are projected to hit $14.1 billion this year. Sacra forecasts cash burn of $27 billion in 2026 and $63 billion in 2027. The company does not expect positive cash flow until 2030. Those numbers come after OpenAI already raised $178 billion in total funding. Its post-money valuation sits at $852 billion following a March round co-led by SoftBank.
OpenAI races to scale compute while chasing profitability
Executives have dialed back some earlier ambitions. Plans once called for $1.4 trillion in infrastructure commitments. That target dropped to $600 billion by 2030. Even so the spending remains enormous. The company has secured stakes in massive data center projects including the Stargate initiative. Partnerships with Microsoft, Oracle, NVIDIA and Broadcom supply chips and cloud capacity. But power, chips and data centers do not come cheap.
Enterprise adoption offers one bright spot. Companies pay for reliability, security and scale. API usage contributes 15 to 20 percent of revenue. Higher-tier subscriptions such as ChatGPT Pro at $200 per month deliver outsized usage. Some subscribers burn $8,000 to $14,000 worth of tokens monthly under those plans. The model converts free users into paying ones at impressive rates. But average revenue per user still trails rivals such as Anthropic according to recent X discussions and reports.
Recent leaks show the pressure. In the first three months of 2026 OpenAI spent $3.7 billion while generating $5.7 billion in revenue. More than half the top line vanished in that single quarter. The Information obtained the shareholder documents that revealed the burn. Reuters quickly picked up the story. The figures fueled fresh debate about when profitability might arrive.
OpenAI itself pushes back on some narratives. It notes revenue now runs at $2 billion per month. That pace grows four times faster than the early trajectories of Alphabet or Meta. Within a year of launching ChatGPT the company reached $1 billion in annual revenue. By late 2024 it hit $1 billion per quarter. Momentum continues. Enterprise now exceeds 40 percent of total revenue. The company expects parity between enterprise and consumer segments by year end.
Valuation chatter refuses to cool. Some analysts target more than $1 trillion at IPO. OpenAI filed its draft S-1 confidentially in June. A listing could come as soon as September though many expect 2027. The private mark sits at $852 billion after the $122 billion raise closed in March. SoftBank, Andreessen Horowitz, Microsoft, NVIDIA and others joined that round. Talks even surfaced about giving the U.S. government a 5 percent equity stake worth roughly $42.6 billion. That idea drew immediate scrutiny on X where users questioned the governance implications.
Competitors add urgency. Anthropic reportedly reached a $47 billion annualized run rate in recent months while OpenAI hovers between $25 billion and $33 billion. Some posts claim Anthropic passed OpenAI on revenue despite smaller user numbers. The gap highlights different strategies. Anthropic focuses on higher pricing and per-user revenue. OpenAI bets on massive scale and broad adoption. Both pour money into frontier models. Neither has cracked consistent profits.
Product releases keep coming. GPT-5.5 Instant rolled out with 52.5 percent fewer hallucinations on high-stakes tasks. New agents, voice capabilities, coding tools and video generation expand the platform. Sora continues to evolve. Ads have begun testing in free tiers. Internal forecasts once predicted $2.5 billion in advertising revenue for 2026 rising to $100 billion by 2030. Those numbers remain speculative but show the direction.
The nonprofit roots still shape decisions. OpenAI operates a hybrid capped-profit structure with the original nonprofit holding ultimate control. That arrangement helped attract early talent and capital. It also creates complexity for public markets. Investors want clear governance and a path to returns. The IPO process will test whether those structures hold.
So the company stands at a crossroads. Revenue growth looks historic. Losses look unsustainable. Compute demands keep rising. Enterprise customers deliver higher quality income. Consumer usage explodes. The bet is that scale eventually drives margins higher. Inference costs must fall. Model efficiency must improve. New revenue streams from agents, commerce and vertical applications must materialize.
Recent web coverage reinforces the tension. Tech Insider detailed the confidential S-1 filing and $850 billion valuation range. Sacra’s ongoing tracking shows the $25 billion run rate and rising burn. X conversations on July 9 highlighted Anthropic’s revenue overtake and concerns about OpenAI offering cheaper models that could erode margins. The debate plays out in real time.
OpenAI projects $280 billion in revenue by 2030 in some internal documents. It also forecasts continued losses until then. The gap between those two lines will define the next several years. Public investors will soon get their chance to judge whether the math works. For now the company keeps spending. It keeps shipping. And the losses keep growing.
WebProNews is an iEntry Publication