Leaked OpenAI Numbers Shift the Narrative: 13 Billion in Revenue, Still Deep in the Red

Monday, 22 June 2026 at 17:40

Gelekte OpenAI-cijfers veranderen het beeld 13 miljard omzet, toch diep in het rood

In 2025, OpenAI generated about $13 billion in revenue but simultaneously reported a net loss of more than $38.5 billion. Those figures, drawn from leaked financials later reviewed by the Financial Times, sparked a firestorm last week. Social media quickly jumped to a headline-grabbing takeaway: AI isn’t just wildly expensive to build—it’s nearly impossible to sell without massive marketing budgets.

A closer look tells a far more interesting story. OpenAI’s numbers reveal the true cost of the fight for AI dominance—not just in technology, but in infrastructure, talent, distribution, and market share.

What are the real numbers?

According to the leaked annual report, OpenAI in 2025 posted:

$13.07 billion in revenue
$34 billion in total expenses
$19.18 billion in research and development
$5.73 billion in sales and marketing
$1.57 billion in general expenses
$38.5 billion net loss

At first glance, the figures are shocking. For every dollar of revenue, OpenAI spent nearly three.

But the top-line numbers don’t tell the whole story.

Why that $39 billion loss is misleading

The most-quoted figure is the roughly $39 billion net loss—suggesting a company financially off the rails.

The Financial Times reports, however, that a massive one-time, non-cash accounting charge is embedded in that loss, tied to OpenAI’s complex restructuring ahead of an eventual IPO. Tens of billions in non-cash costs flowed through the income statement as a result.

Strip out those exceptional items, and the operational loss lands around $8 billion, according to the documents. Still huge—just far less dramatic than the viral $39 billion headline.

That distinction matters because operating losses are a better measure of how the core business is actually performing.

The marketing take needs context

Much of the online debate fixated on a single claim: OpenAI spent 44 percent of its revenue on marketing.

That number comes from the roughly $5.7 billion listed as sales and marketing. Divided by $13 billion in revenue, you indeed get about 44 percent.

But context is everything.

OpenAI isn’t just selling a consumer app. It’s simultaneously trying to:

retain hundreds of millions of ChatGPT users;
convert enterprises to ChatGPT Enterprise;
lock in developers on its API platform;
win government contracts;
stay ahead of Google, Anthropic, and xAI.

In fast-growing software companies, sales and marketing are often among the largest expense lines for years. Enterprise sales, in particular, are costly—and big organizations don’t sign in a week.

So the claim that “AI products don’t sell without ads” doesn’t automatically follow from these numbers.

The real cost center: compute

The standout expense isn’t marketing—it’s research and infrastructure.

OpenAI spent about $19 billion on R&D in 2025—more than triple total sales and marketing.

Behind that lies a fundamental reality of generative AI.

Modern AI models demand enormous amounts of:

GPUs;
data centers;
electricity;
storage capacity;
specialized AI researchers.

OpenAI has previously said it expects to invest hundreds of billions of dollars in compute by 2030. Reuters reported earlier this year the company projects roughly $600 billion in total compute spending by decade’s end.

That underscores why today’s AI firms aren’t like traditional software companies.

Facebook could serve billions with relatively low incremental costs. With AI chatbots, every additional user drives immediate compute costs.

Demand for AI is anything but weak

Another takeaway in the numbers has gotten surprisingly little attention.

OpenAI jumped from about $3.7 billion in revenue in 2024 to over $13 billion in 2025—more than 250 percent growth in a year.

The Financial Times adds that the company scaled from roughly $1 billion in quarterly revenue at the end of 2024 to around $2 billion per month by the end of 2025.

Few tech companies have ever posted growth like that.

The documents therefore suggest the opposite of what some critics argue. Demand isn’t the problem. The challenge is that the cost of staying on top is, for now, growing even faster than revenue.

Why this matters for the entire AI industry

OpenAI’s figures may be the clearest look yet at the economics of generative AI.

For years, the debate centered on whether AI was useful enough. That question is fading. Businesses and consumers are already paying billions for AI services.

The new question: can the sector ever reach the scale needed to earn back its massive infrastructure bills?

That matters well beyond OpenAI. Rivals like Anthropic, Google, Meta, and xAI are also pouring tens of billions into AI data centers, chips, and model development. The next few years will reveal whether those bets produce durable margins—or whether AI remains a structurally capital-intensive industry.

For now, the leaked numbers point to one conclusion: this AI race won’t be won by the company with the slickest chatbot, but by the one willing to invest tens of billions a year, for as long as it takes.