Revenue is skyrocketing, and cash is disappearing just as fast. Wall Street is set to decide whether to buy into the most expensive IPO of all time—even with this hole in the budget.
When a company triples its revenue year-over-year, it’s usually cause for celebration. But with OpenAI, it’s more complicated. According to documents the company sent to its shareholders— first reported by The Information—the creator of ChatGPT generated $5.7 billion in revenue in the first quarter of 2026. At the same time , however , it burned through $3.7 billion in cash during the same period.
Both figures tripled year-over-year. And that symmetry is precisely where the problem lies.
Growth That Undermines Itself
Revenue of $5.7 billion in a single quarter is something almost every tech company in the world would envy. When we add the fact that they’re growing at three times the rate, we’re talking about one of the fastest-growing businesses in history. The catch is that the costs of achieving this growth are rising just as fast.
In the real world, that’s called a problem. Large companies usually gain what’s known as “operational leverage” over time—the more they sell, the cheaper it becomes to serve each additional customer. At OpenAI, this isn’t the case yet. What it sells is inference using state-of-the-art AI models, and that becomes more expensive—not cheaper—with every additional user.
The specific figures for the quarter are as follows:
Revenue: $5.7 billion (roughly 125 billion Czech koruna)
Cash burn: $3.7 billion
Operating loss: $9.3 billion
Net loss: $21.3 billion (largely due to a one-time accounting item of $12.4 billion)
If we break this down into simple math, as Barron’s analysts did, OpenAI loses roughly $1.22 for every dollar it earns. Revenue has tripled, but the loss ratio has remained the same. The company is growing and burning through cash at the same time—and it’s doing both on a record scale.
Why This Isn’t Like Building a Mobile Network
OpenAI’s defenders like to argue that this is a classic capital-intensive infrastructure project—expensive at the start, cheap afterward. But this parallel doesn’t hold up.
Take the construction of a mobile network. Building a cell tower costs a fortune, but once it’s up, it serves a million users for the same cost as a thousand. The cost per customer decreases as scale increases. It is precisely this economics that has made telecommunications and software such profitable businesses.
OpenAI’s GPU clusters don’t work that way. A cluster handling a million queries costs proportionally more than one that handles a thousand. Marginal cost does not decrease with scaling—it increases. And this is precisely what makes OpenAI a business that is structurally different from any trillion-dollar company that came before it.
According to data from the analytics firm Sacra, the cost of inference alone reached $8.4 billion in 2025 and is expected to climb to $14.1 billion this year. That’s a 68 percent increase.
“The implied multiple of 35 times future revenue is priced for a monopoly outcome that does not yet exist.”
Greg Jensen, co-head of the investment division at Bridgewater
A cushion that looks better than it actually is
One number sounds reassuring. At the end of the quarter, OpenAI had over $73 billion in cash and marketable securities, up from $40 billion at the end of December. At the current burn rate, that amounts to roughly a five-year reserve without the company having to raise a single additional dollar.
But that jump from $40 billion to $73 billion didn’t come from business operations. It is largely the result of a massive funding round closed in late March —a $122 billion round, the largest private tech funding round in history, backed by Amazon $AMZN, Nvidia $NVDA, and SoftBank $SFTBY. This round valued the company at $852 billion.
So the cushion is real. But it’s also freshly inflated from the outside, not earned. And that’s a pretty significant difference when monthly spending is counted in the billions.
A trillion-dollar IPO with an uncomfortable question lurking beneath the surface
These numbers aren’t coming at a random time. On June 8, OpenAI confirmed that it had confidentially filed for a U.S. initial public offering (IPO), which could take place as early as September and value the company at up to $1 trillion. That would make it the largest stock market debut in history. Goldman Sachs $GS and Morgan Stanley $MS are leading the IPO.
This is precisely where the quarterly numbers become ammunition for both sides. The bulls point to the revenue curve; the bears point to the chasm below it. Neither side is clearly right, and the S-1 prospectus, which the company will publish before the actual IPO, will be the first opportunity to see the audited financials.
There is no shortage of skeptics. HSBC $HSBC estimates that OpenAI may need over $207 billion in additional capital by 2030, even under an optimistic scenario. Furthermore, internal documents suggest that the company expects a loss of around $74 billion in 2028 alone, before turning a profit in 2030. Meanwhile, it has committed to computing capacity contracts worth up to $1.4 trillion over eight years.
"These are great companies. But a great company doesn’t automatically mean a great investment."
Jay Ritter, director of the IPO Initiative at the University of Florida
A comparison with the competition, however, is not flattering. While OpenAI is losing $1.22 for every dollar of revenue, rival Anthropic is expected to report a small operating profit for the second quarter, according to Barron’s, after doubling its revenue to an annualized $47 billion.
The whole bet can thus be summed up in a single sentence. Investors are supposed to buy a company for $1 trillion that is banking on inference costs falling before it runs out of money. A five-year cash reserve buys time. But it doesn’t buy certainty that the model’s economics will ultimately balance out.