The internal numbers landed at a difficult moment. OpenAI lost $3.7 billion in the first quarter of 2026, according to The Information, putting the company on a pace to burn close to $15 billion this year — a rate that would rank among the largest cash-burn figures in corporate technology history even as the company prepares to sell shares to the public.
The loss is not a surprise inside the company. OpenAI’s spending is dominated by two line items: computing, both for training frontier models and for serving them to hundreds of millions of users, and the multi-year commitments tied to its infrastructure partnerships. Executives have argued for two years that heavy investment now is the price of staying at the frontier, and that revenue growth will eventually outrun the burn. The numbers support part of that case — revenue has continued to climb sharply — but they also give regulators, investors and competitors a concrete figure to point at.
The company’s response to the scale of the losses is visible in its hiring. According to TechCrunch, citing people familiar with the plans, OpenAI is bringing in a series of senior industry executives and advisers ahead of its IPO, a wave of additions aimed at the operational and governance gaps that public investors and their lawyers scrutinize. The hires span finance, compliance and regulatory affairs, the people said, and several are people who have taken technology companies through listings before.
The timing reflects the market window. OpenAI’s shares trade richly in the private market, and the company’s leadership has signaled for months that it intends to go public while the AI trade remains hot. But the same period has brought a second wave of attention: attorneys general in multiple states have opened investigations into whether the company’s products cause user harm, according to people familiar with the inquiries. The probes touch on everything from the accuracy of consumer-facing features to data practices, and they create a compliance workload that the company is now staffing against.
The tension is fundamental. A public offering requires disclosing financials that make the burn rate explicit, submitting to audit scrutiny that private investors tolerated informally, and answering for regulatory risk in a prospectus that lawyers will parse line by line. OpenAI’s own board has debated the sequencing, according to people familiar with the discussions — whether to raise more private capital and delay the listing, or take the process on now while enthusiasm for AI stocks is high.
The company’s competitors are watching the same math. Anthropic, OpenAI’s closest rival, has not disclosed comparable quarterly figures, but its own infrastructure commitments are believed to be of a similar order of magnitude, according to analysts who track the sector. Google DeepMind operates inside a parent that can absorb losses indefinitely. For OpenAI, the difference is that the money must eventually come from public markets, which is why the IPO preparations and the loss disclosure are arriving in the same news cycle.
What the quarterly numbers obscure is the revenue side of the ledger. OpenAI’s product lineup, from consumer subscriptions to API access used by a large share of the industry’s applications, has grown fast enough that the loss, while enormous in absolute terms, is narrowing relative to sales, according to people familiar with the company’s internal forecasts. The question for the public offering is whether investors will price the company on the trajectory or on the burn.
There is also a political dimension. AI regulation is a live issue in Washington and in state capitals, and a company heading into an IPO with open state investigations faces a different kind of scrutiny than a private firm. The hires OpenAI is making, people familiar with the matter said, are designed to demonstrate that the company treats the investigations seriously and has the bench to manage them.
For the wider market, OpenAI’s disclosures serve as a benchmark. Every AI company that follows will now be measured against the same standard of transparency, and investors will be watching whether the burn produces durable revenue. The first quarter of 2026 put a number on the cost of the frontier, and the number is large enough that the IPO, whenever it comes, will be less a celebration than a test.
The burn rate, expressed in annualized terms, exceeds the annual revenue of most companies in the S&P 500, and it dwarfs the early losses of the consumer internet giants that previously defined aggressive scaling. What distinguishes OpenAI’s position is that the spending is concentrated in compute, a cost that tends to fall as hardware improves, rather than in marketing and headcount, which are sticky. Executives have argued, according to people familiar with their internal presentations, that each generation of models is more capable and cheaper to serve than the last, and that this is the mechanism by which the losses eventually shrink. The state attorneys general investigations add a layer of uncertainty that the company’s lawyers are now pricing into every product launch, the people said, and the compliance hires are arriving faster than the investigators’ requests.
OpenAI declined to comment on the figures beyond pointing to its public statements about investment discipline. People familiar with the company’s plans said the IPO remains on track for this year, with timing dependent on market conditions and the progress of the regulatory work now underway.
