OpenAI takes its first step toward an IPO – and will soon have to show the numbers

OpenAI has taken the first formal step toward a public listing. According to reporting by CNBC, Axios and Fortune, the company submitted confidential IPO paperwork to the US Securities and Exchange Commission last week, targeting a listing sometime between September and November this year. Goldman Sachs, Morgan Stanley and JPMorgan Chase are working on the process.

The company will neither confirm nor deny the details. A spokesperson sticks to a line that OpenAI «as part of normal governance, regularly evaluates a range of strategic options», with «focus on execution». Sam Altman has stressed that filing IPO paperwork is not the same as being ready to go public. A confidential filing is exactly that: a draft the SEC reviews privately, before a company decides whether to proceed with a public prospectus.

For business leaders it is tempting to fixate on the number everyone is talking about: a possible valuation approaching one trillion dollars. That is the wrong place to look.

The valuation is not the point

OpenAI was last valued at 852 billion dollars, after a 122-billion-dollar round in late March led by SoftBank. Whether the IPO lands there or closer to a trillion is a question for fund managers, not for a CIO in Oslo.

What actually matters to companies is that one of their most important AI vendors will soon have to do something it has never done before: publish audited figures. An IPO forces a prospectus, and a prospectus forces honesty about the economics.

Fortune lays out the questions the document will eventually have to answer: How fast is OpenAI burning cash on model training and infrastructure? How do revenues split across ChatGPT subscriptions, enterprise deals, the API and Codex? What does it actually cost to serve the models per request? And how much does Altman himself own?

The few numbers already known invite sobriety. OpenAI generated nearly six billion dollars in revenue in the first quarter, but the company is deeply unprofitable. Leadership reportedly grew uneasy after both internal revenue targets and user growth fell short. ChatGPT has stalled around 900 million weekly active users, below what the company had projected.

A loss-making vendor under market pressure

That combination is what boards and executives should note. Over the past two years many organisations have moved real operations – customer service, software development, case handling, analysis – onto OpenAI's models, directly or through Microsoft. The dependency is no longer a pilot. It is production.

When such a vendor goes public while losing money and relying on enormous compute investment, the incentives shift. A listed company answers to shareholders who want a path to profit. That path usually runs through price increases, changed terms, tightened free tiers and harder prioritisation of the highest-paying customers. This is not a prediction but an ordinary dynamic for companies that have to prove profitability to a new audience.

For a CFO that means today's API prices should be treated as perishable, not as a fixed line in the 2027 budget. For a CIO it means architecture that locks the business to a single model vendor becomes a strategic risk worth quantifying. For a CISO and a board it is about vendor concentration: how much daily operation can absorb becoming more expensive, slower or changed at short notice?

What leaders should do now

There is no need for panic, but there is need for preparation. Three steps are sober and concrete.

First, map the actual exposure. Which processes run on OpenAI today, how business-critical are they, and what do they cost per month? Many organisations know this surprisingly poorly because usage has grown in a decentralised way.

Second, build in portability. An abstraction layer that lets you switch models without rebuilding the whole solution is cheap insurance against both price hikes and service disruption. The same goes for keeping at least one alternative vendor in real, tested use – not just on paper.

Third, read the prospectus when it lands. When OpenAI finally publishes the figures, buyers get, for the first time, a factual basis for understanding the vendor's cost structure and incentives. It is rare to see this deep into a critical supplier's economics. Use it in the next contract negotiation.

OpenAI's path to the public markets is first and foremost a financial story. But it opens a window companies have not had before: the chance to assess a decisive vendor on numbers rather than promises. That window should be used before it closes again behind a share price.

Sources and media

• Fortune, «The big questions OpenAI's trillion-dollar IPO filing may finally answer», 22 May 2026: source_url https://fortune.com/2026/05/22/openai-ipo-filing-1-trillion-may-finally-answer-these-big-questions/
• CNBC, «OpenAI to confidentially file for IPO as soon as Friday», 20 May 2026: https://www.cnbc.com/2026/05/20/openai-ipo-filing.html
• Axios, «OpenAI prepares confidential IPO filing», 20 May 2026: https://www.axios.com/2026/05/20/openai-ipo-spacex-musk
• Valuation and round figures from OpenAI's March 2026 raise as reported by the same outlets. Spokesperson quote via Axios.
• Thumbnail: OpenAI Image 2 / hogby.ai