Two racehorses left the gate within a week of each other. Anthropic told the SEC on June 1 that it had confidentially filed to go public. On June 8, OpenAI did the same, with a characteristically wry announcement: "We recently submitted a confidential S-1. We expect it to leak so we're just announcing it."
This morning, the difference between the two horses became clear. Reuters has reviewed Anthropic's confidential draft prospectus and reported its contents: the revenue, the compute bills, the losses, the customer concentration. No equivalent reporting exists for OpenAI. We know the company filed. We do not know what is in the filing. Nearly four months after OpenAI's S-1 reached the SEC, its financials remain dark in a way its rival's no longer are.
That contrast is the story. But it is not the only story. Because even without a prospectus, a steady drip of reported figures, executive comments and signed deals tells us more about OpenAI's shape than the company might prefer. Here is what is credibly reported, what it means for the supply chain behind AI, and what a commerce operator should watch.
What we know, and how we know it
Everything below comes from Reuters, CNBC, Bloomberg, the Financial Times, the Wall Street Journal or company announcements, not from the filing itself. Dates matter, because these numbers move fast.
Revenue is real and accelerating. Reuters reported in March that OpenAI topped $25 billion in annualized revenue as of the end of February, up 17% from $21.4 billion at the end of 2025, citing The Information. Bloomberg reported on August 13 that the run rate had passed $40 billion, roughly doubling from the end of 2025, driven by coding software, subscriptions and a young advertising business. Reuters reported on August 31 that the ad business alone crossed $1 billion annualized. "Run rate" means stretching a recent month's pace across a year. It is not money already earned, but the direction is unmistakable.
Valuation has climbed to near a trillion. CNBC reported OpenAI's March funding round valued it at $852 billion after the money, and CFO Sarah Friar told employees in August that "we raised $122 billion in March, and that gives us flexibility." The New York Times reported in June that Sam Altman pushed his bankers to find a path to a $1 trillion IPO valuation.
Losses and burn are enormous. The Financial Times reported on September 18 that OpenAI projects burning through almost $280 billion by the end of 2030. Reuters could not independently verify that figure, and neither can we. But it fits what the WSJ reported in April: OpenAI had missed multiple monthly revenue targets, lost ground to Anthropic in coding and enterprise markets, fallen short of its goal of one billion weekly ChatGPT users, and Friar had warned colleagues the company might struggle to pay for future computing contracts if revenue did not grow faster.
Timing keeps slipping. In June, CNBC reported OpenAI was gearing up for a debut as soon as the fourth quarter of this year. Days later, the NYT reported the company was leaning toward waiting until 2027. In August, Friar settled it for employees: OpenAI "will be a public company in 2027," maybe sooner if "our business continues to inflect." September, the window some on Wall Street had circled, is nearing its end with no offering, no pricing date and no public S-1. Friar's message about Anthropic was notable: "There is a chance they pull the cover off that confidential file in the coming weeks and become public in September. That's OK, we are running our own race."
Why the numbers are still dark
A confidential filing is not secrecy for its own sake. It lets a company negotiate with the SEC before showing investors, employees and competitors its books. Anthropic is moving toward daylight first. OpenAI chose to wait; that contrast makes the eventual public accounting especially important.
There is also a structural wrinkle Anthropic does not have. OpenAI spent much of the last two years remaking itself from a nonprofit-controlled lab into a public benefit corporation. In October 2025, Microsoft signed a new agreement supporting the change, taking roughly 27% of the new OpenAI Group PBC, valued at about $135 billion, and keeping exclusive IP and Azure API rights until OpenAI achieves AGI, however that milestone gets defined. California's attorney general signed off with conditions, including that the nonprofit keeps control of the PBC through a special class of stock. Going public on top of that machinery is harder than going public on top of a normal company, and the prospectus, whenever it appears, will have to explain all of it in public for the first time.
The supply chain is already public
Here is the paradox for commerce readers: OpenAI's income statement is private, but its shopping list is not. The company has announced its commitments in press releases, because announcing them is part of the strategy.
Stargate, the infrastructure venture with Oracle and SoftBank launched at the White House in January 2025, began as a pledge to build 10 gigawatts of AI data centers for $500 billion. OpenAI says it has already surpassed that, adding more than 3 gigawatts in a single 90-day stretch. The July 2025 Oracle agreement alone covers up to 4.5 gigawatts and more than $300 billion over five years. In September 2025, Nvidia agreed to supply at least 10 gigawatts of systems and invest up to $100 billion in OpenAI. Weeks later, AMD signed a multi-year chip deal worth tens of billions annually, with an option for OpenAI to buy roughly 10% of AMD.
Read those numbers the way a merchant reads a lease. OpenAI has committed to years of rent on compute it has not yet used, betting that demand for intelligence will fill the buildings. The chipmakers, cloud providers, utilities and construction firms on the other side of those commitments collect revenue now. OpenAI collects the bill, and the bet.
The bet is not guaranteed to hold in every location. Bloomberg reported that OpenAI and Oracle dropped plans to expand the flagship Stargate site in Abilene, Texas, after financing complications and shifting demand forecasts. Big commitments can move. So watch the announcements, but also watch the cancellations.
What a commerce operator should watch
First, watch the sequencing. If Anthropic lists first, its public financials become the reference sheet for pricing every AI company, including OpenAI. The multiple investors pay for Anthropic's run rate will anchor what OpenAI can ask in 2027. If you sell to, buy from or build on either company, the first prospectus sets the benchmark for both.
Second, watch for the moment OpenAI's S-1 goes public, likely weeks before a 2027 debut. That document will answer questions the drip-feed cannot: gross margin per dollar of revenue, how much compute spend is training versus serving, customer concentration, and how much of the Stargate obligation is committed versus cancellable. Until then, treat every revenue figure as a run rate from a leak, and every burn figure as a projection.
Third, watch your own vendors. OpenAI's commitments are so large that they will shape chip supply, cloud pricing and power markets that every commerce business touches, whether or not it uses ChatGPT. A supplier racing to fill OpenAI's orders has less reason to discount yours.
Two horses left the gate in June. One just showed the world its racing form. The other is bigger, better funded and still behind the curtain, and the curtain is now part of the race.
Sources and reporting notes
OpenAI's confidential S-1 announcement (June 8): https://openai.com/index/openai-submits-confidential-s-1/ CNBC on the confidential filing and IPO prep (June 8): https://www.cnbc.com/2026/06/08/openai-confidentially-files-for-ipo-prepping-wall-street-for-ai-debut.html Fortune on the filing (June 9): https://fortune.com/2026/06/09/openai-files-confidential-s-1-sec-ipo/ Reuters on $25 billion annualized revenue, citing The Information (March 5): https://www.reuters.com/technology/openai-tops-25-billion-annualized-revenue-last-month-information-reports-2026-03-05/ Bloomberg on $40 billion run rate (August 13): https://www.bloomberg.com/news/articles/2026-08-13/openai-s-revenue-run-rate-tops-40-billion-ahead-of-ipo Reuters on the WSJ report of missed targets (April 27): https://www.reuters.com/business/openai-falls-short-revenue-user-targets-it-races-toward-ipo-wsj-reports-2026-04-28/ Reuters on the FT's $280 billion burn projection, unverified (September 18): https://www.reuters.com/technology/openai-expects-burn-through-almost-280-billion-by-2030-ft-reports-2026-09-18/ Reuters on the $1 billion ad run rate (August 31): https://www.reuters.com/business/media-telecom/openais-ad-business-hits-1-billion-annualized-revenue-run-rate-2026-08-31/ CNBC on Friar's 2027 timeline (August 19): https://www.cnbc.com/2026/08/19/open-ai-ipo-timing-2027-friar.html NYT on leaning toward 2027 and the $1 trillion push (June 25): https://www.nytimes.com/2026/06/25/technology/openai-ipo-artificial-intelligence.html Microsoft on the recapitalization and PBC agreement (October 28, 2025): https://blogs.microsoft.com/blog/2025/10/28/the-next-chapter-of-the-microsoft-openai-partnership/ OpenAI on five new Stargate sites: https://openai.com/index/five-new-stargate-sites/ OpenAI on Stargate progress: https://openai.com/index/building-the-compute-infrastructure-for-the-intelligence-age/ Reuters-syndicated report on the AMD deal and Nvidia terms: https://www.channelnewsasia.com/business/amd-ai-chip-openai-nvidia-deal-technology-5386646 DCD on the canceled Abilene expansion, citing Bloomberg and The Information: https://www.datacenterdynamics.com/en/news/oracleopenai-drop-plans-to-expand-flagship-abilene-stargate-site-meta-in-talks-to-pick-up-crusoe-capacity-with-nvidias-help/
OpenAI's draft S-1 is confidential. Unlike the Anthropic companion story, none of the figures above come from a reviewed prospectus. Revenue figures are run rates from anonymous sources, projections are projections, and no public filing yet exists to check line by line.