On Friday, May 22, 2026, OpenAI quietly filed a confidential S-1 registration statement with the Securities and Exchange Commission. Goldman Sachs and Morgan Stanley are leading the deal. The company is targeting a public listing between Labor Day and Thanksgiving — with September as the early preferred window.
No prospectus is public yet. That’s what “confidential filing” means: the SEC receives and reviews the document privately. The draft becomes public roughly 15 days before the roadshow begins, which means builders will likely get their first look at OpenAI’s official financials sometime in late July or August 2026.
The Numbers That Matter
Before the S-1 details become public, here’s what OpenAI has already disclosed or that analysts have triangulated from funding rounds:
- $25B ARR as of March 2026 (annualized run rate), roughly $2B/month in revenue
- 50 million consumer subscribers (ChatGPT Plus, Pro, Team plans)
- 9 million business users across enterprise and API tiers
- $122 billion raised in the March 2026 funding round at an $852B post-money valuation — the largest private funding round on record
- Targeted IPO valuation: $852B to over $1 trillion, depending on market conditions at listing
Goldman and Morgan Stanley are the same banks Anthropic has in early IPO discussions, per recent Bloomberg reporting — a coincidence that underscores how compressed the AI IPO window is expected to be.
The Timeline from Here
| Milestone | Expected Window |
|---|---|
| Confidential S-1 filed | May 22, 2026 (confirmed) |
| SEC review period | ~60–90 days |
| Public S-1 released | Late July or August 2026 |
| IPO roadshow begins | August–September 2026 |
| Public listing | September–Q4 2026 |
The confidential filing process exists specifically for large, high-profile offerings where premature public disclosure could create market disruption or give competitors advance intelligence. OpenAI qualifies on both counts.
One important note: the filing can be withdrawn. If market conditions deteriorate significantly between now and the roadshow, OpenAI could delay or cancel without any public record of having tried.
What This Means for Builders on OpenAI’s API
The pre-IPO window is probably the most favorable you’ll see
Public companies face quarterly earnings pressure in ways private ones don’t. Right now, OpenAI is still optimizing for growth metrics — user count, API consumption, developer ecosystem size. After listing, they’ll be optimizing for those things and margin expansion, which creates tension with keeping API prices low and access generous.
The Guaranteed Capacity program, launched earlier this year, is an early signal of direction: enterprise customers with committed spend get reserved throughput. Developers without contracts get the residual. This pattern — lock in the big accounts, let everyone else compete for what’s left — is standard practice for any company heading toward a public offering. It will intensify, not soften, post-IPO.
If you’re building a product where OpenAI’s API is mission-critical, now is the time to negotiate terms, not after the S-1 goes public and OpenAI’s enterprise sales team is optimizing for ARR numbers to put in the prospectus.
The S-1 will reveal costs and margin that the AI industry has never disclosed
OpenAI has never published audited financials. When the public S-1 drops, builders will see — for the first time — the actual economics of frontier AI at scale: what training costs, what inference costs, what the margin looks like at $25B ARR, and how the compute procurement agreements with Microsoft shape the P&L.
That document will be more informative than any analyst model built to date. Read it. It will change how you think about the long-term viability of building on any frontier AI provider’s API.
Enterprise-first drift is already in motion
Everything OpenAI has shipped in the past six months — Guaranteed Capacity, the Deployment Company, the $4B deployment venture to help enterprises integrate AI — points toward a business that generates revenue through large committed contracts, not pay-as-you-go API calls.
This doesn’t mean the API goes away. It means the API increasingly becomes a gateway to up-sell enterprise contracts, not a standalone revenue line. The developer experience for individual builders and small teams will reflect that priority order.
Model deprecation may accelerate
Under private ownership, OpenAI could afford to run legacy models indefinitely. Under public market scrutiny, maintenance costs for deprecated models become a quarterly earnings drag that analysts will ask about. Expect deprecation timelines to tighten after listing as the company rationalizes its model portfolio for profitability.
What to Watch in the Public S-1
When the prospectus drops (estimated late July / August 2026), these are the numbers to examine:
Gross margin — Frontier AI infrastructure is expensive. If margins are thin (common in early AI companies), public investors may push for rapid cost-cutting that translates to API pricing changes or model access restrictions.
Revenue concentration — What percentage of revenue comes from the top 10 enterprise accounts? High concentration is a risk flag for public investors and signals continued enterprise-first product prioritization.
Microsoft dependency — OpenAI’s compute agreement and Azure distribution relationship with Microsoft has been the subject of speculation for years. The S-1 must disclose material relationships with major customers and partners. This will be the first public accounting of what that relationship actually looks like financially.
API vs. consumer split — How much of the $25B ARR is API/enterprise vs. direct consumer subscriptions? The answer shapes which roadmap gets investment post-listing.
Unit economics for Claude competition — The S-1 won’t mention Anthropic by name in favorable terms, but analysts will immediately compare OpenAI’s disclosed metrics to Anthropic’s known revenue trajectory (currently ~$30B ARR per April 2026 reports). That comparison will drive the IPO pricing conversation.
The Bigger Picture
OpenAI going public is the closing of one chapter in AI and the opening of another. The era when frontier AI companies could run at enormous losses while optimizing purely for capability milestones is ending. Public shareholders demand a path to profitability, and that path runs through margin expansion, enterprise contracts, and pricing discipline.
For builders, this is not bad news — it’s clarifying news. A public OpenAI is more predictable in some ways than a private one. Quarterly earnings calls create accountability. The S-1 creates transparency. And the competitive pressure from Anthropic, Google, and others means OpenAI can’t simply extract rent from developers without risking ecosystem defection.
The window between now and the roadshow is worth using deliberately. Read the S-1 when it drops. Audit your API dependencies. Understand which parts of your product are truly locked to OpenAI and which could move.
The company that files its S-1 is different from the company that listed. And the company that lists will be different from the one you’ve been building on.
OpenAI’s confidential S-1 was filed May 22, 2026. The public prospectus is expected in late July or August 2026, approximately 15 days before the IPO roadshow begins. Goldman Sachs and Morgan Stanley are the lead underwriters. No listing date has been officially announced.