OpenAI filed a confidential S-1 draft registration statement with the SEC this week, with Goldman Sachs and Morgan Stanley as lead underwriters and JPMorgan also on the deal. The target valuation runs $850 billion to $1 trillion. The public listing window is Labor Day to Thanksgiving — four-to-six months from confidential filing to ticker.

That timing is doing more work than the number. The filing landed within days of Anthropic closing a $30 billion Series H at a $900 billion pre-money valuation — the first time in this market that a rival has overtaken OpenAI’s private mark. The same week, SpaceX’s own IPO prospectus disclosed that Anthropic is paying $1.25 billion per month for GPU compute from xAI’s Colossus 1 facility. Three pieces of news. One inflection point.

OpenAI is going public against a competitor that, by some marks, is now worth more. The IPO is the receipt.

Quick Summary: OpenAI’s Confidential S-1 Filing

Detail	Info
Filing date	On or around May 22, 2026
Filing type	Confidential S-1 draft registration statement
Lead underwriters	Goldman Sachs, Morgan Stanley
Additional banker	JPMorgan
Target valuation	$850 billion to $1 trillion
Listing window	Labor Day to Thanksgiving 2026 (Q4)
Runway from filing to listing	4-6 months
Most recent private mark	$852 billion ($122B April raise)
Competitive context	Anthropic now marked at $900B pre-money
Primary source	Axios reporting

Bottom line: OpenAI just put a roadshow clock on the AI lab race. Public-market discipline is coming for ChatGPT pricing, enterprise contract terms, and product roadmap priorities in ways that change how anyone building on the API should plan the next two quarters.

---

What a Confidential S-1 Filing Actually Means

A confidential S-1 filing is a draft registration statement submitted to the SEC that lets a company start the IPO review process without making its financials public yet. Under SEC rules, the full statement becomes publicly available on EDGAR at least 15 days before the roadshow begins. OpenAI’s financials should hit public view in August or September if the Q4 listing window holds.

Practical translation: the SEC sees the books now. The underwriters see them now. The public won’t see them for several months. The road from confidential filing to public listing typically runs four to six months — which lines up cleanly with the Labor Day to Thanksgiving window the bankers are modeling against.

The underwriter selection is the second tell. Goldman Sachs and Morgan Stanley are the two banks that have run nearly every megacap tech IPO of the last decade. JPMorgan on the deal signals breadth — the kind of bookrunning team you assemble when you’re trying to absorb tens of billions of public-market demand without leaving allocation on the table.

OpenAI’s most recent private mark, set by the $122 billion April raise, was $852 billion post-money. The IPO target range starts at $850B and runs to $1T. That’s a tight floor and a wide ceiling — bankers signal a $1T outcome they think is achievable, anchored to a private mark they don’t want to undercut.

The Competitive Context Nobody at OpenAI Wanted

OpenAI did not pick this week. Or maybe they did, and that’s the more interesting read.

Three news beats landed inside a five-day window:

1. Anthropic’s Series H at $900B. First time OpenAI has been ranked second on private marks. Revenue at $45B+ annualized, gross margin on inference at 70%+, Claude Code anchoring most of the curve.
2. OpenAI’s S-1 filing. Goldman, Morgan Stanley, JPM. Target $850B-$1T. Q4 window.
3. SpaceX prospectus disclosure on Anthropic’s compute. Anthropic is paying $1.25 billion per month — $15 billion annualized — to xAI for GPU capacity at the Colossus 1 facility. First hard number anyone has put on Anthropic’s compute commitments outside the Google and Amazon deals.

Stack those together. The market just learned that Anthropic is burning $15B/year on third-party compute alone, has $45B in annualized revenue, and was priced ahead of OpenAI. OpenAI’s response is to file an S-1 and start the roadshow clock.

Either OpenAI was going to file in May regardless and the Anthropic timing is coincidence, or OpenAI moved the timing forward to define the IPO narrative before Anthropic does. The second reading is more interesting and probably more correct. Anthropic’s October IPO window has been telegraphed for months. OpenAI going public first puts the OpenAI valuation in the public market before Anthropic prices, which is a meaningful piece of competitive choreography.

What Going Public Changes for AI Professionals

The financial press will spend the next two months on the share-price drama. The practical question for anyone running ChatGPT, the API, or an OpenAI-built enterprise workflow is what an IPO changes about pricing, product, and contract terms.

Three things shift the moment a private company becomes a public one.

1. Pricing discipline goes up, customer-friendly experiments go down

Public companies report quarterly. The S-1 will tell shareholders what OpenAI’s revenue lines look like — likely some breakdown of ChatGPT consumer subscriptions, ChatGPT Enterprise, API revenue, and the emerging advertising business. Once those lines are public, every quarter becomes a referendum on whether each one is growing.

The implication: the “discount aggressively to win the enterprise contract” mode that OpenAI ran throughout 2024-2025 gets harder to defend. So does the “give it away free to build mindshare” play that built ChatGPT’s user base. Public-market analysts will ask why ARPU is flat. They will ask why discount accretion is dilutive. They will get answers, and the answers will favor pricing discipline.

If you’re negotiating an OpenAI enterprise contract in Q3 2026, you’re negotiating against the last private cycle of pricing flexibility. After the IPO, the discount machinery slows down. Lock in your multi-year terms now if you have leverage. Wait six months and the same volume tier may price 15-25% higher with thinner room for custom terms.

2. Product roadmap pressure shifts toward revenue, away from research

OpenAI’s history is half research lab, half product company. The S-1 will force the company to break out R&D as a line item. Public-market investors will model that line. They will pressure for ratio shifts — more spend on revenue-generating product surfaces, less on capability research that doesn’t ship.

The research bets don’t disappear. They get justified differently. Frontier model training continues because the next model opens the next pricing tier. Pure-research outputs like alignment papers and capability evaluations get rebranded as risk management. The internal calculus around “should we ship this” gets a louder revenue voice in the room.

Downstream: the things that ship are the things that move quarterly numbers. Expect more product velocity on ChatGPT Enterprise, the API surface, and ads. Expect slower releases on long-horizon research projects. The GPT-5.5 cadence and capability tier choices get evaluated against a revenue model, not a capability roadmap.

3. Enterprise contract terms get more conservative

Public companies don’t make handshake deals. Procurement teams who got custom indemnification, data-residency carve-outs, or pricing locks during the private era will discover that the post-IPO version of OpenAI has shorter answers on the same questions. Boilerplate gets more boilerplate. Exceptions take longer to clear.

If your organization is in active OpenAI procurement right now, get the unusual asks into a signed contract before the listing. The lawyers writing public-company commercial terms are not the same lawyers who wrote private-company commercial terms, even when it’s the same firm.

What’s in the S-1 We Don’t Know Yet

A confidential filing means the financials are with the SEC and the underwriters, not the public. Roughly 15 days before the roadshow starts — call it August or September if the Q4 window holds — the full registration becomes available on EDGAR. At that point, the open questions get answered.

What we’ll find out:

• Revenue by line. ChatGPT consumer subscriptions, ChatGPT Enterprise, API, Sora, advertising. Each line gets a number. Each number gets a growth rate.
• Gross margin on inference. Anthropic’s number is reportedly 70%+. OpenAI hasn’t broken theirs out publicly. The S-1 will. The delta between the two is the first hard read on whether OpenAI’s consumer-and-platform stack runs hotter or colder per dollar than Anthropic’s developer-and-enterprise stack.
• Capacity commitments. Disclosure rules require material contract disclosure. Microsoft Azure spending, the new AWS Bedrock relationship, the Oracle commitments — these all get itemized.
• Microsoft economics. The revenue-share arrangement with Microsoft, the IP licensing structure, the AGI-clause language — all of it gets disclosed in some level of detail. This is the single most-watched part of the filing for institutional investors.
• Concentration risk. Customer concentration, geographic concentration, regulatory exposure. The 10-K boilerplate has to start somewhere, and the S-1 is where it starts.

For competitors, partners, and customers, the August-September timeframe will be the first time anyone outside OpenAI’s inner circle has hard data on what the actual business looks like. That’s a bigger deal than the IPO itself.

The Bigger Picture: Two Trillion-Dollar Listings Inside Six Months

The AI lab IPO calendar through 2026 is now compressed in a way that has no precedent.

OpenAI filed in May 2026 with a Q4 listing target. Anthropic has its October window telegraphed for months but no S-1 filed yet. SpaceX published its prospectus the same week, also targeting Q4.

Three frontier private companies, three trillion-dollar-ish marks, one quarter. The public-market capacity to absorb that volume of new AI-adjacent equity issuance is the unspoken risk on every one of these deals. If macro conditions hold, all three price strong. If sentiment cools, the second mover prices flat and the third prices below private marks.

OpenAI going first matters here. Whoever lands the first frontier-AI IPO sets the comp for everyone else. A successful debut at $1T builds the runway for Anthropic to price at parity. A flat debut at $850B forces Anthropic to discount. The S-1 timing isn’t just a competitive move against Anthropic — it’s a sequencing move for the entire 2026 AI IPO calendar.

Our Take

Filing this week is the right move for OpenAI’s specific position. Filing later means letting Anthropic define the AI IPO narrative through Q3, and there’s no version of that story where OpenAI comes out as the more attractive listing. Better to be the benchmark than the comparison.

What I’d push back on is the gap between the private mark and the operational story. The $852B April raise was priced on growth assumptions that the public market will scrutinize harder than the private syndicate did. The ads revenue ramp is plausible but unproven beyond six weeks. The Microsoft economics are a tax on the gross margin line. The capacity commitments are large enough to look scary in a 10-K risk-factors section.

My read: OpenAI prices the IPO somewhere in the middle of the $850B-$1T range — call it $900-950B — and the share trades sideways for the first quarter as analysts work out the model. That’s not a failure. It’s the normal arc of a frontier private company hitting public-market valuation gravity for the first time. The interesting question is what the second quarter looks like, when the next Anthropic round or Microsoft’s MAI-model hedge against OpenAI starts to reshape the competitive frame.

For AI professionals using ChatGPT and the OpenAI API today, the practical advice is short. Lock pricing now if you can. Read the S-1 the day it goes public on EDGAR. Don’t rebuild your stack around the IPO outcome — the product you’re using on December 1 will be the same product you’re using today, just slightly more expensive and slightly more conservative around the edges. That’s not a tragedy. It’s the post-IPO version of every successful private-to-public transition.

The bigger picture is the one to watch. Two AI labs at frontier valuations, three months apart, in the same public market. Whatever that market decides about the first one shapes how it prices the second. We’ll know which read holds by December.

Frequently Asked Questions

What does “confidential S-1 filing” actually mean for OpenAI?

A confidential S-1 is a draft registration statement that lets OpenAI begin SEC review of its IPO paperwork without publishing financials yet. Per SEC guidance, the full statement becomes publicly available on EDGAR at least 15 days before the roadshow starts. OpenAI’s financials should hit public view in August or September if the Q4 listing window holds.

When will OpenAI actually start trading on a stock exchange?

The target window is between Labor Day (early September) and Thanksgiving (late November) 2026. The exact date depends on SEC review timing, market conditions, and the pace of the roadshow. Treat Q4 2026 as the realistic planning range, not a calendar lock.

What ticker symbol will OpenAI use?

Not yet disclosed. Exchange selection (NYSE vs. Nasdaq) and ticker symbol get confirmed in the public registration statement, not the confidential filing.

Will this make ChatGPT more expensive?

Probably yes over time, but not immediately because of the IPO itself. The structural pressure to grow ARPU and margin is permanent for public companies. Expect aggressive consumer-pricing experiments and tighter enterprise-discount discipline starting in the first full quarter as a public company. The day-of-listing change is zero. The 18-month trend is meaningful.

What happens to OpenAI’s relationship with Microsoft?

The Microsoft revenue-share arrangement, IP licensing structure, and AGI clause language all get disclosed in the public S-1. That’s the most-watched single section of the filing for institutional investors. The relationship doesn’t change because of the listing — it just becomes legible to outside readers.

How does this compare to Anthropic’s IPO timing?

OpenAI’s filing puts pressure on Anthropic to file soon if it wants to hit its October target. Anthropic raised $30B in a Series H at $900B pre-money this month, which buys runway to wait if conditions sour, but the company has telegraphed an October window for months. Expect Anthropic’s confidential filing within the next 30-60 days if the timeline holds.

Should I avoid building on OpenAI’s API until after the IPO?

No. The API surface doesn’t change because of the listing. If anything, post-IPO predictability around pricing and product roadmap makes building on OpenAI marginally safer than the private-era unpredictability. Build for the product you’re using today. Revisit the pricing math each quarter as the public-company discipline kicks in.

Is OpenAI profitable?

Unknown publicly. The S-1 will disclose it when the filing goes public. Most credible estimates have OpenAI deeply unprofitable on a net basis (training and capacity spend) while running healthy gross margins on inference. The exact numbers will define a lot of the IPO pricing conversation.

---

Last updated: May 22, 2026. Sources: Axios · AI Weekly · SEC voluntary submission FAQs · Investing.com on the trillion-dollar IPO test.

Related reading: Anthropic Tops OpenAI’s Valuation at $900B · OpenAI’s $122B Superapp Strategy · ChatGPT Ads Manager Self-Serve Launch · Anthropic vs OpenAI in 2026 · OpenAI Ends Azure Exclusivity.