OpenAI is preparing to file a confidential S-1 with the SEC as early as May 22, with Goldman Sachs and Morgan Stanley advising. The filing itself does not list shares or set a price. It deposits audited financials with regulators and starts a clock that ends with a public offering targeted between Labor Day and Thanksgiving 2026, with September at the earliest and October, November most likely. The number that matters is not on the S-1 yet. It is the gross margin, the customer concentration table, and the compute-cost line that every other AI lab will have to defend against in their next funding round.
What the Confidential S-1 Actually Does
A confidential S-1, filed under the SEC’s 2017 expansion that allows all issuers to submit draft registration statements privately, lets OpenAI submit its draft for non-public SEC review. The public does not see the filing until shortly before the roadshow. What the SEC sees, and what eventually becomes public, includes audited revenue, cost of revenue, customer concentration (what percentage of revenue comes from the top few accounts), debt obligations, and risk factors.
OpenAI’s last private round closed March 31, 2026 at $852B post-money on a reported $122B raise. The reported IPO target exceeds $1T. On ~$25B ARR, that is roughly 40x revenue. For context, no public SaaS company at comparable revenue trades anywhere near that multiple. The S-1 will reveal whether that premium is justified by growth velocity, gross margin, or something else entirely.
OpenAI has also hired Cynthia Gaylor, former CFO of DocuSign, as its first head of investor relations. CFO Sarah Friar told CNBC it is “good hygiene” for a company of OpenAI’s size to “look and feel and act” like a public company. That language is standard pre-IPO signaling. The hiring of a dedicated investor-relations chief is not.
The Numbers That Will Move Every AI Valuation
The headline figures from multiple sources: ~$25B ARR ($2B/month), 900M+ weekly active users, 50M+ subscribers, enterprise at 40%+ of revenue, and estimated 2026 losses of ~$14B. Cash-flow positive is not projected until 2030.
The gross margin line is the one every competing lab should care about. OpenAI’s cost of revenue is dominated by inference compute. If the S-1 reveals a gross margin of, say, 60%, that sets a ceiling for what private-market investors will accept from Anthropic or xAI on the same metric. If the margin is 40%, the opposite: it signals that frontier-model inference is structurally expensive and every lab burning compute to compete is doing so at a loss the market already prices in.
Customer concentration is the other landmine. If a single partner (read: Microsoft) accounts for a large fraction of revenue, that risk factor alone could compress the multiple. Microsoft holds 27% of OpenAI’s equity. The S-1 will have to disclose the revenue relationship separately from the equity stake, and that disclosure will be read carefully by anyone pricing Anthropic’s cloud-partner revenue.
The Anthropic Repricing Problem
Anthropic raised $30B in Series G in February 2026 at $380B post-money. Sacra estimates Anthropic hit $43B annualized revenue by April 2026, up from $9B at end of 2025. Anthropic is in early talks to raise at a pre-money valuation exceeding $900B as of May 2026.
Here is the structural problem: Anthropic’s revenue figures include gross-reported cloud reseller revenue from AWS, Google Cloud, and Microsoft. When a customer buys Claude through AWS Marketplace, Anthropic recognizes the full contract value as revenue, then pays the cloud provider their take. This inflates the top line relative to a net-reporting convention. Direct comparisons to OpenAI’s revenue require an adjustment that neither company has publicly provided.
Anthropic is also projecting its first operating profit of $559M in Q2 2026 on $10.9B quarterly revenue, while OpenAI remains deeply unprofitable. If Anthropic can demonstrate profitability before OpenAI’s S-1 goes public, the narrative writes itself: the second-largest lab is already making money while the market leader burns cash. Filing first gives OpenAI the chance to set the valuation benchmark before that profitability story hardens into the default comparison.
Meanwhile, Claude Code annualized revenue reached $2.5B, with business subscriptions quadrupling since the start of 2026. Enterprise represents over half of Claude Code revenue. One analysis estimated 4% of all public GitHub commits worldwide were authored by Claude Code. That is a specific, measurable penetration metric in a way that “900M weekly active users” is not, and public-market investors will notice the difference.
Governance Landmines
OpenAI’s corporate structure is, by public-company standards, unusual. The OpenAI Foundation holds 26% equity and appoints, and can remove, all PBC board members; Microsoft holds 27%; employees and other investors hold the remaining 47%. A nonprofit controlling board composition at a company targeting a $1T+ market cap will generate SEC comment letters. The risk-factor section of the S-1 will have to address this directly, and proxy advisory firms will have opinions.
On May 18, a federal jury rejected Elon Musk’s lawsuit against OpenAI and Sam Altman in under two hours, finding Musk waited too long to sue. The verdict removes an overhang for the IPO timeline, but the governance question lingers in the news cycle during the roadshow window. That is not a legal risk to the listing. It is a narrative risk to the premium.
The 2026 IPO Queue
OpenAI is not the only company queuing for public capital. SpaceX confidentially filed with the SEC in April 2026. Both companies share Goldman Sachs, Morgan Stanley, and JPMorgan as bankers.
The bank-capacity constraint is real. The same underwriting teams cannot run two IPOs of this size simultaneously. Timing matters for more than narrative positioning: it determines which deal gets the best analyst coverage and which gets the leftovers.
What Changes When AI Labs Have Quarterly Earnings
The shift from private rounds to public markets changes the valuation mechanics for the entire sector. Today, Anthropic and xAI raise money by telling a story about trajectory. The investors who buy in are sovereign funds, large venture firms, and strategic partners who accept that audited financials are not available and that revenue multiples are calculated from self-reported or third-party-estimated figures.
Once OpenAI’s S-1 is public, every one of those investors can compare the next Anthropic pitch deck to audited OpenAI numbers. What is the gross margin? What is the revenue per employee? What is the customer concentration? What is the compute cost as a percentage of revenue? These become specific, comparable benchmarks rather than abstract narratives.
The burden of justification shifts from the buyer (who previously had to decide whether to trust a private company’s numbers) to the seller (who now has to explain why their numbers look different from the public comparable). That is a structural change in how AI valuations get set, and it happens the moment the S-1 drops, not the moment the stock starts trading.
OpenAI’s reported target exceeds $1T. Whether the market agrees depends on those audited margins. Whether Anthropic can defend $900B depends on the same numbers. The S-1 is the first domino. Everything after it is a repricing event.
Frequently Asked Questions
What happens to the Musk lawsuit after the jury verdict?
Musk vowed to appeal the decision to the Ninth Circuit. The two-hour jury verdict removes the immediate overhang, but an appellate filing keeps the governance dispute active through the IPO roadshow window. If the Ninth Circuit agrees to hear the case, OpenAI’s risk-factor section will need to address ongoing litigation from a co-founder who alleges the company abandoned its founding mission, a narrative that complicates the premium narrative regardless of legal merit.
How does SpaceX’s IPO compete with OpenAI for bank capacity?
SpaceX filed its confidential S-1 on April 1, 2026, seeking roughly $75B at an estimated $1.75T valuation. Axios reported OpenAI’s filing plan ‘seems timed to take some shine off’ SpaceX’s offering. The practical constraint: Goldman Sachs, Morgan Stanley, and JPMorgan are underwriting both deals, and the same equity sales teams and research analysts cannot run two offerings of that size in parallel. The deal that prices second gets the leftover allocation of institutional attention.
Has a nonprofit-controlled public company ever listed near this valuation?
No. The closest public-company analogues are dual-class structures at Meta and Alphabet, which concentrate voting control with founders. OpenAI’s structure is different: a separate philanthropic entity (the OpenAI Foundation) appoints and can remove every board member, with no shareholder recourse. Proxy advisory firms ISS and Glass Lewis have historically recommended voting against governance structures that limit shareholder influence, and their guidance on this one will likely draw institutional scrutiny that most mega-IPOs avoid.
What happens to xAI’s next funding round if OpenAI prices below the $1T target?
The ‘AI lab premium’ that justified private valuations above $400B compresses immediately if the public market rejects OpenAI’s asking price. xAI secondary buyers would benchmark against the actual trading multiple rather than the private-round narrative, and the repricing would be mechanical: if OpenAI trades at 25x revenue instead of 40x, every competing lab’s implied multiple gets recalculated downward using the same discount. xAI’s exposure depends on the timing and terms of its next secondary, which have not been reported in detail.