Prices as of September 3, 2026, 6:07 a.m. Eastern. Record compiled from 24 dated sources and re-checked September 3; see the sourced record file for every citation.
When will OpenAI IPO? Not this year, and probably in the first half of 2027. OpenAI's chief financial officer, Sarah Friar, told employees on August 19 that the company "will be a public company in 2027." As of September 4, 2026, Kalshi prices an IPO confirmed before June 1, 2027 at 69-77 cents, before March 1 at 34-42, and before January 1 at 11-13; 70 cents means a 70% chance. You can't buy OpenAI stock directly today; the one public door is Robinhood's RVI fund.
On August 19, Sarah Friar, OpenAI's chief financial officer, told employees at an all-hands meeting that the company "will be a public company in 2027," and sooner only if "our business continues to inflect." As of 6:07 a.m. Eastern on September 3, the people with money on the question agree with her. On Kalshi, where a price in cents is a probability, the contract that pays $1 if OpenAI's IPO is confirmed before January 1, 2027 is quoted at 13 cents by buyers and 17 by sellers: about a 15% chance it happens this year. The contract for before June 1, 2027 sits at 71 to 87 cents, about a 79% chance, and last traded at 75. The market's answer to "when is OpenAI going public" is spring 2027, and in the week since this page last read the book it has moved further from an early listing, not closer: the before-February contract has lost 12 cents.
The reason she said 2027 is in numbers that came out the day before. The Wall Street Journal reported on August 18, per SiliconANGLE, that OpenAI's second-quarter revenue rose to $6.7 billion from $5.7 billion while its operating loss widened to $12.3 billion from $9.3 billion. The same day, Anthropic's second-quarter revenue of $11.6 billion passed OpenAI's for the first time, with a $559 million adjusted operating profit. Friar's other line to the room, per The Next Web: "The IPO is not a finish line, it is a milestone, another fundraise." On Anthropic going first: "We are running our own race."
If you searched "can I buy OpenAI stock," the honest answer is not directly, and not yet. OpenAI's own rule, as quoted by the Journal in May: "All OpenAI equity is subject to transfer restrictions. This means that OpenAI equity cannot be directly or indirectly transferred unless the seller first obtains OpenAI's written consent." The one public door is Robinhood Ventures Fund I (NYSE: RVI), a closed-end fund that bought about $75 million of OpenAI common stock on April 17, 2026: a sliver of an $852 billion company inside a fund, not OpenAI shares. Until a ticker exists, the only OpenAI you can trade is the date, and what counts as an IPO on this contract is stricter than the headlines.
Stay ahead of the markets.
Daily insights and expert picks on Kalshi, Polymarket, and what's moving markets.
Free forever. Unsubscribe anytime.
The Board
As of September 3, 2026, 6:07 a.m. Eastern (Kalshi, cents = probability):
| IPO confirmed | Yes price (bid/ask) | Reads as |
|---|---|---|
| Before January 1, 2027 | 13¢ / 17¢ | about 15% |
| Before February 1, 2027 | 14¢ / 19¢ | about 17% |
| Before March 1, 2027 | 35¢ / 55¢ | about 45% on a book 20 cents wide; last trade 42 cents |
| Before April 1, 2027 | 27¢ / 56¢ | about 42% on a book 29 cents wide; last trade 48 cents |
| Before June 1, 2027 | 71¢ / 87¢ | about 79%; last trade 75 cents |
Each contract pays $1 if, before its date, the SEC declares OpenAI's registration statement effective, the IPO is priced, or an exchange assigns OpenAI a ticker. A confidential draft, a press report, a share sale to employees, or a CFO naming a year settles nothing. The 2026 rungs are all but gone: the before-September contract settled No at the end of August, before October 1 is quoted 0 to 2 cents, November 4 to 7, December 4 to 9.
The August record explains the shape. On August 20, the before-2027 contract fell from 25 cents to 20 (an intraday low of 14) on 2,331 contracts, and the single biggest trade of that week came at 10:24 a.m. Eastern on August 21: 2,532 contracts of before-February at 37 cents, with the taker on the NO side, a bet against a confirmed IPO before February 1. Most of what has changed since this page's August 26 read happened in four minutes in the middle of the night. Between 1:50 and 1:54 a.m. Eastern on August 28, one side sold the early rungs across the board: 2,737 contracts of before-2027 at 13 to 19 cents, 354 of before-February at 15 to 24, 106 of before-April at 46 to 51 and 15 of before-March at 29, every fill with the taker on NO. The before-February book, quoted 28 to 29 on August 26, reset to 14 to 19 that night and has stayed there; its last print was 18 cents on August 30. We looked for a reason and found none we can cite: no filing, no report and no company statement dated August 27 or 28 touches the IPO, and OpenAI's news on August 27 was an open letter on AI cyber threats and an office expansion in Brazil. The latest fill on the whole ladder came at 10:26 p.m. Eastern on September 2, when someone sold 128 contracts of before-December at 4 cents. The price moved; the record did not.
Two rungs need explaining, and one of them is the only rung whose book sits higher than the seller left it: February reset to 14 to 19 that night and has not moved off it, March did. The before-March contract was swept to 29 cents in those four minutes and then bought back twice on September 2, at 39 cents that morning and at 42 on 52 contracts at 8:34 p.m. Eastern, the freshest print among the 2027 rungs. It is quoted 35 to 55 now, and its midpoint of 45 sits above April's 42, which cannot be a view on a ladder where every later date must be at least as likely as the earlier one; it is two wide books disagreeing. And the before-May contract, which on a screen looks like it fell 10 cents this week, did not trade down at all: it is quoted 34 to 75 because a bid was pulled from 61 cents, its last trade was 69 cents at 12:47 a.m. Eastern on September 2, and the 156 contracts that changed hands there since August 26 all printed between 59 and 69. A wide book is a missing order, not a view, which is why that rung stays off the table above.
The before-2027 contract carries 37,433 open positions; the settled September contract left 150,036 behind. Polymarket, pulled at the same moment, had "OpenAI IPO by December 31, 2026" at 8 to 10 cents on about $953,000 of lifetime trading, down from 16.5 on August 26, a bigger move than the Kalshi before-2027 contract's, which went from 16 to 17 cents to 13 to 17 over the same week.
What the Models Think
We also put the question to our panel of forecasting programs, which we grade in public against real settlements, without showing them the price, only the sourced record below. Three of the panel's eight seats ran this board. Their middle answer, in the August 26 run, put an IPO confirmed before January 1, 2027 at 6%, well under the 17-cent market of that day and still under the 13 to 17 it is quoted now, and the widest gap between any two of them was 3 points. The reason, in plain words: the CFO's own guidance is 2027, no public filing exists on EDGAR, and the $7 billion employee share sale on August 10 relieved the pressure that usually forces a listing. One model: "firms that publicly guide to the next calendar year rarely beat it by months." All three named the same thing that would move them, a public registration statement appearing on EDGAR, and the same tail that keeps the December contract alive: a strong Anthropic debut in October pulling OpenAI's timeline forward. Worth knowing: a first pass in which the same panel could see that week's prices came back at 14%, close to the market; shown only the record, it went to 6. The market has real money behind it. The panel has only the record. These are model estimates, not predictions of fact and not financial advice.
What The CFO Actually Said, And What The Numbers Behind Her Say
The September plan died in public over ten weeks. On May 20, the Journal reported, per Decrypt, that OpenAI would file confidentially within days, with Goldman Sachs and Morgan Stanley as underwriters, "targeting a September listing." On June 8, per TechCrunch, the confidential draft went in, a week after Anthropic's, with a company statement that reads like a shrug: "We have not decided on timing yet; it may be a while because there are things we want to do that are easier as a private company." On June 25 the New York Times reported, per Yahoo Finance, that advisers had put two options in front of Sam Altman: list in 2026 below $1 trillion, or wait for 2027 and the trillion. Forbes' account of the same reporting had Altman calling any cut to the target a nonstarter. The cautionary tale was two weeks old. SpaceX listed on June 12, opened at $150, ran above $225, and was back at $151 by June 26.
Then came the $7 billion tender on August 10, in which OpenAI itself bought shares from current and former employees at the March round's $852 billion valuation. Altman's line to staff, quoted by Bloomberg: "we did not have our best 12 months ever, which is mostly my fault, but we are about to have our best 12 months to date." A company that has just written its employees a $7 billion check is not a company that needs a listing for liquidity, and TechCrunch read it exactly that way. Reuters reported on February 20 that OpenAI targets roughly $600 billion in compute spending through 2030; that is the bill an IPO has to help pay, and the reason "another fundraise" is not a throwaway phrase.
The calendar explains the shape of the board better than the headlines do. Under the SEC's rules on the age of financial statements, the numbers in a registration statement cannot be more than 135 days old when it goes effective. For a December year-end company, September 30 figures go stale around February 12, 2027, and audited full-year figures become mandatory after roughly February 14. A deal on third-quarter numbers has to be done by mid-February; a deal that misses waits for the audit, which means late March at the earliest. That is why before-February trades at 14 to 19 while before-April is quoted 27 to 56 and before-June is about four chances in five. It is also why Anthropic's ladder, same filing month and same rulebook, looks nothing like this one: a confirmed Anthropic listing before November 1 is quoted 87 to 89 cents, up from 76 to 78 on August 26, and before December 1 at 89 to 91; before October 1 is 20 to 21 cents and traded 5,516 contracts in the 24 hours before this read. Bloomberg reported on August 20, per Yahoo Finance, that Anthropic was preparing to file its IPO paperwork publicly as early as the end of the month; as of September 3, EDGAR's company index shows no registration under Anthropic's own name, only investment vehicles that hold its shares. Traders think one of them is weeks away and the other has not started.
The Lawsuit, The "Backstop," And The Structure An Underwriter Can Sell
None of these dates existed in a sellable form a year ago. On October 28, 2025, OpenAI closed the restructuring that turned its nonprofit into the OpenAI Foundation, holding about 26% of a for-profit called OpenAI Group PBC, with Microsoft at about 27%, worth roughly $135 billion at the time. Delaware Attorney General Kathy Jennings issued a statement of no objection after more than a year of review, on the condition that the foundation "will retain control and oversight over the newly formed PBC, including the sole power and authority to appoint members of the PBC Board." California's attorney general did the same.
Elon Musk spent two years trying to undo that. His 2024 suit asked a court to remove Altman and Greg Brockman, unwind the for-profit, and steer damages back to the charitable mission. On May 18, 2026, a jury in Oakland found his claims barred by the three-year statute of limitations, and Judge Yvonne Gonzalez Rogers adopted the verdict. Musk's response, per Newsweek: "There is no question to anyone following the case in detail that Altman & Brockman did in fact enrich themselves." He called the outcome a "calendar technicality" and said he would appeal to the Ninth Circuit; OpenAI's lawyer William Savitt told Bloomberg Law that "this decision is not a technical decision, it's a substantive one." Our search on September 3 found the vow but no docketed appeal in this case. Three weeks after the verdict, the confidential draft went to the SEC. That is the order the record shows.
Friar has her own entry. At the Journal's Tech Live conference in November 2025, asked whether she meant a federal backstop for the company's chip and data-center financing, she said "Exactly," describing the backstop as what "allows the financing to happen." That night, per CNBC, she wrote on LinkedIn: "OpenAI is not seeking a government backstop for our infrastructure commitments. I used the word 'backstop' and it muddied the point." Altman said the company neither had nor wanted a government backstop for its data centers. The episode matters here because the financing question she was describing, a $1.4 trillion commitment over eight years against a company that lost $12.3 billion in a quarter, is the question a public prospectus has to answer in writing, and the same executive is now the one putting a year on the answer. Neither episode shows up in the trade record this page can see. What the restructuring did do is make a registration statement writable at all, which is why this ladder can price a confirmation before June 2027 at 79 cents instead of nothing; the appeal that has not been docketed is a risk-factor paragraph in that document, not a price.
More on this: Bears vs Panthers Kalshi Picks: Carolina Lost Both Starting Tackles · Browns vs Jaguars Kalshi Picks: A Touchdown Isn't Enough · Buccaneers vs Bengals Kalshi Picks: Cincinnati Couldn't Stop Anyone · Cardinals vs Chargers Kalshi Picks: The Money Is On The Cardinals · OpenAI Millennium Prize Odds: Navier-Stokes Doesn't Count
What Moves The Price From Here
- A public registration statement on EDGAR, any day. A confidential draft is invisible; the public version must post at least 15 days before the road show, and pricing follows the road show. The day it appears, a clock of roughly four to six weeks starts, and whichever rung sits at the end of that clock heads toward $1. This is the one thing all three seats that ran this board named first, and it is free to watch.
- Anthropic's debut, which the ladder prices for October. Its before-October contract is quoted 20 to 21 cents and its before-November contract 87 to 89; Bloomberg's August 20 report had the public filing coming as early as the end of August. It has not appeared: EDGAR shows nothing under Anthropic's own name as of September 3, so the four-to-six-week clock above has not started, and every day it does not start pushes the before-October rung toward zero and hands the debut to November, which the book already prices at 87 to 89. A strong first day gives OpenAI's bankers a comparable and pulls the December contract (4 to 9 cents) up; a SpaceX-style fade pushes more of this ladder into spring.
- The February wall, February 12 to 14, 2027. Audited 2026 numbers arriving on schedule keep late March and April live; any sign of the audit dragging sends probability toward May, June, and the seven months of 2027 that sit past this ladder's last rung.
The before-2027 contract stops trading at 11:59 p.m. Eastern on December 31, 2026. The last rung on this board closes at 11:59 p.m. Eastern on May 31, 2027.
Hero illustration: OddsShopper, in the house collage style. Sam Altman photo by Steve Jurvetson, licensed CC BY 2.0; photos cropped, toned, and composited.
Prices quoted are live Kalshi market prices as of September 3, 2026, 6:07 a.m. Eastern, and will keep moving until these contracts settle; the first rung closes December 31, 2026 and the last on May 31, 2027. OddsShopper covers prediction markets as an independent analyst: any model or panel figures referenced are model estimates, not predictions of fact and not financial advice. This article does not advocate for or against any candidate or party. Trading involves risk; contracts can go to zero.
