Anthropic is now expected to start marketing its IPO in mid-October at the earliest and complete the listing days before the November midterms, according to a Reuters exclusive published Friday evening, September 4. Kalshi's date ladder on the deal reset around that calendar within the hour, and then kept moving. The contract that pays on an IPO confirmed before October 1 collapsed to a few cents on the wire and stayed there. The contract that pays before November 1 closed near 84 cents after a volatile Friday night, then was sold down to 80 cents on Saturday with no new report behind it. Halloween is still the line the whole market trades around, and as of September 5, 2026 the market is charging more for the chance the deal misses it.
The Quick Answer
The market says the Anthropic IPO gets confirmed in October, most likely in its final week, and it is a little less sure of that than it was on Friday night. Kalshi prices a confirmation before November 1 at 80 cents as of Saturday afternoon, down from 84, which is the market putting one chance in five on the Reuters calendar slipping past Halloween or on Anthropic choosing to wait out the election. Our five price-blind AI models, frozen since mid-August, had the same rung at 37%. They were right that the September sprint would fail, and they still sit well below the market on October.
New to Kalshi event contracts? Start here with how they work. Every price below carries its stamp, about 4:15 p.m. ET on Saturday, September 5, 2026, and refreshes on our standard market-pulse cycle while this story is moving. Free: The PM Weekly Brief covers this ladder and the rest of the week's prediction-market moves in one email.
What Actually Settles This Market
| Venue | Kalshi |
| The Question | "When will Anthropic officially announce an IPO?" — a ladder of yes/no contracts, one per deadline. The July, August and September rungs have settled NO; three new October rungs (before the 10th, 17th and 24th) listed Friday, September 4 |
| What Settles It | A rung pays YES if, before its date, the SEC declares Anthropic's Form S-1 effective, OR the IPO is priced, OR a securities exchange assigns Anthropic a ticker (per Kalshi's settlement rules) |
| Event Ticker | KXIPOANTHROPIC-DATE (series KXIPOANTHROPIC; each rung trades as KXIPOANTHROPIC-DATE-26OCT01 and so on) |
| Prices Below As Of | About 4:15 p.m. ET, Saturday, September 5, 2026 |
The title says "announce." The rules say something much stricter. All three settlement triggers sit at the end of the going-public process, and Anthropic has only taken the first step: on June 1, 2026 the company publicly confirmed it had confidentially submitted a draft Form S-1 to the SEC, with "the number of shares to be offered and the price" not yet set. A confidential draft is none of the three triggers, so it settles no rung of this ladder.
As of September 5 the S-1 is still confidential, a full-text search of SEC EDGAR for Anthropic filings since June 1 returns no S-1 or amendment, and no date, price range, ticker, exchange or share count has been set. That is why the before-July, before-August and before-September rungs settled NO in a row, and it is the entire explanation for Friday evening. The same map for the other AI lab is on our OpenAI IPO ladder, where the by-January rung was quoted at 11 bid, 16 offered at the same Saturday-afternoon stamp, even though OpenAI filed its own confidential draft on June 8, 2026.
How The Calendar Slipped
Four reports in a month, each later than the last. On August 13 the Financial Times, as relayed by Fortune, said Anthropic could launch as soon as October at a valuation approaching two trillion dollars, with the IPO "still under discussion." On August 20 Bloomberg reported the company was preparing to publicly file as early as the end of that month, which sent the before-October contract into a weekend spike. On August 27 The Information, as relayed by Reuters, moved the prospectus to after Labor Day and the listing to late September or early October, adding that existing shareholders might sell in the deal and that the lockup could run longer than customary.
Then the calendar moved again, this time with a mechanism attached. On September 3 Bloomberg reported that Anthropic was finalizing a $15 billion revolving credit facility, led by Morgan Stanley with Goldman Sachs, JPMorgan and Citigroup in prominent roles, the same four banks leading the IPO. On September 4 Reuters' Echo Wang reported, citing people familiar with the matter, that marketing now begins in mid-October at the earliest, the listing lands days before the midterms, and the public prospectus, once expected the following week, is not expected until late September. Reuters ties the two stories together: the credit facility closes first, bank analysts then meet the company, and companies typically leave a few weeks between those meetings and a public prospectus. The plans are subject to change, and Anthropic declined to comment.
The public filing went from "end of August" to "after Labor Day" to "late September." The listing went from "as soon as October" to "days before the midterms." Every step pushed the first settlement trigger later, and the market's job on Friday was to price how much later.
Run the Reuters sequence as a clock and count backward. The midterms are Tuesday, November 3. A listing "days before" that is October 28 to 30, and an IPO is priced the evening before it lists, so pricing lands October 27 to 29. Marketing "in mid-October at the earliest" means the roadshow opens around October 13 to 15, which gives it the usual two weeks before pricing. Behind that sits the public prospectus in late September, the analyst meetings a few weeks before it, and the credit facility closing first. Every link in that chain is at its earliest date, and the last link still lands only two to four days before the November 1 contract stops paying. That arithmetic, not the headline, is what the rest of this page is about. Saturday brought no fifth report, only the Reuters story echoing through the weekend press, and the ladder moved anyway. That is the part worth reading closely.
What Friday's Tape Said
The before-October contract had been bid up all afternoon, on nothing in the day's reporting, to a 50-cent print, most likely by traders paying for The Information's after-Labor-Day calendar, which still left a narrow lane to a priced, tickered offering by the end of September. When the Reuters wire hit that evening, the contract was sold from 30 cents to a penny inside an hour. It finished the session at 3 bid, 4 offered.
Three things about the rest of the ladder matter for the read. The before-November contract took two sell sweeps on the wire, one at 6:42 p.m. ET from 90 to 85 and a second three minutes later that printed as low as 61 on a few dozen contracts, and buyers had it back at 90 by 7:07 p.m. before the 8 p.m. hour sold 790 contracts at 83 and 84. The before-December contract rose after the report, closing six cents above Thursday, which is the tell about how the market read it. And the last flow on the crashed October rung was buying: somebody spent the late evening collecting a 3-cent lottery ticket on a trigger arriving before the prospectus is even public.
Three new rungs inside October (before the 10th, 17th and 24th) listed at 5 p.m. Friday with almost no book. Their quotes were placeholders on Friday night, and one of them stopped being a placeholder on Saturday afternoon.
What Moved On Saturday
Friday night's version of this page said the midterm math had not actually been repriced. On Saturday, somebody repriced it, four cents' worth, and on almost no volume.
The before-November contract traded 1,486 contracts between Friday's 10:20 p.m. stamp and Saturday's 4:15 p.m. stamp, against 14,454 during Friday's session. Nearly all of it was selling. It had printed 85 just after 11 p.m. In the first hour after midnight it was sold to 81, to 79 by 2 a.m., and printed 74 just after 3 a.m. on a scrap of a trade, all against resting bids. A 300-contract sale at 10:43 a.m. printed 67, the low of the day, and 514 contracts changed hands in that hour at 67 to 76. Buyers then paid 78 to 84 through the 1 p.m. hour, and the last real print was 222 contracts sold at 80 at 2:46 p.m. The book closed the afternoon at 80 bid, 81 offered. The whole move is a little over a thousand contracts of net selling into a thin Saturday book, which is not a reprice by a whale; it is the absence of the buyer who was there at 85 on Friday.
The before-October contract did the opposite and did it in size: 33,354 contracts since the Friday stamp, 65,000 over the trailing 24 hours, almost all of it inside a 3-to-6-cent range, and most of it buyers. The 10 p.m. hour on Friday alone put 13,322 contracts through at 4 to 6 cents with every one bought, the largest a 4,974-contract lot at 5 cents. Someone is treating a trigger before September 30 as worth a nickel, and the market has been happy to sell it to them at that price all weekend. It sits at 4 bid, 5 offered.
The one October rung that found a book was before-October-24. At 3:09 p.m. ET Saturday, 1,659 contracts were bought in one burst at 51 to 65 cents, the first sizable burst in the fast version of the Reuters calendar: pricing by October 23, from a roadshow that starts in mid-October. Its quote is still 15 bid, 62 offered, wide enough that the number is a trade, not a price. The before-October-10 and before-October-17 rungs still have a few hundred contracts each and stay off the board.
Further out, before-December eased to 93 bid, 95 offered after touching 96 overnight, and before-January firmed to 95 bid, 96 offered on a single contract. The shape of Saturday's move is the November rung giving back four to five cents while the rungs on either side of it barely moved, which is the market shifting a sliver of probability from "October" to "November or later" rather than re-reading the whole calendar.
Why Halloween Is The Line
The before-November contract closes at 12:59 a.m. ET on November 1, so it pays only on a trigger that lands on October 31 or earlier. Halloween is the line.
Now put the clock from the calendar section against it: pricing on October 27 to 29, a debut on October 28 to 30, and a contract that pays through October 31. Every part of that calendar sits inside the contract, but the last part sits inside it by a business day or two. That is what 80 cents buys: the Reuters timeline holding with a business day or two of slack, priced with one chance in five that the slack is not enough. On Friday night the same rung asked 84 for the same calendar; the four cents the market gave back on Saturday is the price of "at the earliest."
Read as a distribution, subtracting each rung from the next, the ladder now puts about 76 points on a trigger firing in October, about 14 on November and roughly 6 on later or never. Friday night's read was about 80, 11 and 5. The gap between the November and December rungs is the price of a slip of even a few days, or of a decision to wait for the election to pass, and it got three points wider over the weekend. Two weeks ago this page had the October mass split between September and October; September settled at zero and its mass moved to one month rather than scattering. Polymarket's by-date ladder, as tracked by CryptoSlate, reads the same October calendar on that venue's own resolution terms.
The new October rungs say the same thing from the other side. A trigger before October 24 needs one of the three settlement triggers by October 23, most naturally a pricing, which means barely a week of marketing from a mid-October start, and Saturday's 1,659-contract burst at 51 to 65 cents was the first sizable buyer to call that fast version closer to a coin flip than a long shot.
The Board
Every price is a cent value on a one-dollar contract, so 80 cents means the market implies roughly an 80% chance. Market prices are the midpoint of each rung's bid and ask as of about 4:15 p.m. ET Saturday, September 5, rounded toward the bid where the midpoint lands on a half cent. The hero chart at the top of this page still carries Friday night's prices; the table is Saturday's.
The "AI blend" is the equal-weight average of the five models from our August 16 panel run, after a revision round; the panel ran before every report above and has not been re-run, so its numbers price an older information set. "Edge" is the blend minus the market, in points. The before-September rung settled NO on September 1 with the panel at 2%, scored as a hit. The three new October rungs are left off until they have a two-sided book.
| Confirmed Before | Market | Fable | Opus | Sonnet | GLM | DeepSeek | AI blend | Edge |
|---|---|---|---|---|---|---|---|---|
| Sep 1, 2026 | Settled NO | 2 | 2 | 2 | 2 | 2 | 2% | hit |
| Oct 1, 2026 | 4¢ | 10 | 10 | 10 | 10 | 15 | 11% | +7 |
| Nov 1, 2026 | 80¢ | 37 | 34 | 35 | 32 | 45 | 37% | −43 |
| Dec 1, 2026 | 94¢ | 55 | 53 | 55 | 50 | 65 | 56% | −38 |
| Jan 1, 2027 | 95¢ | 66 | 64 | 66 | 60 | 75 | 66% | −29 |
| Feb 1, 2027 | 96¢ | 72 | 72 | 73 | 67 | 82 | 73% | −23 |
| Mar 1, 2027 | 97¢ | 77 | 78 | 78 | 72 | 87 | 78% | −19 |
| Apr 1, 2027 | 96¢ | 81 | 82 | 82 | 76 | 90 | 82% | −14 |
Every seat on this panel is graded against real market settlements — records to date: Fable 87% on 1,549 graded calls · Opus 87% on 1,621 · Sonnet 85% on 1,597 · GLM 82% on 2,941 · DeepSeek 81% on 2,985. Recomputed daily; the full scoreboard is public.
The October and November rungs quote a cent wide. The December rung is two cents wide and the 2027 marks past January come off thin books with five- and six-cent spreads, which is how March can quote above April; a cumulative ladder that quotes out of order at the far end is telling you those rungs carry no information. The liquid front of the curve is the signal, and the 2027 numbers are approximations.
The most interesting row is the November one, and for a new reason. On Friday night the gap between the market and the panel on that rung was 47 points. On Saturday the market closed four of them on its own, on no news, on fewer than 15 hundred contracts. That is not the panel being vindicated; a 43-point gap is still enormous, and the panel's 37 was built on a timeline that could slip into 2027. It is the market admitting that "mid-October at the earliest" carries an "at the earliest" in it. The October row still cuts the other way: the panel's 11 sits above a 4-cent market for a rung that needs the SEC, a pricing or a ticker before the end of September while Reuters says the prospectus is not even public until then.
Five of the panel's eight seats priced this board on August 16; the other three were unavailable for that run. The panel re-scores when the story's next hard fact lands, a public S-1 on EDGAR, and every number here gets graded in public once the market settles, on the full graded scoreboard.
More live boards from the same panel: OpenAI IPO ladder (its by-January rung quoted 11/16¢ at the same 4:15 p.m. ET stamp, against Anthropic's 95¢) · GTA 6 release-date board. Kalshi prices fetched September 5, 2026.
Where The Panel Stood, And Why It Reads Differently Now
When the panel ran on August 16 its distribution had the same peak as the market's but a much flatter shape: October the single most likely month at about 26 points, real mass in November and December, and roughly 18 points on a slip deep into 2027 or a shelved deal. Its one-line thesis was that record-breaking IPOs slip, and this would be the largest ever.
Opus came into round one pricing before-November at 43, then cut it to 34 after reading the more skeptical seats:
The reported October pricing target concentrates trigger-firing mass into October and November, but the S-1 is still confidential and reports are not filings. — Opus
GLM stayed the most conservative seat: for an unprecedented two-trillion-dollar deal the SEC review runs long, roughly four to seven months from confidential filing to effectiveness, so a trigger before November is compressed. DeepSeek, the most aggressive seat, kept the fullest late tail while trimming its before-November mark.
Three weeks later the scorecard cuts both ways. On September the panel was simply right, and "reports are not filings" is the logic that dragged before-October from a 50-cent Friday print to 4, which makes the panel's 11 on that rung look generous rather than slow. On the rungs that matter from here, the market has moved away from the panel, and Saturday moved it back by a hair. The panel priced slippage, and slippage is what the four reports delivered, but each report moved the date by weeks rather than months and came with more specifics than the last: named banks, a credit facility, a marketing window, an election deadline. The panel priced a timeline that could slip into 2027. The market is pricing one that has slipped into late October and has an election to stop it slipping further, and on Saturday it started charging a little for the possibility that the election stops the deal instead. Both can be true, which is why the run's numbers appear with their date on them instead of being nudged toward the news.
What Moves It Next
- The Credit Facility Closes, Then The S-1 Goes Public. Per Bloomberg and Reuters the sequence is facility, analyst meetings, then a public prospectus in late September. A public filing settles nothing on its own, because none of the three triggers has fired, but it starts the market's visible clock toward effectiveness, and it usually names the exchange and the proposed ticker symbol before the exchange formally assigns one. That is why the panel's read was that it pulls the near rungs sharply upward: it converts a well-sourced timeline into a document with dates in it. If late September passes with EDGAR still empty, the before-November contract is the one that bleeds, the way before-October did on Friday, and Saturday showed how little selling it takes to move it.
- Marketing Starts In mid-October. On time, it compresses the remaining path to a pre-Halloween pricing and makes Saturday's before-October-24 buyer look early rather than wrong. A start that slips to the third or fourth week of October puts the November rung on a knife's edge and lights up the December rung instead.
- A Ticker Gets Assigned, Or The Deal Prices. Those are two of the three settlement triggers themselves. A price range in the prospectus is a signal, not a trigger; the rung only pays when the IPO is actually priced, the S-1 is declared effective, or an exchange assigns the ticker. An early ticker turns the nickel stack on the October rung into dollars.
- A Weak Window, Or A Decision To Wait Out The Election. A company that raised 65 billion dollars in May and is lining up a fifteen-billion-dollar credit facility does not need the proceeds, so a poor IPO tape or a choice to let the midterms pass pushes mass across the November line. Saturday's four cents is the first of that being bid.
Until one of those fires, the confidential draft remains the only hard fact on the board.
The Bottom Line
The next checkable step is a database entry, a public S-1 on EDGAR in late September. If it lands on schedule, the before-November contract gets its Friday-night price back and the December contract becomes the floor. If late September passes with EDGAR empty, the 80 on the November rung does what the 50 did on October's, and Saturday's thin selling is what the first hour of that looks like. Friday was the third time in a month this market showed it knows the difference between a hard fact and a well-sourced timeline; Saturday was the first time it priced the difference between "mid-October" and "mid-October at the earliest."
Prices on this page are Kalshi's book. If you also trade on the Polymarket App, code OS4 gets new users a $50 trading bonus on a $10 deposit. Affiliate link; we may earn a commission.
18+ Only. Restrictions and eligibility requirements apply. Not available in all jurisdictions. Trading is risky. 100% loss can occur. See polymarket.com/tos for more information. The Polymarket US App serves as an independent software provider and affiliate of Polymarket US and Polymarket Clearing, the CFTC-regulated exchange and clearing organization.
More on this: OpenAI IPO Odds: The Other AI Lab's Kalshi Ladder · How Kalshi Event Contracts Work · The Full Graded Model Scoreboard · OddsShopper Prediction Markets Hub
