Prediction-Market Pricing of the 2026 Mega-IPOs - Blog | Dune
Prediction-Market Pricing of the 2026 Mega-IPOs
What Kalshi and Polymarket are pricing for the SpaceX, OpenAI, and Anthropic IPOs, and where the two venues disagree, read through Dune's prediction-markets dataset.
Research • June 25, 2026 • 4 min read
Filippo Armani Data Content Creator at Dune
Kalshi prices the announcement, Polymarket the listing
Kalshi and Polymarket both run "IPO by date" markets on all three names, and they settle on different events. Kalshi resolves when the company confirms an IPO: the SEC declares the S-1 effective, the deal prices, or an exchange assigns a ticker. Polymarket resolves when shares actually open for trading. In a normal IPO these two events are one to two business days apart, so the two venues should track closely. Where they pull apart, it reflects the two crowds pricing the same dated question differently, often because the cutoff falls near the expected listing. The dates align cleanly: Kalshi's "before the 1st of a month" is the same cutoff as Polymarket's "by the last day of the month before."
SpaceX priced its IPO months before the filing
SpaceX listed in mid-June and both venues' June contracts have stopped trading, so this is the one outcome we can grade.
By April 1 both venues already priced a SpaceX IPO by June 30 at about 72 to 74 percent, ahead of any public filing, and the confidential S-1 that day moved nothing.
Dune
Line chart
@dune with Dune
Two later events did the work, and both venues reacted on the same day each time:
- May 15, the accelerated-timeline report (roadshow, pricing, listing window, ticker SPCX): both jumped from the high 70s to above 90 percent.
- May 20, the public S-1 with the confirmed ticker: both moved to about 97 percent.
The two venues tracked each other almost exactly
They crossed 90 percent on the same day and held within a point or two of each other all the way up, then resolved together when shares opened on June 12. Kalshi reached 99 percent on June 4, two days before Polymarket on June 6, a small gap that closed at the listing.
Late volume rotated to the venue still pricing an open question: over the market's life Kalshi traded more, $2.66M against $2.17M taker-side, but in the final two weeks nearly all flow was on Polymarket, about $420k against roughly $1k, as the announcement settled and only the listing remained live.
Dune
Column chart
@dune with Dune
OpenAI is drifting toward 2027
OpenAI has repriced lower across every window since early June. Its chance of an IPO by September 30 fell from about 59 percent (Kalshi) and 39 percent (Polymarket) on June 2 to 24 and 15 percent now.
Dune
Line chart
@dune with Dune
The December 31 window dropped from 86/71 to 64/53
Dune
Line chart
@dune with Dune
Both venues now put a 2026 listing below even odds for most of the year and only modestly likely by December, and the two sit close together, so they broadly agree on the timing. The move lines up with the May 20 reports of a confidential filing around May 22, after which the timing odds drifted out rather than in.
Anthropic is now priced to list before OpenAI
Anthropic prices ahead of OpenAI on a like-for-like basis. Comparing the two on Polymarket alone, where both are measured by first trade, Anthropic is higher at every cutoff: 21 percent versus 15 percent by September 30, and 74 versus 53 by December 31. The market sees Anthropic reaching public trading sooner, despite filing its confidential S-1 two weeks later, on June 1.
Anthropic also shows the widest split between its two venues of any name here. Kalshi puts a 60 percent chance it confirms an IPO by September 30; Polymarket puts a 21 percent chance it is trading by then. Since confirmation and trading are normally days apart, that 39-point gap is not a long lag. It reflects uncertainty about whether the listing lands before or after September 30, concentrated at that one cutoff, with confirmation expected on the early side of it.
Dune
Line chart
@dune with Dune
The gap narrows to 15 points by December 31, at 89 and 74, as both events move inside the window.
Dune
Line chart
@dune with Dune
OpenAI draws the most live volume; Anthropic's is catching up
Lifetime taker-side notional:
- SpaceX: about $4.8M, now settled.
- OpenAI: about $1.24M, roughly 70 percent on Polymarket, the most active live market.
- Anthropic: about $0.47M, split $271k Kalshi and $197k Polymarket, with Polymarket ahead over the last two weeks.
Anthropic's Polymarket book grew from $12k to $197k in three weeks, and its price correction tracked that growth. SpaceX traded almost $5M over its life but went quiet the moment it listed. Anthropic's Polymarket book filled in only after it filed, and now leads Kalshi on recent flow.
Dune
Column chart
@dune with Dune
What to take from it
- Kalshi and Polymarket look like the same event-contract marketplaces, but their markets diverge most of the time. These three IPOs, alongside Fed rate decisions, are the closest cross-venue comparables we could find in the Economic and Finance categories.
- The markets put Anthropic public before OpenAI, despite Anthropic filing two weeks later. On a like-for-like basis, where both are measured by first trade on Polymarket, Anthropic is higher at every date (21 versus 15 percent by September 30, 74 versus 53 by December 31). The reading rests on a few thinly traded contracts, but the ordering has held for weeks.
- Even at low notional, these contracts give an independent, real-time read on deals that stay private until the S-1. Pricing comes from third parties rather than the issuer or its underwriters, so it reflects external positioning rather than sell-side guidance, a sentiment signal not otherwise available on pre-IPO names.
All charts and figures above are built on Dune's prediction-markets dataset, which standardizes Kalshi and Polymarket trades, prices, and resolutions into one schema.
Methodology notes: Figures as of June 24, 2026. The two contracts settle on events usually one to two business days apart, so a spread between venues measures disagreement at a given date, not a filing-to-listing lag. Each venue's curve gives a median date only for its own event. The two venues have separate users with no arbitrage between them, so a spread can persist.