AI Market Intelligence Daily

1. Sam Altman's "ill-advised" remark on a 2026 IPO is now fully priced, and the real action is in the valuation ladder, not the date. The headline contract "Will OpenAI not IPO by December 31, 2026?" slid back from 97% to 93%, and the mirror "Will OpenAI IPO by December 31 2026?" ticked down to 6%, a modest drift, but the more telling move is in the closing-cap contracts on Polymarket. The band for a $1.25T-to-$1.5T close rose from 8% to 12%, the $1T-to-$1.25T band from 3% to 5%, and the "less than $500B" contract finally printed a real 1%. The center of gravity for an eventual OpenAI listing is migrating upward, toward $1.25T and above, even as the probability of a listing this year stays low. Traders are separating the two questions: not this year, but when it happens, price it near a trillion.

2. The biggest percentage moves of the day are weekly-reset leaderboard contracts settling as the week closes, and they carry less signal than their size suggests. "Top Coding AI this week?" for ChatGPT on Kalshi jumped from 12% to 98%, the largest single move tracked, while "Most popular AI this week?" for OpenAI rose from 14% to 99% as the Tencent leg collapsed from 90% to 3%. These are end-of-week settlements, not new information, and they should be discounted accordingly. The durable signal underneath is that Claude remains the consensus best model overall on the December 31 contract at 69%, even as ChatGPT sweeps the week of coding and usage. One genuine divergence stands out: "Will Meta have a top-ranked AI model before 2027?" rose from 6% to 14%, a real step up that tracks directly with the news that Meta's Muse agent is now the number 2 app in the US.

3. The GPU pricing ladder is repricing sharply higher, and this time it looks like fresh money, not the previous day's churn. The Kalshi contract asking whether B200 spot price would be above $7 at 4 PM ET on Sep 18 jumped from 23% to 56%, the above-$7.50 contract from 6% to 21%, and the above-$0.70 RTX 5090 contract from 13% to 55%. That is a coordinated bet that hourly compute prices spike over the next two weeks, consistent with Jensen Huang's claim that Nvidia will grow 70% next year and with OpenAI pausing Pro subscriptions to free capacity for Astra demand. The open question is whether Sep 18 confirms a genuine supply crunch or just a spot-price wobble.

4. Anthropic is consolidating its position as the firm most likely to IPO first, while OpenAI's math frontier is getting dinged. "Will OpenAI or Anthropic IPO first?" now prices Anthropic at 97% to 8% on Kalshi, and "Anthropic IPO before 2027?" sits at 84%. The news backdrop supports the split: Anthropic's CEO outlined a plan to pace the frontier and the company disclosed distillation campaigns by Alibaba, Moonshot AI, and DeepSeek, a story of controlled scale, while OpenAI is publicly deferring its listing and fighting a public feud with mathematicians. That feud shows up in the prices: "Will OpenAI announce another Millennium Prize solution before Jan 1, 2027?" fell from 79% to 47% in a day, the largest conviction drop among the math contracts, suggesting traders are discounting the lab's math pipeline as its reputation with the mathematical community sours.

5. The AGI complex softened across the board, a quiet but consistent drift worth tracking. "OpenAI announces it has achieved AGI before 2027?" fell from 28% to 21%, and Kalshi's "Will OpenAI announce the creation of AGI? (Before 2027)" fell from 26% to 18%, with the broader "any AI model reaches 1520" contract easing from 68% to 63%. None of the headline news is AGI-specific, so this reads as a slow fade in the 2027 AGI odds rather than an event-driven move, but it points in the same direction as the IPO repricing: traders are pulling back on both the date and the magnitude of what happens by year-end. The one counterpoint is "AI wins IMO gold medal in 2026?" holding steady at 12%, which keeps a floor under the math-story complex even as the Millennium contracts fall.