AI Market Intelligence Daily

1. The B200 compute-price market on Kalshi is the day's standout move: the contract asking whether B200 per-hour price will be above $10.97 on December 31 jumped from 3% to 84% in a single ~7-hour window, the largest. change of the 88 markets that moved. The surrounding rungs tell the story of a repricing, not a coin flip: the $10.37–$10.67 thresholds now sit at 81–82%, the $10.97 level at 84%, and the $13.37 rung at just 25%, implying the market now expects December GPU spot pricing to land somewhere in the mid-$11s per hour. That is a big swing from a market that had essentially priced in B200s getting cheaper by year-end — consistent with the week's backdrop of hyperscaler capacity strain (Amazon's data-center push) and steady demand for frontier-model training runs. Google is the clear winner in this week's model-share markets, and the moves are near-uniform: every Kalshi contract above the low-20s percent threshold has converged to 97–99% within the last 24 hours, with the sharpest reprices on the 22.5% and 21.4% rungs (up 2,400% and 1,840% respectively). At the same time, Gemini 4 is now an 83% probability of shipping before November 1 (58% for the tighter October 16 deadline), and the "next Gemini Pro before November 16" market more than tripled from 10% to 32%. The pattern reads as traders reacting to an imminent or partially-announced Gemini launch: weekly share data and release-timing odds are moving in the same direction, which is the cleanest news-to-price alignment in today's set. The share markets also show what Google's week cost the rest of the field. OpenAI's "market share above 18.1%" contract fell 49% (57% to 29%), Anthropic's above-2.5% contract collapsed 82% (17% to 3%), and DeepSeek's 22% rung only rose modestly to 18% while its higher rungs (23.7%–24.2%) all sit at 1%. The "best AI" end-of-year board tells the same story: Gemini at 38% and Claude at 52% now carry the board, while ChatGPT (8%, down from 10%) and Grok (3%, down 36%) have been sold off. Notably, this reshuffling has no obvious single headline behind it — the top news stories cover Amazon's NDA policy, an OpenAI safety resignation, and Disrupt conference items — so the share repricing is most plausibly driven by the underlying weekly usage data traders see rather than by today's press. OpenAI's equity-market odds drifted lower across the board even without a triggering headline. "OpenAI IPO before 2027" slipped from 3% to 2%, "OpenAI not IPO by December 31, 2026" is now a firm 96%, and the "highest 2026 IPO market cap" question for OpenAI sits at 1%. Anthropic is the de facto IPO favorite at 82% before 2027, with 57% odds of posting the largest 2026 IPO market cap. The drift is small in absolute terms but directionally consistent with the day's OpenAI news — a safety employee publicly resigning over culture — plus the broader "only 2% of consumers are buying superintelligence" sentiment piece; the market is quietly widening the gap between the two companies' near-term listing expectations. A few divergences are worth flagging. The Grok story — reports that Grok "encouraged Trump to capture Venezuela's president" — should read as a reputational negative for xAI, yet "xAI has a top-ranked model before 2027" actually doubled from 7% to 14%; the move is small in absolute terms, but its direction runs against the headline. Separately, Apple's tightening of macOS Full Disk Access "due to new risks from AI agents" and the White House's "super intelligence" rebrand both point toward more regulation, yet "independent AI safety audits become U.S. law before 2027" collapsed 89% (9% to 1%) in the same 7-hour window as the B200 spike, and "AI regulation by 2027" sits at just 9% — the market is treating today's governance news as noise, not signal.