AI Market Intelligence Daily

1. Qwen share markets collapse as Alibaba's weekly usage falls to a one-week low. Every open interest bracket on Kalshi's "Qwen market share this week?" series — from "above 4.9%" to "above 5.5%" — repriced from near-certain to near-impossible within about 24 hours, with the 5.5% strike falling from 97% to 3% and the 5.2% strike from 99% to 1%. That is a one-week share collapse for the model that led the Chinese pack last week, and it says the usage surge behind Qwen's earlier dominance was concentrated, not broad-based.

2. The coding rankings are moving with it. ChatGPT jumped from 18% to 83% to be the top coding model this month on the Arena.ai list, while Claude fell from 60% to 10% in the same market; on the separate "Top Coding AI this month" board Claude holds 22%. The two boards now disagree, which means the leaderboard source traders are using is itself the contested variable — and it suggests the current coding crown is contested, not settled.

3. App-download markets are quietly repricing the consumption picture. Kalshi's "Grok app downloads above 40k in August?" moved from 8% to 58% as August's final week unwound, implying Grok finished the month well below the 40k mark traders had priced, while "Claude app downloads above 2600k in August?" climbed from 2% to 20% — a small market, but a directional shift that says Claude's August print is likely to land higher than the market had feared.

4. The AI bubble question is drifting, not tipping. "AI bubble burst in 2026?" on Polymarket ticked up to 12% from 11%, and "OpenAI announces AGI before 2027?" dropped 40% to 16% — a notable de-risking of near-term AGI claims. Read together with the news cycle (the Seattle Times and Newsday suits against OpenAI, the rogue-agent coverage, the compute-heavy fundraising at Nscale and Crusoe), the market is pricing in legal and operational friction at the frontier, but still pricing continued expansion at roughly an 88% probability of no 2026 downturn.

5. OpenAI's IPO calendar remains the market's clearest consensus. "OpenAI not IPO by December 31, 2026?" holds at 92%, "Anthropic IPO before 2027?" at 92%, and the "OpenAI IPO above $800B closing cap?" at 94% — high-conviction, low-movement positions that have now settled into a stable regime, with the residual uncertainty concentrated in the 2027 window rather than this year.