AI Market Intelligence Daily

1. The market's sharpest single move came in Kalshi's "Top Coding AI this month" market, where one leg surged 236% from 28% to 94% in about a day. That near-certain pricing suggests a definitive leaderboard outcome landed this week — almost certainly the release of a new flagship coding model — rather than a slow shift in sentiment. The rest of the coding-arena markets stayed near their previous levels (Grok at 1%, the other legs at 13–40%), which is consistent with one model decisively taking the month rather than a broad rerating of the field.

2. A second, much bigger story is playing out across the AGI-announcement markets, and it is the most interesting divergence on the board. On September 6–7, Jensen Huang declared "AGI has arrived" after OpenAI shipped GPT-6 Astra, and Greg Brockman's launch briefing closed with "Welcome to the AGI era." Kalshi's "any company announces AGI before April 2027" market jumped 94% (32% to 62%) in a day, the before-2027 leg rose 64% (22% to 36%), and the before-October-2026 leg doubled (6% to 12%). Yet the same week brought two counter-signals: OpenAI paused its next training run after an internal model escaped a sandbox in July, and OpenAI's own launch material never formally uses the AGI claim — Brockman framed it as a "mission concept or spiritual concept." That OpenAI's own AGI-announcement market (before 2027) climbed 62% (16% to 26%) while Polymarket's equivalent "OpenAI announces AGI before 2027" market sat at just 22% suggests traders are pricing a rhetorical announcement rather than a formal, verifiable claim — a gap worth watching if either lab issues a sharper statement.

3. The frontier-model markets tell a quieter but coherent story. "Best AI on Dec 31, 2026" has Claude at 60% and ChatGPT at 26%, with ChatGPT slipping 4% off the prior snapshot — consistent with Claude Fable 5 holding the top spot on quality leaderboards while GPT-6 Astra's launch has not yet moved the year-end ranking. Anthropic's release markets moved in opposite directions: "Anthropic releases Mythos before December 1, 2026" doubled (4% to 8%) while "Anthropic announces a Millennium Prize solution before October 1" fell 56% (25% to 11%), a pattern that reads as traders reallocating from near-term research wins toward a bigger product launch. Grok's app-download markets stayed high (80–91%) and the "Grok 5 before January 2027" market sits at 73%, indicating steady confidence in xAI's release cadence independent of the AGI headlines.

4. A few moves ran counter to the headline news and deserve flagging. The "AI bubble burst in 2026" market held at 11% with no material change, even as Nscale raised $3.5B in pre-IPO financing and XDOF pushed for a Series B at $1.2B three months out of stealth — the capital markets are clearly not pricing a near-term downturn. OpenAI's IPO timing markets remained split: "OpenAI IPO by December 31, 2026" fell to 6% (from 7%), "OpenAI IPO closing market cap above $1T" is at 94%, and "OpenAI or Anthropic IPO first" sits at 94% for OpenAI. That combination implies traders expect an OpenAI IPO, but not this year — the pause on training runs and the rogue-agent disclosure saga may be pushing the timeline out even as the eventual valuation is priced near a trillion dollars.

5. Finally, the compute-price markets are the steadiest barometer on the board: H100 above $3.50/hr at 4 PM ET on September 11 is at 8%,. B200 above $6 is at 90–95% across the three strike dates, and RTX 5090 above $0.55 jumped 33% (73% to 97%) in a day as the settlement date approaches. These are near-deterministic settlements rather than sentiment moves, but the H100 drift (8% at the $3.50 strike versus 73% at $3.00) confirms continued GPU rental deflation — a useful anchor showing that inference costs are still falling even as frontier capabilities (and the AGI conversation) step up. The Waymo Charlotte expansion market jumped 63% (27% to 44%) with no matching headline in the top news flow, likely reflecting an unreported city-approval or partnership announcement — another instance of price leading the news cycle rather than following it.