AI Market Intelligence Daily

2. The largest single move is the "RTX 5090 compute per hour above $0.60 on Sep 25" contract, which surged from 1% to 54% in roughly 24 hours. Neighboring strikes moved with it: the $0.55 strike went from 17% to 60%, the $0.65 strike from 1% to 5%, and the "price up next week" contract rose from 46% to 57%. The move is directional across the whole ladder rather than an isolated outlier, which suggests traders are repricing a sustained price level for consumer GPUs rather than reacting to a single data print. The same pattern shows up in the A100 ladder (the $1 strike up from 25% to 40%) and the H100 ladder (the $3.75 strike up from 4% to 15%), so the repricing spans generations, not just one SKU.

3. That compute repricing lines up with the clearest macro signal in today's news: Neocloud Lambda securing $1B in debt specifically to buy more chips. Debt-financed GPU expansion is exactly the kind of demand shock that pushes hourly rental prices up, and the fact that compute markets moved in the same direction as that headline is the strongest news-to-price correlation in the report. Nvidia's $3.5B investment in MediaTek and the "AI advantage is moving beyond the GPU" story reinforce the same theme: capital is still flowing into the chip supply chain, and the prediction markets are pricing in tighter, more expensive compute for the next several months.

4. The model-release markets point the other way, with traders shifting their end-of-year pick away from OpenAI and toward Anthropic. Claude now sits at 70% for "Best AI on Dec 31, 2026," while ChatGPT fell from 16% to 12% and Grok is only 10%. GPT-6 odds actually ticked up for an October release (35% to 45%), so the shift is not about OpenAI slowing down; it is about Claude being seen as the stronger model at year end. That is a notable divergence from the news sentiment, because the day's headlines are dominated by Sony Music and Warner suing Anthropic over intellectual property theft, a reputational headwind for exactly the model the markets are crowding into. The market is clearly discounting the lawsuit as a non-event for model quality, which is a risk to watch if the case gains more traction.

5. OpenAI's IPO markets show fading near-term listing expectations and a lower implied valuation. The "OpenAI IPO by Dec 31, 2026" contract dropped from 14% to 13% while the "not IPO by Dec 31" market rose from 84% to 88%, and the "IPO before 2027" market sits at just 16%. The market-cap ladders for IPO day were repriced down across the board: the $750B to $1T band fell 48%, the $1T to $1.25T band fell 37%, and the $1.5T-plus band fell 35%. The market is not just pricing in a later listing; it is pricing in a smaller one. Against the backdrop of Apple's trade-secret case against a former employee accused of stealing data for OpenAI, a market that prices a 2026 listing at 13% is treating IP and legal overhang as a real delay factor.