Rising debt burdens and data center capex among hyperscalers, paired with collapsing inference prices from Chinese open-weight models like Qwen and GLM, are the main factors keeping Polymarket odds for an AI industry downturn low at 6% by year-end and 19% by mid-2027. Recent reports highlight OpenAI’s revised revenue figures, Oracle’s credit downgrade and payment delays, and token-price indices halving since June, underscoring the mismatch between build-out costs and monetization. Traders see these as classic late-stage bubble signals—high valuations, concentrated VC flows, and interest-rate sensitivity—yet note that sustained demand and capability gains could still delay resolution. Key near-term catalysts include hyperscaler earnings and any further price cuts or regulatory scrutiny on energy use.