AI Bubble Burst by 2026: Prediction Market Odds

The market defines an AI industry downturn as the occurrence of at least three specific triggering events within a 90-day window before the market's expiration at the end of 2026. These triggers are: NVIDIA’s stock falling 50% from its all-time high; the iShares PHLX Semiconductor ETF dropping 40% from its peak; a bankruptcy filing by OpenAI or Anthropic; OpenAI getting acquired; the H100 rental price falling to $1.00 or below for five consecutive days, as tracked by the SiliconData Silicon Index; or the stock of a major AI hardware supplier—Taiwan Semiconductor, ASML, Broadcom, Arista Networks, or Super Micro Computer—closing 50% below its all-time high. With a current probability of 14.2%, the market implies that a severe industry retrenchment is seen as a tangible risk but remains the less likely scenario. The resolution conditions are demanding: they require a simultaneous breakdown across hardware, software, and infrastructure. For instance, a sharp decline in Nvidia’s value alone would not be enough unless accompanied by two other failures. The two earlier-dated outcomes, for downturns by December 2025 and March 2026, both trade at 0%. This means that, at the time of pricing, the market sees virtually no chance that three of the six conditions will be met within the required 90-day windows ending on those dates. Observable indicators—such as stock prices, company solvency, and GPU rental costs—do not suggest an imminent cascade. The total market volume of over $2.9 million indicates meaningful capital behind these assessments. The 14.2% figure is not a trivial tail risk; it suggests that roughly one in seven market participants consider a serious unwind plausible by the end of next year. The market can resolve early if three conditions are met at any point before the expiry, but the description stresses that a resolution to ‘Yes’ requires official confirmations or a consensus of credible reporting. The 0% probabilities for the nearer dates indicate that none of the watchpoints are flashing, reinforcing the view that any downturn, if it materialises, is more likely to develop later in the forecast period.