About the AI Bubble Oracle

An open, data-driven tracker for the question everyone in AI is quietly asking: are we in a bubble, and if so, has it burst yet?

What this is

The AI Bubble Oracle takes a codified definition of an "AI bubble burst" — the six-part condition from the Polymarket market "AI bubble burst in 2026?" — and tracks, from public data, exactly where we stand against it in real time. Each condition (NVIDIA and semiconductor drawdowns, a major supplier collapse, an OpenAI/Anthropic bankruptcy or acquisition, and H100 rental prices cratering) gets its own live chart, and the contract's "3 of 6 within 90 days" rule is evaluated in front of you.

It is deliberately transparent about how it decides. The Data Sources section documents every feed, the processing pipeline that normalizes it, and — importantly — the ambiguities and options baked into the rules (what counts as an all-time high, how the 90-day window is read, which H100 index applies), so you can see how each interpretation changes the answer.

Where it's going

Tracking a single market's resolution conditions is the starting point, not the destination. The larger goal is to build a living dashboard of the data sources and indicators that actually help anticipate an AI downturn — well beyond one market's rulebook.

• Wire up the conditions that are still manual today: bankruptcy dockets and acquisition news.
• Add leading indicators — compute prices, capex and datacenter buildout, funding rounds, model-economics signals — that may foreshadow a turn before the headline conditions trip.
• Make every indicator inspectable and every assumption a toggle, so the tool argues its case rather than asserting a number.

Open source

The whole thing — data collectors, the evaluation engine, and the site generators — is public. Fork it, check the methodology, or open an issue.

The data refreshes automatically every day via a GitHub Action, and the site redeploys on every push.