AI bubble diagram: how a bubble inflates, step by step

What does an AI bubble diagram show from start to finish?

The diagram has six steps: a real advance, rising prices, easy funding, euphoria, a tipping point, and a correction. The horizontal axis is time and the vertical axis is price. A second line, revenue, climbs more slowly, and the space between the two lines is the gap that defines the bubble. Reading the picture means asking which step each piece of news belongs to.

The diagram is a map, not a forecast. Real cycles skip steps, repeat steps, or stretch one step over years. What the map gives you is an order of events. When you know the order, you can ask what normally comes next and which observable number would confirm it, and you can write the question down with a date.

What starts an AI bubble in the first step of the diagram?

Every bubble diagram begins with a real advance that changes expectations. For AI, the advance is a set of models that write, code, and answer questions at a quality many people find useful. The EQOS Asia AI study projects productivity gains of +40%, which shows why expectations moved. The first step is genuine, and that is why it convinces so many people.

At this step prices follow earnings, and the gap is small. Early buyers are paid for being early. The question for a reader is how long that first step lasted and how much of the later price rise it explains. A first step that explains only part of the rise leaves room for the next steps. Note the date of each data point you use, so the first step stays separate from the later ones.

How do rising prices pull new buyers into an AI bubble in step 2?

In the second step, rising prices become news. Buyers who ignored the advance notice the gain and join. The EQOS Asia AI study records valuations up ×4 in 18 months, which fits this step. Each new buyer lifts the price a little more, and the lift attracts the next buyer.

This loop is the core mechanism of the diagram. Price becomes the reason to buy, instead of earnings. Watch for the moment when explanations of a purchase shift from what the product does to how far the price has risen. That shift is the signal that step two is giving way to step three. Press coverage and fund inflows are two public places to see it.

How does easy funding extend an AI bubble in the third step?

Rising prices make funding cheap. Companies raise money on high valuations, and lenders accept the valuation as comfort. The AI industry spends about $200B a year on research and deployment (EQOS Asia AI study, April 2026). Spending at this scale builds data centres and chips that will need customers for many years.

Funding extends the bubble because it pays for growth before revenue arrives. It also raises the stakes: the larger the spending, the more revenue the system must produce. MIT 2025, cited by the study, finds that 95% of deployed AI projects fail in production, so revenue per project is the number to watch in this step.

What does euphoria look like near the top of an AI bubble?

Euphoria is the step where stories replace numbers. Predictions grow bolder, new terms appear, and a doubt is treated as a lack of vision. Error data offers a counterweight: the study notes that 47% of enterprise users took at least one major decision based on hallucinated content, a reminder that the tools still make mistakes.

Euphoria shows in behaviour. People who never followed markets ask for tips, firms add AI to their name or pitch, and projects launch before use cases exist. None of these signs fixes a date. Together they tell you the diagram has reached the step where the gap is widest. Record three such signs with dates, and you hold a simple euphoria index for your own use.

What tips an AI bubble from the top into a correction?

A tipping point is a shock that arrives when the system is stretched. Typical shocks are a funding squeeze, a missed earnings target, a new rule, or a public failure. The EQOS Asia AI study scores the enterprise AI market at φ(S) = 0.337, in the fragile zone, where the system works and a shock can tip everything.

The AI Act compliance date of 2 December 2027, listed by the study, is one scheduled rule change. The weakest dimension in the study is Communication at 0.305, just above the 0.3 stop threshold, so a public incident on explainability is a plausible shock. The diagram shows the shape of the fall, and the trigger decides its timing.

Where does AI sit on the AI bubble diagram as of October 2026?

Mapping the EQOS Asia figures onto the diagram gives a mixed picture. The advance is real, prices have risen fast, and funding is large. Results lag, with 95% of projects failing in production. These facts place AI between steps two and four, and the study’s verdict, bifurcation, means the next step can be a soft landing or a break.

Use the map as a routine. Each quarter, place the new figures on the diagram, note the step, and record the date. If you want a score for your own exposure, the Λ simulator gives a first estimate in 10 minutes, and the EQOS Asia 30-minute pre-diagnostic turns it into a ranked set of options.

Frequently asked questions

About the author

Stéphane Jambu

President, EQOS Forecast Asia

Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile