Dot-com bubble chart vs AI bubble chart

What does a dot-com bubble chart show at first glance?

A dot-com bubble chart shows 3 features: a long climb through the second half of the 1990s, a sharp top, and a steep drop. The Nasdaq Composite peaked at 5,048.62 on 10 March 2000, and the NASDAQ-100 reached its low of 1,114 on 9 October 2002, down 78% (Wikipedia, Dot-com bubble). The drop from the top to the low took 31 months.

Those 2 numbers come from 2 different indexes. The first measures the Composite and the second the NASDAQ-100, so a line that joins them mixes 2 scales. Name the index on every axis. A reader who sees a clean line from 5,048 to 1,114 should ask which series the line draws, and which dates it uses.

Should a dot-com bubble chart use a linear or a log scale?

A linear scale draws equal distances for equal point changes. At the top of the bubble, a move of 100 points looks as large as it did at the bottom, although it is a far smaller percentage. A log scale draws equal distances for equal percentage changes, so the early years of the climb stay visible next to the final surge.

Use a linear scale when the question is about money lost, and a log scale when the question is about speed. The EQOS Asia AI study states its growth figure as a percentage over time, a ×4 rise in 18 months, which reads naturally on a log scale. Pick one scale and state it in the caption.

How can a dot-com bubble chart and an AI bubble chart share one axis?

Rebase both series to 100 on a common start date, then plot months from that date and not calendar years. The 2 curves then answer one question: how fast did each one rise? The AI side comes from the EQOS Asia AI study, which records valuations up ×4 in 18 months, an average of about 8% a month. A ×4 rise moves the series from 100 to 400.

The start date decides the picture. Choose a start that you can justify before you see the result, such as the first month of the funding wave, and write it in the caption. The dot-com series follows the same rule: rebase from 1995 for the full rise, or from 1999 for the steepest year, when the Nasdaq Composite rose 85.6% (Wikipedia, Dot-com bubble).

What does a 78% fall on a dot-com bubble chart mean for recovery?

A fall of 78% leaves 22% of the starting level. To return to the start, the series needs a gain of about 355%, because 100 divided by 22 is 4.5. This is why a drawdown chart reads worse than a price chart: the drop looks like a single step, and the climb back looks like a mountain. The arithmetic applies to any index.

Apply the same rule to AI. If valuations that rose ×4 in 18 months fell by 78%, they would stand at 88% of their starting level, which shows how much of a rise a deep fall can erase. This is an illustration of arithmetic, and it carries no forecast for the AI bubble.

Which second line belongs beside the price on a dot-com bubble chart?

Price alone hides the reason for the fall. A second line for the price-to-earnings ratio shows how much buyers paid for each unit of profit. In late 1999 the Nasdaq ratio stood above 200 (Wikipedia, Dot-com bubble). When the ratio rises while earnings stay flat, the chart shows expectations growing faster than results.

The AI version needs a revenue line. The EQOS Asia AI study records about $200B a year spent on AI research and deployment, and 95% of projects failing in production (MIT 2025). Plotting spending beside revenue per user each quarter shows whether the gap narrows. If the lines meet, the boom reading gains weight, and if they part, the bubble reading does.

Which mistakes make a dot-com bubble chart vs AI bubble chart misleading?

4 mistakes recur. A window that starts at the bottom of a drop makes any later move look large. Two indexes joined on one line mix their scales. Two axes with different units make unrelated lines look alike. And a caption without a source or date makes the chart impossible to check. Each mistake changes the story without changing a single data point.

A reader can defend against all 4 with the same habit: ask for the source, the start date, the index, and the scale. As of October 2026, the AI side of any chart rests on a short history of a few years, so a pattern match with 1995 to 2002 is a hypothesis to test and not a finding. A test needs a rule that could fail.

How do you build your own dot-com bubble chart vs AI bubble chart?

5 steps build a fair chart. Name the index, fix the start date, rebase both series to 100, choose the scale, and cite the source with its date. Do the work in that order, because each step fixes a choice that the next step builds on. Save the chart with its settings, so a colleague can redraw it and obtain the same picture.

The EQOS Asia Λ simulator estimates a φ(S) score for your own case in 10 minutes, and the pricing page lists the formats for a deeper study. The enterprise AI market scores 0.337, in the fragile zone. A chart shows the shape, and a score gives the number you can recompute.

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