Dot Com Bubble vs AI Bubble: 2000 and 2026 Compared in 4 Charts

The dot com bubble vs AI bubble comparison comes down to four measures, and in 2026 they split two ways. Long-run valuations and market concentration are at or near dot-com extremes, while valuations on expected earnings and the damage done so far are nowhere close. Here is each measure, with the numbers for 2000 and for today.

The question is live again because the Nasdaq Composite closed at a record on October 6, 2026 and then fell into November, with the Nasdaq-100 entering a correction on November 2. Every selloff in this cycle brings the same search: is this 2000? The data says some of it rhymes and some of it does not.

44.2

Shiller CAPE at its December 1999 peak

40.9

Shiller CAPE on Sept 30, 2026

-77.9%

Nasdaq, March 2000 peak to October 2002 low

-6.8%

Nasdaq, Oct 6 record to Nov 4 close

Long-run valuations are close to 2000, forward earnings are not

The Shiller CAPE ratio divides the S&P 500 by ten years of inflation-adjusted earnings. It peaked at 44.2 in December 1999, the highest month in Robert Shiller's data. J.P. Morgan Asset Management put it at 40.9 at the end of September 2026, and Shiller's own file shows about 40.7 for October. On this measure the market is within about 8% of the dot-com record.

The forward price-to-earnings ratio, which uses analysts' estimates for the next twelve months, tells a different story. FactSet recorded 24.4 for the S&P 500 on March 24, 2000. On October 30, 2026 it was 18.7, below both its five-year average of 19.8 and its ten-year average of 19.1, FactSet reported. The gap between the two measures is the core of the argument today: prices are high against the past decade of profits, but not against the profits companies are expected to earn next year.

Forward estimates carry their own risk: they depend on analysts' assumptions about future profits. If AI capital spending slows, today's earnings estimates for chipmakers and cloud providers would be the ones exposed.

The market is more concentrated than in 2000

On concentration, 2026 is past the dot-com mark. Technology stocks made up 39.4% of the S&P 500's market value at the start of June 2026, above the nearly 35% reached in March 2000, according to a Reuters analysis of LSEG Datastream data. J.P. Morgan Asset Management estimated that the ten largest stocks made up about 40.6% of the index at the end of September.

Strategists disagree on what the concentration means. Fortune reported in May that, in the view of Goldman Sachs economist Pierfrancesco Mei, market breadth today is far less narrow than in 1999 to 2000. Julian Emanuel of Evercore ISI said the rally since March 2026 feels like 1999.

Our guide to Magnificent Seven concentration risk explains why a heavy top weighting matters for index investors either way.

The biggest company is far bigger, relative to the old leader

Cisco briefly became the world's most valuable company in March 2000, closing at $579.1 billion on March 24, just ahead of Microsoft, the Spokesman-Review reported at the time. Nvidia closed at about $4.99 trillion on November 3, 2026, according to 24/7 Wall St. Measures of Nvidia's valuation on next year's earnings vary widely depending on which estimates are used, so we do not show one here.

The drawdown so far is a fraction of 2000 to 2002

The Nasdaq Composite closed at 5,048.62 on March 10, 2000 and bottomed at 1,114.11 on October 9, 2002, a loss of 77.9% over two and a half years, according to St. Louis Fed (FRED) data. It took about 15 years to reclaim the 2000 peak, AFP reported in 2015. This time, the index closed at a record 27,599.89 on October 6, 2026. Its lowest close since was 25,164.36 on November 2, about 8.8% below the record, and it ended November 4 at 25,736.51, down 6.8%.

That is not evidence that the worst is over. In 2000 the Nasdaq also had sharp drops and strong rebounds before the decline became clear, as our dot-com bubble timeline shows. It does mean that, so far, 2026 looks like a correction rather than a collapse.

AI spending matters more to the economy than the 1990s build-out did

One measure where today already exceeds 2000 is the economy's reliance on the build-out. St. Louis Fed researchers estimated that AI-related investment added 0.97 percentage points to US GDP growth in the first nine months of 2025, compared with 0.81 points from the comparable categories in 2000. That was 39% of all growth, against 28% in 2000.

That cuts both ways. Spending at this scale props up growth while it lasts, and a slowdown would take a larger share of growth with it. The research group Epoch AI estimated that AI data centers, chips and networking alone came to about 0.8% of US GDP in the first quarter of 2026. Spending plans this year, and how much of them is funded with debt, are covered in our AI capex analysis.

What the AI Bubble Index says about 2000 vs now

The AI Bubble Index read 63.3 at the November 3 close, in the Normal band, down from 64.4 a week earlier. The part built for exactly this comparison is the long-run core, which uses only measures that existed before 2000 so today can be ranked against the dot-com years. It reads 80.1, well above the headline. Fragility, at 72.2, sits above Euphoria at 62.8, and sentiment and attention is the hottest family at 92.8.

The methodology page sets out the test the core was built to pass: reach its top tenth during 1998 to 2000 and stay out of it in quiet years. A high core reading describes conditions; it does not say when or whether a decline like 2000 to 2002 follows. The dashboard shows the core's history back to 1990.