AI Stocks vs the Dot-Com Bubble: The Valuation Numbers, Side by Side
Cisco at 140× forward vs Nvidia at a fraction of that — and the two dimensions where 1999 still rhymes.
Put the two eras on one axis
The cleanest way to test "is AI a rerun of 1999?" is to price both eras the same way. At the March-2000 Nasdaq peak, the market's four horsemen traded at these forward multiples: Cisco ~140×, Oracle ~150×, Sun Microsystems ~120×, and the era's "cheap" mega-caps — Microsoft ~60×, Intel ~50×.
Today's AI leaders, on trailing earnings (as of 2026-07-02):
| Company | TTM PE today | 1999 counterpart |
|---|---|---|
| Nvidia | 29.8× | Cisco at ~140× forward |
| Microsoft | 23.2× | Microsoft-1999 at ~60× forward |
| Alphabet | 27.4× | Oracle at ~150× forward |
| Meta | 21.2× | Sun Micro at ~120× forward |
The gap is not subtle. Nvidia — the most expensive-feeling stock of the era — trades at roughly 30× trailing earnings, a fraction of Cisco's 1999 forward multiple. The 2025-26 AI trade has been carried substantially by earnings, not just multiple expansion: the driver decomposition shows it year by year.
Where the 1999 comparison does bite
- Index concentration. The Magnificent 7 hold 31.5% of the S&P 500 — above the 2000 top-seven share. The single-theme exposure is real even if the multiples are tamer.
- Broad-market valuation. The S&P's CAPE at 40.7× is in the 99th percentile — 2000 is one of only two episodes higher.
- Capex cyclicality. 1999's fiber build-out became 2001's glut. Today's AI datacenter capex (tracked in the M7 capex panel) is the same shape of bet: current earnings funding future capacity whose returns are not yet proven.
The honest verdict
On single-stock valuation, today is not 1999 — no leader trades within shouting distance of Cisco's 140×. On index structure (concentration, one narrative carrying the market, capex acceleration), the rhyme is legitimate. Both things are true at once, which is exactly why the question refuses to die.
FAQ
Is Nvidia as expensive as Cisco was in 1999?
No. Nvidia trades at about 30× trailing earnings; Cisco peaked at roughly 140× forward earnings in March 2000 — several times higher on a comparable basis.
How does the AI boom compare to the dot-com bubble?
Single-stock multiples are far lower (NVDA 30× vs CSCO ~140×), but index concentration is higher: the Mag 7 hold 31.5% of the S&P 500, above the 2000 top-seven share.
Computed from /api/mag7/ai-valuation.json and /api/mag7/concentration.json, refreshed each trading day.