Is The Nasdaq Selloff 2000 Or 2018?
Follow the whole story: every piece in this series is on our AI storyline page.
What Does Everyone Already Know?
The Nasdaq 100 closed Monday more than 10% below its Oct 6 record. It gained 0.7% on Tuesday, when chip stocks bounced after three losing sessions, and 0.9% on Wednesday, ending about 9% below the record. The debate is the same one every correction brings: is this a bubble bursting, or a pullback in a bull market?
The two cases are very different. From March 2000 to October 2002, the Nasdaq 100 lost about 83%. To get back to even from there takes a gain of almost 500%. In late 2018, the Nasdaq fell about 24% from its late-August high to Christmas Eve, then regained it within about four months.
| The number | 2000 | Today | Score |
|---|---|---|---|
| Valuation: forward P/E of the leaders | About 52 (median) for the seven largest; Cisco over 100 | In the 20s for the Nasdaq 100, in published estimates this year | 2018-like |
| Profits: what the leaders earn | Cisco earned about $2.7B in a year | Microsoft brought in $58B of operating cash in one quarter | 2018-like |
| Funding: who pays for the build-out | Mostly debt: telecom carriers borrowed to lay fiber | Mostly the giants’ own cash, but they now spend about $1.15 per $1 of operating cash, and Amazon and Alphabet are borrowing too; Oracle and CoreWeave borrow heavily | Moving toward 2000 |
Two of three numbers say correction. The third says the risk is growing.
See more lines: where the numbers come from
2000 forward P/Es are as of March 2000 for the seven largest companies at the time, from a Goldman Sachs comparison as reported in the press (median about 52; Cisco about 102, Oracle about 85, Microsoft about 53, Intel about 42). Published estimates of the Nasdaq 100’s forward P/E this year range from the low 20s to the high 20s; we haven’t measured it. The $1.15 is the four cloud giants’ Q3 2026 capex (about $202 billion) divided by their operating cash flow (about $175 billion), our estimate from reported figures; the full calculation is in our piece on why nobody stops spending on AI, the first question listed at the end.
Are We Paying 2000 Prices?
No. In March 2000, investors paid a median of about 52 times next year’s expected earnings for the seven largest companies, and more than 100 times for Cisco, then the most valuable company in the world. This year, published estimates put the whole Nasdaq 100 somewhere in the 20s times forward earnings. That’s expensive by history. But it’s about half of 2000, or less. Score: 2018-like.
Do Today’s Leaders Actually Make Money?
This is the biggest difference. Many 2000 Nasdaq companies had little or no profit. Even Cisco, the profitable leader, earned about $2.7 billion in its 2000 fiscal year while worth more than $500 billion at the peak.
Today’s leaders are among the most profitable companies ever built. Microsoft brought in $58 billion of operating cash in the quarter ended September, up 29% from a year ago. Alphabet earned about $26 billion of profit in the same quarter. Azure grew 48% and AWS 39%. When prices fall on companies earning that much, the earnings give investors something solid to value, which 2000’s leaders didn’t have. Score: 2018-like.
Who Is Paying For The Build-Out?
Here’s the number that doesn’t fit the comfortable story. In 2000, the build-out of the internet was paid for largely with borrowed money. When the money stopped, the builders went bust and their suppliers’ orders vanished.
Today the four cloud giants mostly pay with their own cash, though Amazon and Alphabet together borrowed a net $74 billion in the third quarter. In that quarter, their capital spending ran about 15% ahead of the cash their businesses brought in. A year ago it was well under that line. And the companies that don’t have that cash are already borrowing heavily. Oracle had about $125 billion of borrowings at the end of August, plus another $288 billion of data-center leases that haven’t started yet. CoreWeave carried $35.6 billion of debt at midyear and holds a junk rating.
The market has already drawn the line. Over the last 12 months, Oracle is down 50% and CoreWeave 58%, while Microsoft is up 8%. The debt-funded edge of AI is in its own 2000. The cash-funded center is not, yet. Score: mixed, moving toward 2000.
What Should You Do This Week?
If you can’t tell which year it is, size from the worse one. The idea is simple: decide the most you could lose on your AI holdings without changing your life or your plan, then work backward from the 2000 case, not the 2018 one.
An example. Say you have $1 million invested and could live with losing $150,000 on the AI part. In a 2018-style drop of about 25%, you could hold $600,000 of AI stocks. In a 2000-style drop of about 75%, you could hold only $200,000. That $400,000 gap is the price of not knowing. Sizing for 2000 means you won’t be forced to sell at the bottom if it turns out to be 2000, and you still own a meaningful stake if it’s 2018.
Three practical steps:
- Count what you own. Index funds included. Many investors own far more AI than their direct holdings suggest.
- If you’re over your 2000-case size, trim in stages, not all at once. In a taxable account, use any losses from this year’s selloff to offset gains before Dec 31.
- If you’re under it, add at prices you set in advance: a ladder of limit orders at, say, 10%, 20% and 30% below today, or a cash-secured put at the price you’d buy. Option prices change daily, so check the bid and ask before you trade. This is education, not personal advice.
And watch the funding number every quarter. If the four giants come to finance the build-out mainly with bonds rather than cash, the scorecard flips to 2000 on the third count, and it’s time to revisit your size.
What’s The Bigger Idea?
Know the worst case before you need it. The worst case isn’t volatility; it’s the specific loss you’d live through, and how long it lasts. For the Nasdaq 100 that history is a fall of 83% and a recovery measured in years. You don’t need to believe it will happen to plan for it. You only need to make sure that if it does, you don’t have to sell.
On valuation and profits, this is 2018. On funding, it is drifting toward 2000. Today’s leaders cost about half of what 2000’s did, or less, and earn vastly more. But they now spend more than they bring in, and the debt-funded builders are already down 50% to 58%. Size your AI holdings for the 2000 case, buy any further drop at prices you set, and recheck the funding number each quarter.
What Should You Ask Next?
- Why can’t anyone stop spending on AI?
- How much AI do you really own? Count it in three numbers
- The stock you researched just fell 30%. Fear is a planning mistake
We’ll rescore these three numbers after every quarter’s earnings. Follow the story, and send us the number you’d add to the scorecard.
Is Your Portfolio Sized For 2000 Or For 2018?
Corrections test concentration, not conviction. If you have $1 million or more invested, see how exposed your portfolio is to the AI trade in a free 15-minute call with a Trefis partner firm.
For questions about this article or underlying research, email help@trefis.com.