What If We’re Not in 2000 Yet?

The Four Optimist Reasons

1. Cycle clock	2. Productivity	3. Earnings	4. Valuation
Bespoke’s launch-to-launch analogue maps today closer to Sept. 1998 than March 2000.	AI adoption is broadening and could lift output per worker.	Today’s leaders have real profits and strong earnings momentum.	Expensive, yes—but earnings and balance sheets are far stronger than at the 2000 speculative peak.

MAYBE WE’RE LOOKING AT THE WRONG YEAR

Welcome to Stock Talk. For the last year, we’ve asked a lot of questions about the similarities between today’s AI boom and the dot-com bubble. And there are plenty.

Huge technology spending. Excitement over a revolutionary new technology. Semiconductor stocks soaring. And investors worry that we’ve seen this movie before. But today, we’re going to turn that argument upside down. What if the AI bears have the right movie, but the wrong scene?

What if this is another 1995-2000 dot-com-like technology cycle…but instead of being in March 2000, near the end, what if we’re closer to September 1998? That would make an enormous difference.

And a fascinating chart from Bespoke Investment Group gives us a reason to consider exactly that.

Today, I’m going to give you four reasons for optimism:

One: The market clock may say 1998—not 2000.

Two: Productivity may be entering its strongest phase.

Three: Today’s stock prices have something many dot-com companies didn’t have, real earnings.

And four: Today’s earnings make this very different from the speculative peak of 2000.

And that brings us close to Ed Yardeni’s “Roaring 2020s” thesis.

POINT ONE — THE CLOCK MAY SAY 1998, NOT 2000

VISUAL 1 — Bespoke page 13: Nasdaq Composite after Netscape launch vs. ChatGPT launch. Launch dates, 944-day returns, and the March 2000 endpoint.

Nasdaq Composite	Netscape Cycle	ChatGPT Cycle
Launch	12/19/1994	11/30/2022
First 944 days	+128.5%	+141.1%
Relative point in cycle	~Sept. 1998	~Sept. 2026
Final dot-com peak	3/10/2000	???

Let’s start with the most important chart. Netscape launched on December 19, 1994. That helped open the Internet age to millions of people. ChatGPT launched on November 30, 2022. That arguably did something similar for artificial intelligence.

Now line up the two cycles by days since launch. At 944 days, Bespoke found the Nasdaq Composite had gained 128.5 percent after the Netscape launch and 141.1 percent after the ChatGPT launch.

The similarity is remarkable. But here’s what’s even more important. Look at what happened next.

The blue Netscape line didn’t peak in 1998. Far from it.

The Nasdaq eventually reached a cumulative gain of roughly 592 percent by March 10, 2000.

That does not mean today’s Nasdaq is going to do the same thing. Markets don’t repeat like clocks.

But it raises an important question. Why are we assuming today’s AI cycle is already at the 2000 finish line? Bespoke’s time-aligned comparison argues something very different:

The calendar says 2026. The analogue might say 1998 not 2q 2000. 1998 wasn’t the end of the dot-com boom. It was the beginning of its most powerful stage.

THE INTEREST-RATE PARALLEL

This analogy gets more interesting when we add the Federal Reserve. The 1990s weren’t one straight line higher. The Fed raised rates. Bond yields moved around. The Asian financial crisis hit. Then LTCM nearly collapsed in 1998. The Fed responded with liquidity.

Today, the Fed is again wrestling with strong economic growth, inflation and a technology investment boom. Bespoke calculates the Fed’s recent communication has shifted hawkish, while the market has been debating whether Warsh’s Fed might hike again.

And the 10-year Treasury yield is approaching 4.8 percent. That’s a headwind.

But remember: interest rates were also a headwind during the late 1990s—and stocks kept climbing because earnings and productivity overwhelmed that headwind. That brings us to reason number two.

POINT TWO — PRODUCTIVITY: THE SECRET WEAPON

VISUAL 2 — Productivity flywheel: AI Investment → Workers Produce More → Productivity Rises → Unit Costs Fall → Profit Margins Rise → Real Wages Can Rise → GDP + Earnings Grow Faster.

This may be the most important part of the bullish argument. AI isn’t valuable simply because people can ask ChatGPT questions. The economic prize comes when businesses use AI to produce more output with the same number of workers.

That’s productivity.

Imagine a company has 100 workers. Last year they produced 100 widgets. This year, because of AI, those same workers produce 110. The company can make more money without increasing labor costs by 10 percent.That can mean higher wages, higher profits, lower inflation, and higher economic growth—all at the same time.

That’s essentially the heart of Yardeni’s Roaring 2020s thesis: technology-driven productivity allows real economic growth, wages and corporate profits to improve without automatically producing runaway inflation. And we’re starting to see evidence that AI is spreading beyond Silicon Valley.

Bespoke cites a New York Fed survey showing AI adoption among service companies rising from 25 percent in 2024 to 61 percent in 2026. Manufacturing adoption jumped from 16 percent to 51 percent.

But more than 90 percent of manufacturers and 75 percent of service companies still describe their AI investment as only minimal to modest.

Think about that. Adoption is exploding—but investment is still relatively small for most businesses.

That doesn’t sound like the end of an economic cycle. It sounds more like the implementation phase is beginning.

POINT THREE — FEMO, NOT JUST FOMO

VISUAL 3 — FEMO vs. FOMO comparison table.

	Dot-Com Late Cycle	AI Cycle Today
Technology	Internet	Artificial Intelligence
Investor driver	Increasingly FOMO	FEMO + FOMO
Mega-cap profits	Much smaller	Enormous
S&P EPS trend	Strong but nearing cycle peak	Accelerating
Profit margins	Lower	Near record levels
Capex	Telecom / Internet	AI / data centers / power

This is where the 2020s could be fundamentally different from 2000. The late dot-com boom increasingly became about FOMO, Fear Of Missing Out. There is pockets of this in the markets right now but far from the 1999-2000 speculation I lived through.

Companies with little revenue could achieve enormous valuations. Some didn’t have earnings. Some didn’t even have a believable path to earnings. Today we have speculation in some areas.

But the biggest companies driving the AI boom are also producing enormous profits. Yardeni has given today’s version a different acronym: FEMO—Fabulous Earnings Momentum.Yardeni calculates S&P 500 EPS increased 19 percent year-over-year in Q1 2026. In Q2, reported EPS growth surged to roughly 50.7 percent—although mark-to-market investment gains inflated that figure. So lets strip out those one-time gains, underlying Q2 earnings growth around 25 percent.

And analysts currently expect approximately 23.6 percent EPS growth in Q3 and 27.9 percent in Q4.

That’s not a forecast of a distant future. Those are earnings being generated now. Yardeni’s 2027 consensus EPS estimate has climbed to approximately $419. Forward S&P earnings calculated by Factset has reached about $402 last week—another record.

This is why the distinction matters. Stock bubbles normally become dangerous when prices keep rising but earnings stop supporting them.Right now earnings aren’t falling. They’re accelerating.

AND THE PHYSICAL ECONOMY IS RESPONDING

VISUAL 4 — Bespoke page 14: Data-center vs. office construction spending. Highlight $9.5B in March 2020, $75.2B in July 2026, and the 665% increase.

Bespoke’s page 14 chart provides another remarkable visual. Data-center construction has gone nearly vertical since AI took off. In March 2020, annualized data-center construction spending was only about $9.5 billion. By July 2026: $75.2 billion. That’s an increase of roughly 665 percent.

Meanwhile, office construction has fallen to around $44.8 billion. That’s not just financial speculation.

Those are actual buildings. Actual electrical systems. Actual cooling. Actual semiconductors. Actual jobs. And Bespoke finds that the three job categories with the biggest increases in postings this year are mechanical, industrial and electrical engineering—exactly the skills required for this physical AI buildout.

POINT FOUR — YES, STOCKS ARE EXPENSIVE. BUT THIS ISN’T MARCH 2000

VISUAL 5 — Four reasons the optimist says “earlier.”

	Bullish Evidence	What Could Break It?
1. Cycle clock	Bespoke analogue ≈ Sept. 1998	Analogue stops working
2. Productivity	AI adoption spreading rapidly	Productivity fails to materialize
3. Earnings	20%+ underlying growth	Earnings estimates roll over
4. Valuation	High P/E supported by rising EPS	Rates/P/E compression overwhelm EPS

Now let’s address the biggest argument from the bears. Valuation.

Stocks aren’t cheap. That’s true. Yardeni has discussed a forward P/E around 22 times, versus approximately 25-27 times near the 1999 technology bubble peak. One source I reviewed had the 1q2000 peak at >30x calendarized EPS.

And there’s another warning sign.Bespoke shows the S&P 500 ETF’s dividend yield has dropped below 1 percent for the first time since the 1999–2001 period.

Meanwhile, the 10-year Treasury yield is close to 4.8 percent. The gap between the two is the most negative since 2002. That’s real competition for stocks. So don’t ignore valuation.

But don’t confuse “expensive” with “the cycle must end tomorrow.”Bespoke makes an important observation. That stock-versus-bond yield spread was actually even worse during much of the 1990s.

And stocks continued rising. Why? Earnings. Productivity. Economic growth.

ONE MORE REASON I’M NOT READY TO CALL 2000

Look underneath the S&P 500. This doesn’t look like a market where every investor is blindly chasing AI.

In fact, quite the opposite. Bespoke shows the Nasdaq 100 has struggled since June. Semiconductors have corrected. AI infrastructure stocks peaked earlier this summer.

Yet the S&P 500 is still up more than 12 percent in 2026. Health care led the summer. Financials gained more than 12 percent. Energy is up roughly 43 percent year-to-date through September 3. Biotech has surged. Value remains in an uptrend.

And the equal-weight S&P has been much healthier than you’d expect if this were simply a one-trade AI bubble.  Yardeni sees something similar. As of mid-August, his “Impressive 493”—the S&P 500 excluding the Magnificent Seven—was up 17.6 percent year-to-date, versus only 3.8 percent for the Magnificent Seven.

That’s broadening—not narrowing.

THE ROARING 20s CASE

So let’s put it together. Yardeni currently assigns an 80 percent subjective probability to his Roaring 2020s scenario and only 20 percent to a recession-driven bear market.  He’s raised his 2026 S&P 500 target to 8,400. And he’s still targeting 10,000 by the end of 2029.

That sounds outrageous until you look at the math. From the level Yardeni cited in August, getting to 10,000 required roughly another 28.5 percent, or only about 7½ to 8 percent annualized price appreciation over the remaining years of the decade.  That’s hardly dot-com-style mathematics.

It’s mostly an earnings story.

CLOSE — THE OPTIMIST

So here’s my takeaway. I’m not saying stocks can’t correct. They will at some point, we’ve been looking for some weakness for a few months starting October. And I’m not saying some  AI stocks aren’t expensive. Some absolutely are. And I’m certainly not saying the Bespoke chart guarantees we’re going to repeat 1998 through 2000.

It doesn’t. But investors should be careful about making the opposite mistake too.

Don’t declare the end of a cycle simply because prices have gone up.Bull markets usually don’t die because somebody notices they’re old. They die because something underneath them breaks.

Earnings. Liquidity. Credit. The economy. Or monetary policy.

And today? Earnings are accelerating. AI adoption is broadening.

Business investment is strong. Productivity has a chance to improve.

The market itself is broadening beyond mega-cap technology.

And the Bespoke launch-to-launch clock says something worth thinking about:

Maybe this isn’t March 2000. Maybe it’s closer to September 1998.

And if that’s even approximately correct, then the biggest mistake investors make over the next few years might not be owning too many stocks.It might be becoming too pessimistic, too early.

That’s why today’s Stock Talk isn’t called “The Bubble.” It’s called “The Optimist.”

Because sometimes the most dangerous four words in investing aren’t: “This time is different.”

Sometimes they’re: “This has to end.”

Selected Sources / Production Notes

• Bespoke Investment Group, The Bespoke Report, September 4, 2026 — especially pages 13–15, 18–22 and 28.
• Ed Yardeni / Yardeni QuickTakes — Roaring 2020s, Fabulous Earnings Momentum, current earnings and market target commentary.
• General information only. Not financial advice. Past performance does not guarantee future results.