Stock Talk - Dot-Com vs. AI Stock Cycles: “The Optimist” Stock Talk Update September 11, 2026

Episode Date: September 14, 2026

The AI boom keeps getting compared to the dot-com bubble. But what if the comparison is right—and the timing is wrong? A Bespoke Investment Group chart compares the Nasdaq after the launch of Netscape in 1994 with the Nasdaq after the launch of ChatGPT in 2022. At a comparable point in the two cycles, today lines up closer to September 1998 than the March 2000 dot-com peak. That doesn’t mean history will repeat. Historical comparisons aren’t forecasts. But it raises an important question: What evidence would tell us this AI cycle is nearing the end—and what evidence would suggest it may still be developing? In this episode of Stock Talk, we examine four major pieces of the argument: The historical Netscape-versus-ChatGPT market comparison, the spread of AI adoption and productivity, corporate earnings momentum, and today’s valuation and interest-rate risks. We also look at the massive increase in data-center construction, the difference between FOMO and what Ed Yardeni calls “Fabulous Earnings Momentum,” and evidence that market leadership has broadened beyond the Magnificent Seven. The goal isn’t to predict where stocks go next. It’s to ask a better question about where we may be in this technology and market cycle—and what evidence investors should watch from here. If you enjoy market analysis that looks at both the opportunity and the risk, subscribe to the channel and turn on notifications for future episodes of Stock Talk. If you’re approaching retirement and want to understand how markets, income needs, taxes, Social Security, withdrawal decisions, and risk fit into your personal retirement strategy, contact Oak Harvest Financial Group to start a conversation about your retirement plan.   Important Disclosure: This content is for general informational and educational purposes only. It is not individualized investment advice or a recommendation to buy or sell any security or pursue any investment strategy. Investing involves risk, including possible loss of principal. Historical performance and historical comparisons do not guarantee or predict future results. Estimates and market data discussed in this video may change. YouTube Chapters These timestamps follow the actual uploaded cut. 00:00 AI vs. the Dot-Com Bubble 00:18 What If We’re Comparing It to the Wrong Year? 01:03 Four Tests for the AI Boom 01:33 The Netscape vs. ChatGPT Nasdaq Chart 02:45 Why the September 1998 Comparison Matters 03:07 Subscribe to Stock Talk 03:17 What Happened After 1998? 04:05 AI’s Real Economic Test: Productivity 04:52 AI Adoption Is Surging 05:47 The Data-Center Construction Boom 07:25 FOMO vs. Fabulous Earnings Momentum 08:07 What Earnings Are Telling Us 09:25 The Biggest Problem: Valuation 09:51 Stocks vs. 10-Year Treasury Yields 11:03 Is the Rally Broader Than AI? 11:47 The “Impressive 493” vs. Magnificent Seven 12:30 So Is This Another Dot-Com Bubble? 13:10 The Four Tests That Matter From Here 14:00 The Real Lesson From 1998 vs. 2000 14:36 The One Question Investors Should Ask 15:03 What This Means for Your Retirement

AI boom keeps getting compared to the dot-com bubble. Heck, we've done it here for about 18 months, far before most other people. There are a lot of good reasons for this comparison. Technology spending is exploding. Semiconductor stocks have soared. AI is being called a once-in-a-generation technology. One chart from Bespoke Investment Group raises a very different question. What if we're comparing today to the right technology boom, but the wrong year?

Bespoke compared the NASDAQ after the launch of Netscape in 1994. with the NASDAQ after the launch of CHETGBT in 2022. Based on where those two market cycles line up in time, today's market looks a lot closer to September 1998 than March 2000.com peak. That doesn't mean that stocks are about to repeat what happened after 1998. This is a historical comparison, not a forecast,

but it raises a much more useful question. What evidence would tell us if this AI cycle is near the end and what evidence would tell us if it may still have time to develop. I live through 1999 to 2000 market, and some of today's similar areas are very real. So are the differences. That's why I don't want to turn to this is a simple bull-vers-bair argument. I want to test the comparison. We're going to look at four things.

Where the bespoke market cycle comparison places us, whether AI is actually improving productivity, whether corporate earnings support today's stock prices. And finally, the biggest issue might be with the optimist. domestic case, valuation, and interest rates. Okay, let's start with the bespoke chart comparing the NASDAQ after the launch of Netscape with a NASDAQ because this is what made me look at the whole argument once again. Netscape launched on December 19, 1994, for millions of people, it helped open the door to the internet.

ChatGBT launched on November 30th, 2022, arguably it did something similar for artificial intelligence. Bespoke then lined up the NASDAQ based on the number of days following each loan. So when you look at this chart, we're not comparing the calendar year 1998 with the calendar year of 2006. You're comparing how the NASDAQ performed at the same number of days after each major technology launch. During the first 944 days after the Netscape launch and the NASDAQ gained 128.5%. During that comparable period of 944 days after CHETGT launched, the NASDAQ had gained about 141%. The paths are surprisingly similar, but a similar path up to one date tells us nothing certain about what happens after that date.

The economy is different, interest rates are different, the companies are very different, valuations are different, and markets don't follow old charts like train tracks. What makes this chart interesting is where the 944 day point lands in the older cycle. On the Netskate timeline, it does not land in March 2000.com peak. It lands around September 1998. And that question we're going to keep coming back to throughout this video. Does the rest of the evidence look more like a cycle that's exhausted or one that's still developing? If you like this kind of market analysis, subscribe to the channel and hit the notification bell.

That's what Stock Talk has built around and there are a lot of moving pieces in this market that we're going to keep following. Now look at what's happened to the older Netscape line after that comparable period. The NASDAQ did not peak in September 1998. The eventual.com peak came on March 10th. 2000. By that point, the NASDAQ cumulative gain from the Netscape launch had reached roughly 592%. That number needs a giant warning label around it. It's a historical performance. It's not expected return for today's NASDAQ. There's no basis for assuming today's market will follow the exact same path. This historical comparison could stop working at any time. The useful point is much

narrower. A powerful technology rally by itself does not tell us exactly where we are in the market cycle. So if you want to know whether the boom is actually mature, we need to take some evidence more than that stock chart. The next piece of evidence may be much bigger. The biggest economic promise of AI isn't that you can ask a chatbot question. It's whether businesses can use AI to produce more without needing the same increase in people and costs. Imagine a business has 100 workers producing 100 widgets. Then it introduces AI tools and those same 100 workers can produce 110 widgets. has increased without employment, having increased 10%. That's productivity.

In the right environment, better productivity can support profits, wages, and economic growth while reducing some cost pressure. But there's a huge condition attached to this argument. AI actually has to deliver those gains. Spending billions of dollars on A doesn't automatically make a business or an economy more productive. It's one of the biggest things that could break the optimistic case.

Still, the adoption numbers are worth watching. Bespoke cites a New York Fed survey showing AI adoptions among service companies rising from about 25% 2004 to 61% now in 2006. Among manufacturers, AI adoption rose from about 16% to 51%. Those are big increases. But now look at the other side of the data. More than 90% of manufacturers and 75% of service companies still describe their AI investments as only minimal to modest.

That's an interesting combination. AI use is spreading very quickly, but many businesses are still spending relatively little on it. One interpretation is that we're still early in the implementation stage. Another possibility is that companies experiment with AI and eventually decide the productivity gains don't justify much more spending. We don't know which outcome wins yet. That's why actual productivity matters more than AI hype. Now look at the second bespoke chart because this one shows the AI boom moving out of the stock market and into the physical economy. This chart compares the U.S. Data Center construction spending with office construction spending.

Focus first on the data center line in March of 2020 annualized U.S. data center construction spending was roughly $9.5 billion. By July of this year in 2006, it had climbed to about $75.2 billion. That's an increase of roughly 665 percent. Now compare that with office construction. Office construction was around $44.8 billion. So the point of this chart is not simply that data centers are growing. It's that data center construction has become a major category, a physical investment, and is moved above office construction,

and the figure shown here. This AI boom isn't only happening on the stock market screen. We're talking about actual buildings, electrical systems, cooling equipment, semiconductors, infrastructure, and jobs. Bespoke also points out that mechanical, industrial, and electrical engineering were among the job categories with the largest increases posting this year, which lines up closely with the infrastructure needed

to build data centers. That's real economic activity. But here's the catch. Building something does not guarantee the investment will earn a good return. We've seen giant capital spending booms before. Some created enormous value, other created too much capacity.

So the question isn't whether companies are spending. They clearly are. It's whether the spending eventually produces enough revenue profit and productivity to justify the cost. That brings us to one of the biggest differences between today's AI boom and many of the companies at the center of the dot com mania, earnings. Toward the end of the dot com boom, FOMO, that's fear of missing out, became a powerful voice. Companies with very little revenue could reach enormous valuations. Some had no meaningful earnings. Some had no believable path to profits. We absolutely have speculation in parts of

today's market, too. And we shouldn't pretend otherwise. But many of the largest companies driving today's AI investment cycle were already highly profitable businesses. Ed Yardini has jokingly called today's environment FEMA. Fabulous earnings momentum. The name is catchy, but the numbers are what matter. S&P 500 earnings per share increased about 19% year-over-year in the first quarter of 2006. Reported second quarter growth was roughly 50.7%, but that headline number needs an important adjustment because marked-to-market investment gains boosted it. So go ahead. Let's strip out the effects of those gains, and underlying second quarter earnings growth was closer to 25%. That's still a very strong number. Analysts in source material a facts set we're expecting roughly 23.6%

earnings growth in this quarter, the third quarter, and 27.9% in the fourth quarter. Those are estimates, not guarantees, and they can change quickly. That's important because earnings are also one of the clearest ways this optimistic case could fail. Earnings estimates start falling, profit margins weaken, or AI spending grows much faster than the profits it creates. Today, highs valuations could be much harder to defend. So I'm not saying strong earnings make stocks safe. They don't. Markets can fall while earnings are growing. The point is that today's largest technology companies generally have a much stronger earnings foundation and many of the speculative businesses associated the late.com bubble. That makes the comparison.

much more complicated than simply saying technology stocks went up a lot. So this must be another March of 2000. Okay, now let's get to the part of this argument that could make even an optimist uncomfortable. There's no reason to dance around it. Erdini has discussed forward S&P 500 earnings of around 22 times. In the late 1990 technology bubble peak, those estimates were running 25 to 27 times depending on the measure. Some of the measures I've looked at in the first quarter of 2000 said it was over 30 times earnings. But valuation isn't the only warning sign. Bespoke points out that the dividend yield on the S&B 500 ETF has fallen below 1% for the first time since 1999 through 2001 period. The same time, the 10-year Treasury yield is approaching 4.8%. The visual here is important because

you're comparing the income available for stocks with the yield available from Treasury securities. At roughly 4.8%, Treasury yield creates a real competition for stocks that are expensive in some measures and offering a dividend yield below 1%. When interest rates stay high, investors may be also less willing to pay for very high prices for future earnings. So valuation could become a problem here. Interest rates could make them a problem. Slowing earnings could become a big problem. An AI spending that fails to generate strong returns could also become a problem. Revolutionary technology doesn't automatically mean every investment tied to that technology is a good investment at every price. The internet taught investors that lesson the hard way. Some internet companies went on to change the world, while many stocks still suffered enormous losses.

That's why the better question isn't simply, is AI real? Clearly it is. The harder question is, what price are investors paying for the growth they expect? There's one more clue underneath the surface, this market that makes today's picture more interesting. If this were becoming a pure, late-stage AI frenzy, you might expect market leadership to get narrower, with more money chasing fewer technology stocks. But recently, we've seen something very different. The NASDAQ 100 struggled after June. Semiconductor's corrected. Some AI infrastructure stocks peaked early in the summer, yet the broader S&P 500 was still up more than 12% in 2026 through this period covered by this analysis.

Healthcare led during parts of this summer. Financials gained over 12%. Energy was up roughly 43% year to date through September 3rd. Biotech rallied, value remained in an up trend, and as of mid-August, your Dini so-called impressive 493, that's the S&P 500 excluding the Magnificent 7, was up around 17.6% year-to-date compared to the 3.8% of the Magnificent 7. The point of this visual is market breadth. it shows that during this period, a larger group of S&P 500 companies was participating in the market's gains. It said the market depending on only seven largest technology-related companies.

But even this isn't a forecast. Breath can reverse, leadership can change, strong year-to-date performance tells us what has happened, not what happens in the future. Still, it challenges the simple idea that this market is being held up by only a handful of giant AI companies. So where does this leave us? It's today AI boom, another dot-com bubble. There are some similarities, and some of them are impossible to ignore. The transformative technology, huge capital spending, fast-rising stocks, high expectations,

all of this deserves respect. But saying this looks like the dot-com era is not the same thing as proving, we're already at March 2000. That's the distinction that matters. The bespoke Netscape versus ChatGBT chart raises the possibility of that, measured from the launch of a major consumer technology, today's cycle resembles an earlier point in the dot-com timeline. But it does not tell us what happens next. So instead of using 1998 as a prediction, I think it's more useful to use it as a test. The optimistic case is going to hold together. AI needs to produce measurable productivity gains. Earnings need to remain strong

enough to support investment. The physical buildout eventually needs to become productive economic returns. Credit conditions need to remain healthy, and market leadership needs to remain reasonably broad. If those things begin to break, the historical analogy becomes much less important. On the supportive side, we want to see productivity continue to improve earnings remaining strong, AI investment begin producing returns, and market breadth remain healthy. On the risk side, we'd be watching for productivity to disappoint, earnings estimates to roll over AI capital spending to outrun profits it creates, credit conditions to weaken, or interest rates to put more pressure on valuations. That's the real lesson from this entire comparison. Not that 2006 is 1998 or 2000,

not that stocks have to keep rising, and not that the risks of a bubble should be ignored. Historical analogies become dangerous when we start treating them like forecasts. The dot-com error reminds us that transformational technology and excessive speculation can exist at the same time. Both can be true. AI may dramatically change the economy, and some AI investments may still disappoint. The market may continue higher, and it may also experience major corrections long before the economic AI story is finished. So the next time somebody says, this looks like 2000, there's only one question worth asking, which part? 1995, 1998, March of 2000. The calendar can't answer that question, and neither can one chart. earnings, productivity, interest rates, credit, valuations, and actual business results will.

Same kind of movie, maybe, same scene. We don't know yet. And that's exactly what investors should be watching. If you're watching all of this and are wondering what a market like this means for your own retirement, that's a very different question from trying to guess whether NASDAQ is in 1998 or 2000. Your retirement plan has its own income needs, tax decisions, risk level, social security choices, withdrawal strategy. and timeline. Those decisions should be based on your own personal circumstances, not on historical market chart. If you'd like Oak Harvest Financial Group to take a look at your retirement plan

to help you understand the risks, tradeoffs, and decisions in front of you, call Oak Harvest Financial Group and start that conversation. All content contained with an Oak Harvest podcast expresses the views of the speaker and is for informational purposes only. It is based on information believed to be reliable when created, but any cited data, indicators, statistics, or other sources are not guaranteed. The views and opinions expressed herein may change without notice. Strategies and ideas discussed may not be right for you, and nothing in this podcast should be considered as personalized investment, tax or legal advice, or an offer or solicitation to buy or sell securities.

Indexes such as the S&P 500 are not available for direct investment, and your investment results may differ when compared to an index. Specific portfolio actions or strategies discussed will not apply to all client portfolios. Investing involves the risk of loss, and past performance is not indicative of future results.