Five Reasons the AI Boom Isn't in 'Dot-Com' Bubble Territory (Yet)

Are we in an AI bubble?

That's the question on a lot of investors' minds right now. Nothing seems to matter more given how much the generative-AI boom has helped drive the U.S. stock market in recent years.

Since OpenAI's ChatGPT launch in November 2022, the S&P 500 Index has returned around 70%. And Nvidia (NVDA), Microsoft (MSFT), Meta Platforms (META), and other mega-cap tech stocks have added more than $15 trillion in market value.

Sam Altman, founder of OpenAI, sounded a somewhat tepid warning in August saying that people might be getting “overexcited” about AI.

More recently, Mark Zuckerberg, founder and CEO of Meta Platforms, admitted that it's "quite possible" that we're in an AI-induced bubble, or at least headed for one.

Zuckerberg drew parallels between the current AI boom and other large infrastructure build-outs in the past that led to bubbles, such as the 19th century railroad mania and the late 1990s dot-com bubble. And he alluded to the hundreds of billions of dollars that Meta Platforms and other hyperscalers are throwing at data centers and other AI-related investments, with little regard for return on investment.

Indeed, there are striking similarities between the current environment and that of the late 1990s.

Judging from equity valuations, few investors are truly worried right now. But the big worry should be that we’re headed for a repeat of the dot-com bust.

The bad news is that there probably will be a bust after such unrestrained investment.

The good news, if you could call it that, is that the comparisons between the peak of "irrational exuberance" in the late 1990s and the current AI boom are misguided.

I think many people have forgotten just how bonkers the dot-com era was.

In the lead up to the bubble peak, people lost their minds...

It was a proper mania complete with unbelievably reckless trading, out of control price action, and beyond absurd valuations.

While things may seem crazy now, in many ways there’s no comparison to the situation in late 1999 and early 2000.

Here are five reasons why the dot-com bubble was bigger and badder than the current AI boom, and why we’re not in a mania phase… at least not yet.

No. 1: Dot-Com Era IPO 'Pops' Remain Unmatched

Nothing screams "FOMO" (fear of missing out) like a massive IPO return on the first day of trading...

Last March, Newsmax (NMAX) went public in a $75 million IPO. The offer price was $10 per share. Inexplicably, buyers were so aggressive that the stock closed its first day of trading above $83 – up more than 730%. The IPO set a record for its first-day "pop."

There have been other spectacular IPOs this year. For example, Figma (FIG) and Circle Internet (CRCL) posted first-day gains of 250% and 168%, respectively.

The dot-com bubble featured its own massive IPO pops. For example, on December 9, 1999, shares of computer server and workstation company VA Linux Systems were up nearly 700% on their first day of trading.

However, the dot-com IPO market was far more ludicrous than the one today.

In 1999, there were more than 470 U.S. common stock IPOs with an offer price of at least $5. The average and median first-day return from the IPO offer price to the close on the first day of trading were about 71% and 57%, respectively.

The momentum continued into 2000, when 380 companies went public. Shares popped an average of 56%.

There have only been 67 IPOs this year. And their average and median first-day gains were 35% and 14%, respectively.

While the new-issues market is certainly getting hot, the IPO fever isn't nearly as intense as it was during the dot-com bubble.

No. 2: The Dot-Com Bubble Had Crazier Stock Price Moves

In 1999, telecom company Qualcomm (QCOM) decided to shed businesses to focus on advanced wireless technology and chip design.

The market loved the strategy so much that Qualcomm's stock rocketed 2,600% higher that year, rising from a market cap of about $4 billion to more than $113 billion.

Moves like that helped power the Nasdaq 100 Index, which was the epicenter of the dot-com bubble. The Nasdaq 100 includes the 100 largest nonfinancial companies listed on the Nasdaq stock exchange. And it contained many of the high-flying tech and telecom stocks, like Qualcomm, that went parabolic during the late 1990s bull market.

From early 1995 to March 2000, the S&P 500 returned more than 260%. But the Nasdaq 100 soared by an amazing 1,080%.

Since the lows in 2020, the Nasdaq 100 is up about 240%. That's solid, but it's nothing compared with the late 1990s bull run.

The following chart shows the Nasdaq 100 since 1985.

It's a log chart, so equal distances on the y-axis have the same percentage changes. It also shows the long-term trend line and standard deviation channels.

The dot-com mania is clearly visible as the big outlier. All the other moves look tame in comparison.

No. 3: Index Valuations, When Measured Properly, Are Still Below Record Levels

You may have seen the scary headlines warning of sky-high valuations, like this one from the Wall Street Journal last month:

U.S. Stocks Are Now Pricier Than They Were in the Dot-Com Era

Here's another one from Yahoo! Finance:

This Warren Buffet indicator is bright red. Why it could be worse than the 1999 bubble – and how to prepare.

Then there was a research note from Torsten Slok, the chief economist at private-equity and alternative asset management firm Apollo Global Management:

AI Bubble Today Is Bigger Than the IT Bubble in the 1990s

It went on to assert that the top 10 companies in the S&P 500 today are more overvalued than they were back then.

I could go on and on.

It can be confusing for investors reading these warnings because there's no universally agreed-upon way to value the broader stock market. And every valuation metric has flaws.

Take the price-to-sales (P/S) ratio. Put simply, it compares the market value of a stock or index with annual sales (revenues).

Currently, P/S shows the market to be far more expensive than it was in 2000. Yet, the P/S ratio is one of the worst metrics for valuing the S&P 500 because it's blind to margins and profitability.

Other metrics, such as market cap to gross domestic product (or gross national product) and Tobin's Q have their own fatal flaws. The cyclically adjusted price-to-earnings ("CAPE") ratio, one of the most popular valuation metrics, has faults, too. The CAPE ratio is the real (inflation adjusted) index value divided by the 10-year average of real earnings for the index. It smooths out the variation in earnings over the business cycle.

The CAPE ratio peaked at more than 44 in 1999. It's currently around 40, suggesting this is the second-highest valuation the market has ever attained.

Faith in the CAPE ratio has waned because there has been a structural shift in the indicator only seen with the benefit of hindsight. The CAPE ratio's average up to 1990 was 14.6. Ever since, the average has been 27. And it has been above 27 since 2016.

Valuation metrics can be useful tools. But if the tool you followed made you bearish since 2016, then how was the tool of use?

It's nice to smooth out earnings volatility. But averaging 10 years of trailing earnings is like staring in the rearview mirror with binoculars.

The S&P 500 is now dominated by Big Tech companies. They're highly profitable and growing quickly.

Nvidia, Microsoft, Apple (AAPL), Alphabet, Amazon (AMZN), Meta Platforms, Broadcom (AVGO), and Oracle (ORCL) now account for about 35% of the S&P 500's market cap.

Aggregate earnings for these companies have grown from about $90 billion a decade ago to around $560 billion today – a more than sixfold increase.

Not only that, but these Big Tech companies are also expected to earn $670 billion over the next four quarters. Perhaps these estimates are a bit too high. But the growth has been undeniable and is one reason why the CAPE may be overstating current valuations.

Arguably, the trailing price-to-earnings (P/E) ratio, based on the last four quarters of earnings, and the forward P/E, based on consensus analyst estimates, are better.

These metrics have their flaws as well. But we can still use them to make a reasonable comparison between periods.

Here's the forward P/E for the S&P 500 since 1995:

It's currently at 23, which is high but not quite as high as the dot-com-bubble peak of around 26.

No. 4: The Largest Companies in the S&P 500 Are Still Cheaper

Microsoft led the bull market in the late 1990s...

Its stock peaked in December 1999, three months before the broader market. But the company remained the largest in the S&P 500, with a roughly $560 billion market cap when the index topped on March 24, 2000.

On that day, Microsoft's trailing P/E and forward P/E were both above 60. But that was nothing compared with the second-largest company, Cisco Systems (CSCO), which had a forward P/E of 137. Oracle was another top-10 stock in the S&P 500 with a triple-digit P/E of around 118.

The table below shows the full list of the top 10 stocks on March 24, 2000.

The average forward P/E was 53, although that was dragged up by the triple-digit outliers. The median forward P/E was about 35.

I've also included P/E ratios that treat these stocks as a 10-member market-cap-weighted index, called S&P 10. (Note, this is not the same as a market-cap-weighted average of the P/E ratios.)

The S&P 10's forward P/E ratio was nearly 39 back then. And I think that's the most relevant number.

The table below shows the same ratios and statistics for the largest stocks in S&P 500 as of September 22, the day of the highest index close as I write.

There's only one set of triple-digit P/Es, and that's Tesla (TSLA). Microsoft is the only company that appears on both lists, and its forward P/E is about half of what it was in March 2000.

Every summary statistic in the AI boom table is below its corresponding value in the dot-com bubble table.

The chart below compares the S&P 10 P/Es.

This clearly shows that the largest stocks in the S&P 500 are still cheaper now than they were at the height of the dot-com bubble. (This is also true using free-cash-flow yield, a valuation method that deserves its own write-up.)

The largest companies today might be expensive, but they're not record expensive. I know that's hardly comforting, but its further evidence that we're not at mania levels.

I will also concede that the S&P 500 is more concentrated than ever. The top 10 stocks in the index accounted for around 25% of total index market cap at the peak in 2000. Today, they total roughly 40%.

No. 5: The 'P/E Above 100' Club in the Dot-Com Era Was Larger

The Nasdaq 100 was completely devastated as the dot-com bubble burst. From its peak in March 2000 to the end of 2002, it fell as much as 83%.

That's why headlines today warning about record valuations catch people's attention. It suggests that a similar outcome is possible, or even likely.

Yet, the Nasdaq 100 reached a truly absurd valuation in 2000. Its trailing P/E reached nearly 100.

On the bright side, the largest stocks in the Nasdaq aren't nearly as expensive as they were during the dot-com bubble. The Nasdaq 100's trailing P/E is currently about 34. On the other hand, the index is also well above the 10-year average P/E of about 27.

A triple-digit P/E is a nosebleed valuation. Only companies with great businesses that are growing very quickly deserve a P/E that high. And even then, those stocks will probably be decimated from time to time as they grow into their valuations.

To look at stocks with a P/E above 100, I will broaden my universe out to the top 1,000 largest U.S. companies by market cap. (Note, this is not the Russell 1000 Index. It includes companies that aren't in any index.)

Among the largest 1,000 stocks, there are currently 29 with a forward P/E greater than 100. Their median P/E is 137.5.

I already mentioned Tesla. The "P/E above 100" club also includes highfliers like Palantir Technologies (PLTR), Cloudflare (NET), Snowflake (SNOW), Roku (ROKU), CrowdStrike (CRWD), Figma, and Circle Internet.

Not to be outdone, the dot-com bubble had its own list of crazy valuations, such as Cisco, which I mentioned earlier. Stocks with forward P/Es greater than 100 in early 2000 also included Broadcom, VeriSign (VRSN), eBay (EBAY), Qualcomm, Time Warner, and Oracle.

On March 24, 2000, there were 62 stocks with forward P/Es greater than 100. And their median P/E was nearly 247.

So, there were more than twice as many egregiously valued large- and mid-cap stocks during the dot-com bubble as there are today. And their valuations were much more extreme back then.

The Verdict: We're a Far Cry from Peak 'Dot-Com' Territory

So, there you have it... The dot-com bubble featured a hotter IPO market and more shocking price moves. It was more broadly expensive than the current market. That's also true for the top 10 stocks. And tech stocks were far more expensive in early 2000 than they are now. Plus, the number of stocks with nosebleed valuations was far higher during the dot-com bubble.

Don't get me wrong... Things are getting crazy.

The stock market has gotten extremely expensive. Just because it's not at a record valuation doesn't mean we shouldn't be worried. At the very least, we should expect relatively low, long-term returns from here.

There are also signs of excessive speculation everywhere, including soaring options volumes, ballooning leveraged exchange-traded fund assets, meme stock (and coin) pumping, pre-revenue companies attaining multibillion-dollar valuations, and quantum-computing stocks going nuts.

By all means, be disciplined and follow your investing strategy. Make sure your portfolio is diversified. Hold some defensive stocks. Sell stocks that have unrealistic expectations built into their valuations. Hold some cash.

But whatever you do, don't claim that this market is crazier than the peak of the dot-com bubble.

Good investing,

Alan Gula, CFA