Seven Ways the AI boom Is Different From 1929 and 1999

Since history tends to rhyme, market forecasters have been having a field day comparing the AI buildout with 1929 and 1999.

Today, we’re going to point out seven useful distinctions, starting with the data point that triggered today’s observations.

In September, the S&P 500 Index bucked its historical trend of trading lower for the month with a modest 0.5% gain.

But it did so with 78% of non-AI stocks in the index declining – a healthy stealth correction behind the scenes.

On only two occasions has the stock market hit record highs when twice as many stocks are declining as those advancing: 1929 and 1999.

So… that’s today’s comparison.

Rather than add to the cacophony of worry, we thought we’d try to identify the most useful distinctions and trade accordingly.

When it comes down to it, hindsight being as perfect as any sight, 1929 was a margin-and-credit boom in radio and electrification stocks; 1999 was an equity-financed internet-and-telecom boom; the AI boom has features equivalent to both.

All three had – or have – a technological revolution at their center.

The AI buildout is distinct as it’s an industrial capex boom financed through corporate cash flow, debt, private credit, vendor financing and public-market enthusiasm.

Here are seven ways those distinctions matter to the individual investor trying to manage his own money today:

1. The AI buildout is more physically demanding than the dot-com boom.

The dot-com boom needed servers, routers, fiber and office space. Telecom executives buried fiber all over the country, and Wall Street funded too much capacity too quickly.

AI needs something heavier: chips, data centers, cooling systems, gas turbines, transformers, transmission lines, backup generation, water, land and long-term power contracts.

Goldman estimates global AI investment will exceed $1 trillion in 2026, including $581 billion in the U.S., and that U.S. AI capex could rise from 1.8% of gross domestic product (GDP) in 2026 to 2.8% in 2028.

That gives AI a more industrial character. The internet boom could be financed with stories, stock issuance and fiber. AI executives need electrons by the gigawatt.

2. The AI leaders are richer than the dot-com leaders.

Many dot-com companies had no profits, thin revenues and business plans written in incense smoke. A company could add “.com” to its name and receive a market multiple fit for a railroad baron.

The AI boom has stronger anchor tenants. Microsoft (MSFT), Alphabet (GOOGL), Amazon (AMZN), Meta Platforms (META), Oracle (ORCL), Nvidia (NVDA), Broadcom (AVGO), Advanced Micro Devices (AMD) and Micron Technology (MU) are real companies with revenue, customers, engineers and balance sheets. That makes the boom sturdier than 1999 in its first phase.

But sturdier does not mean safer at any price. Reuters reported that Goldman expects the largest U.S. hyperscalers to spend about $800 billion on capex in 2026, with consensus expecting $1.1 trillion in 2027. When strong companies spend at that scale, investors must judge the return on capital, not just the technology.

3. The weak point in AI is financing capacity.

The dot-com boom ended when investors stopped funding companies that could not turn traffic into profits.

The AI boom will likely weaken as investors question whether the next dollar of capex will generate enough revenue, productivity or pricing power to justify the financing. That pressure will show up in debt markets, credit spreads, free cash flow, power constraints and shareholder resistance to dilution.

Reuters reported that Morgan Stanley estimates AI infrastructure will require $1.5 trillion in external financing by 2028, while Nvidia has partnered with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR on financing platforms intended to mobilize more than $500 billion for AI infrastructure. That is the giveaway. Nvidia is not only selling picks and shovels. Nvidia and Wall Street are helping finance the miners.

4. The power constraint separates AI from 1999.

Dot-com investors worried about eyeballs, clicks, fiber glut and burn rates.

AI investors must also worry about the grid.

The EIA projects U.S. power consumption will rise from 4,195 billion kWh in 2025 to 4,288 billion kWh in 2026 and 4,356 billion kWh in 2027, driven by data centers and electrification. That turns AI into a utility, energy, construction and permitting story.

This is why the AI boom can spill into natural gas, turbines, copper, uranium, grid equipment, backup generators and land. The software story has a smokestack now.

5. 1929 was driven more directly by stock-market leverage.

The 1929 boom had new consumer technologies, radio stocks, automobiles, electrification and a belief in a permanently higher plateau. But the signature financial feature was stock market leverage. The Federal Reserve’s history notes that brokerage houses, investment trusts and margin accounts allowed ordinary people to buy stocks with borrowed money.

The Fed’s own research says broker loans and NYSE market capitalization both more than doubled between 1926 and the 1929 peak. After the crash, Congress gave the Fed power to set margin requirements partly to prevent another credit-backed stock-market bubble.

AI has leverage too, but that leverage sits more within corporate balance sheets, private credit structures, vendor financing, data-center deals, and power commitments. The danger is less “retail investors bought stocks on 10% margin,” and more “corporate executives signed commitments that assume heroic future demand.”

6. The policy backdrop is different.

In 1929, the Fed was young, the safety net was thin, deposit insurance did not yet exist, and policymakers were still learning what a modern financial crash could do.

In 1999, Alan Greenspan’s Fed had cut rates after the LTCM crisis and Y2K fears, and investors believed central bankers would cushion market accidents. Then the Fed tightened into 2000, and the NASDAQ cracked.

In 2026, investors expect central bankers and Treasury officials to react quickly to financial stress. That expectation can extend the boom. It can also encourage investors to take more risk because they assume officials will arrive with liquidity before the bodies cool.

That is where Empire of Debt still earns its keep. When investors come to believe policymakers will underwrite every large boom, they stop treating risk as a cost and start treating it as a public utility.

7. The AI boom has real productivity earlier in the cycle.

The internet eventually transformed commerce, media, advertising, logistics and software. But in 1999, much of that payoff still sat beyond the horizon.

AI is already changing code writing, customer service, drug discovery, legal work, accounting, marketing, logistics and semiconductor design. That gives today’s boom a stronger business case than many dot-com stories had.

The catch is timing. A technology can be real, useful and revolutionary while investors still overpay for the companies building it. Railroads were real. Radio was real. The internet was real. The question for investors is always the same: who earns the return after the financing, depreciation, competition and debt service take their bite?

The 8th bonus similarity: the market’s “bad breadth.”

The uncomfortable resemblance across 1929, 1999 and today is the source of today’s chart.

In 1929 and 1999, the headline indexes masked weakening participation. Today, a small group of AI-linked leaders can keep the S&P 500 near records while most stocks quietly fall away. That does not guarantee the same ending. It does tell investors that the market is relying on fewer horses to pull a heavier wagon.

When breadth narrows, the index becomes more fragile. The leaders have to keep leading. If Nvidia, Microsoft, Meta, Alphabet, Amazon, Oracle or the chipmakers stumble, the index has fewer backups.

In the end, for investors chasing AI stocks, the balance-sheet test will matter more than the story.

If the dot-com boom asked, “Can these companies get customers and make money?”

The AI buildout asks an even harder question: “Can these companies earn enough return on enormous fixed investment before financing costs, depreciation and competition catch up?”

The core distinguishing feature: AI is a technology bubble story on CNBC. On the balance sheet, it is a capital-intensive industrial cycle.

The clean comparison is important.

1929: stock-market leverage, investment trusts, broker loans, speculative margin buying, weak policy backstops.

1999: equity issuance, venture capital, telecom overbuild, internet euphoria, unprofitable companies with revolutionary language.

AI buildout: profitable giants, real productivity gains, trillion-dollar capex, power shortages, private credit, corporate debt, vendor financing and a race to prove that the cash flows arrive before the capital bill overwhelms the story.

Our Grey Swan forecast follows naturally.

AI can run longer than skeptics expect because the leading companies are real and the productivity gains are already visible.

We suspect the announcement on August 10, 2026, that Nvidia had reached an agreement with BlackRock and four other leading Wall Street banks to turn “AI compute” into an “investible asset class” will be the turning point for financial historians.

The final showdown will not arrive with trumpets and fanfare. It will arrive first in credit spreads, reduced capex guidance, delayed data-center projects, failed power deals, equity dilution, and CFOs discovering religion after the bonds start trading badly.

You’ll recall that Oracle’s credit spreads, which we discussed last week, are already historically out of whack.

For now, the party in a few hyperscalers rages on: laissez les bons temps rouler!

Today, Andrew’s Pro trade will help you partake… until the fat lady starts belting out a new tune.

~ Addison

P.S. Last week’s Grey Swan Live! with Shad Marquitz was a masterclass in what to look for in resource investing.

We shared our insights from last week’s trip to Colorado to interview mining executives, as well as what we’re seeing in the resource space now. The replay is up on the site here. Whether you’re a seasoned resource investor or not, you’ll get some valuable information out of this livestream.

What happens when the systems we’ve spent the last century building stop working the way they’re supposed to?

John Robb has spent decades studying exactly that question — and on Thursday at 2 p.m. ET for Grey Swan Live!, he’s going to explain why he thinks we’re watching the beginning of the “end of mass.”

From AI and politics to drones and the future of warfare, this conversation could completely change the way you look at the next decade.