The AI Bubble Is Growing Faster Than the Dot-Com Boom

The servers are humming more loudly than before. When you drive by a new AI data center in northern Virginia or outside of Phoenix, you’ll notice windowless structures rising from dusty lots, encircled by humming transformers and chain-link fences. Inside, cooling systems roar, electricity meters spin, and racks of chips glow in fluorescent light. It has an industrial feel that the internet didn’t have. Material. Pricey.

And the word “expensive” is appropriate. A small number of AI startups have increased their market value by almost $1 trillion in the last year alone, many of them without making a profit. The profits of Nvidia have increased. Capital expenditures promised by Big Tech are comparable to the GDP of small nations. It appears that investors think AI will change everything, including how your refrigerator orders groceries and how legal research and drug discovery are conducted.

Perhaps they are correct. However, it’s also possible that we have already seen this film.

The internet seemed unstoppable in the middle of the 1990s. The 1995 Netscape IPO energized markets. Businesses that only had a website and a memorable name added “.com” to their brands, and their stock prices skyrocketed. The routers and switches that drove the growth of the web were sold by Cisco. Investors believed they were on the verge of something huge.

Indeed, they were. They weren’t, either. From 1995 to its peak in 2000, the Nasdaq increased by almost 400%. Then it fell apart. Pets.com turned into a joke. Years of gains were lost by Cisco. Microsoft needed over ten years to reach its previous peak. The AI boom of today is comparable, albeit quicker.

This market is evolving five times faster than the internet era, according to a recent statement made by former Cisco CEO John Chambers. Observing 2026 earnings calls, that doesn’t seem overly dramatic. A company loses $100 billion in market value in one afternoon one quarter after making a mistake with an AI product demo. After announcing a new model release, it bounces back the next day.

There isn’t any brewing volatility. It’s a built-in feature. The unnerving thing is the scale. In absolute monetary terms, analysts contend that the current AI bubble, if it is one, is several times bigger than the dot-com boom. Companies like Amazon and Alphabet now have capital spending plans that total hundreds of billions of dollars. The number of data centers is increasing. Power companies are competing to supply the anticipated demand for electricity.

The expenditure itself serves as a rationale for further expenditure.

At the heart of it all is Nvidia, which sells chips as quickly as they can be produced. Analysts say there are “zero signs of slowing” in demand. However, recent stock reactions have been muted even by blowout earnings. That slight change—strong outcomes met with hesitancy—feels familiar.

Businesses that announced ambitious growth plans in 1999 received rewards. The market demanded proof by 2000.

This time, there is also a structural difference. The companies that finance AI infrastructure—Google, Amazon, and Microsoft—make a ton of money. They are not brittle startups that are consuming a lot of venture capital. Because of this, the current boom has a stronger foundation than the dot-com era.

However, thousands of AI startups are vying for attention beneath those titans.

Founders are pitching “AI-powered” versions of everything on LinkedIn, including scheduling bots, marketing tools, and legal assistants. Some people are smart. A lot of them are derivative. It’s still unclear if the majority of these companies are merely riding investor fervor or if they will ever turn a profit.

Large language models might find it difficult to generate a real “killer app” that can support current valuations, according to a skeptic analyst who recently made this claim. The costs of scaling are enormous. The cost of improving models increases exponentially. Revenue models are still unclear.

But there is still hope. The cultural momentum is hard to ignore. Research labs are no longer the only places where AI exists. It is used by students to write essay drafts. Attorneys test it for reviewing documents. It is used by programmers to autocomplete software. Over half of all internet traffic is now generated by automated bots. The technology is authentic. The adoption is genuine.

Whether revenue growth will keep pace with the influx of capital is the question.

Telecom firms constructed extensive fiber networks during the dot-com era in anticipation of the unending demand. Overcapacity crushed balance sheets as demand slowed. Hyperscalers are currently constructing data centers at a dizzying rate. In retrospect, those expenditures might appear excessive if enterprise AI adoption stagnates or margins contract.

Everyone seems to be aware of the risk and moving forward despite it.

AI may be in a bubble, as Sam Altman himself has admitted. At the height of a manic episode, such openness is uncommon. However, recognizing a bubble does not cause it to burst. In fact, it occasionally reinforces the idea that things are different now—more sophisticated and grounded.

AI businesses are already making a sizable profit, unlike in the late 1990s. The profits made by Nvidia are real. AI-related cloud services are rapidly becoming profitable. Domain names are not the only tangible products available.

Expectations are still very high. Years of faultless performance are reflected in stock prices. Capital markets anticipate that the productivity gains from AI will materialize without any problems with public trust, energy bottlenecks, or regulatory backlash.

It’s a big wager. Fraud is not necessary for market crashes. All they need to outrun reality is expectations.

It’s hard to avoid feeling both awed and uneasy when passing one of those new data centers and listening to the continuous mechanical drone. The infrastructure under construction is enormous. The level of ambition is astounding. The amount of money at risk is enormous.

Perhaps we are in 1996, with a long way to go before any reckoning. Perhaps the music is already playing too loudly, and it’s 1999.

More quickly than the dot-com boom, the AI bubble is expanding. It seems indisputable. It remains to be seen if it results in long-term dominance or train wrecks. One thing is certain, though: corrections, if they occur, happen just as fast when capital is moving at this rate.