In March 2000, a company that made routers was briefly the most valuable business on the planet. Cisco built the boxes that moved internet traffic, and investors were sure the internet would need endless amounts of them. They were right about the internet. They were wrong about the price. Within two and a half years the stock had lost nearly 90% of its value, and it did not close above that March 2000 record of $80.06 again until December 2025, more than 25 years later.

That story is being retold a lot right now, because the AI boom rhymes with it in uncomfortable ways. A widely shared YouTube breakdown, AI Bubble vs Dot-Com Crash, lays out the case well. We checked its main arguments against current figures. Some hold up strongly. A few need context. And the comparison breaks down in places that matter.

The spending is bigger than anything the dot-com era saw

In the late 1990s, telecom companies laid fibre optic cable for a level of traffic that would not arrive for years. The demand was real, just early, and the companies that borrowed to build it went bust waiting.

Today the build-out is concentrated in four companies. Amazon, Microsoft, Alphabet and Meta are on track to spend roughly $700 billion to $725 billion on capital expenditure in 2026, according to Forbes, up about 75% from around $410 billion last year. Most of it is going into data centres, chips and power for AI. That spending is expected to eat up nearly all of their operating cash flow this year, and some of it is now being funded with debt.

The bet is that demand for AI keeps growing fast enough to fill all that capacity. There are early signs of real revenue. Microsoft says its AI business has passed a $37 billion annual run rate, and Google Cloud reports a backlog of more than $460 billion. But the gap between what is being spent and what AI currently earns is still wide.

Nvidia is today’s Cisco, with one big difference

If Cisco sold the picks and shovels of the internet gold rush, Nvidia sells them for AI. Almost every major AI model is trained and run on its chips, and data centre sales now make up the vast majority of its revenue.

The difference is valuation. At its peak, Cisco traded at well over 100 times its expected earnings. Nvidia, for all the excitement, is priced far more modestly, with recent estimates putting it somewhere between the high teens and the mid twenties times next year’s expected earnings. It also earns far more on each sale than Cisco did, with gross margins around 75%.

The similarity is dependence. Cisco’s sales collapsed when the startups and telecoms buying its gear ran out of money. Nvidia relies heavily on a small group of very large customers. If a handful of them slow their spending, or shift to their own custom chips, as several are trying to do, Nvidia feels it directly.

Money that goes round in circles

This is the argument that gets the most attention. Money in the AI industry often moves in loops. A chipmaker or cloud company invests in an AI lab. The lab then spends that money buying chips or computing power from the same investor or its partners.

OpenAI is the clearest example. In its funding round this February, Nvidia invested $30 billion and Amazon committed $50 billion, both companies that also sell it the chips and cloud capacity it runs on. Nvidia has also agreed to guarantee up to $105 billion in financing for data centres that will be filled with its own chips for OpenAI, the Motley Fool reports. Separately, OpenAI has a $300 billion data centre deal with Oracle, which in turn buys Nvidia hardware.

Critics say this can make demand look stronger than it is, because the same money shows up as growth for more than one company. Not everyone agrees it is that sinister. Economist Noah Smith argues it looks more like old-fashioned vendor financing, where a supplier helps a customer buy its products, than fake revenue. His point is subtler: these deals tie everyone’s fate together. The system becomes less exposed to any single company failing, and more exposed to the whole AI trade going wrong at once.

Who is selling, and who is holding

The video also points to top executives, including Jensen Huang, Jeff Bezos and Mark Zuckerberg, selling large amounts of stock near record highs. That is true, but worth reading carefully. These sales mostly happen through plans filed months in advance, and founders whose wealth sits almost entirely in one company sell regularly to diversify. On its own, it is a weak warning signal.

The stronger point is about concentration. At the end of 2025, the ten largest companies made up about 41% of the S&P 500, according to RBC Wealth Management. At the height of the dot-com bubble, the figure was around 27%. Anyone holding a simple US index fund today is making a much larger bet on a handful of AI-linked companies than they may realise.

Where the comparison falls apart

The dot-com bubble was full of companies with no profits and sometimes no real business. Pets.com is the famous example. The companies leading the AI boom are the opposite. Google, Microsoft, Meta, Amazon and Apple run enormously profitable businesses in search, advertising, cloud, retail and devices. If AI disappoints, they can cut spending and carry on. Most of them will not disappear the way so many dot-com names did.

But a company surviving and its share price doing well are two different things. Cisco never came close to going under. It stayed profitable and kept growing. People who bought its shares at the top in 2000 still waited a quarter of a century to get their money back. That is the real lesson of the comparison: the risk is less about AI failing, and more about paying too much for its success.

What this means for India

India has an unusual place in this story. It has sat largely outside the AI rally. Only about 16% of MSCI India’s market value is exposed to AI, compared with 70% to 80% in Korea and Taiwan, Business Today reports. That has hurt over the past year, with the Nifty and Sensex down around 11% while those AI-heavy markets surged. Some analysts now describe India as the market’s “anti-AI trade,” and Motilal Oswal has said a meaningful slowdown in AI could push global money back towards emerging markets like India.

Closer to home, many Indian investors have added US exposure in recent years through Nasdaq and S&P 500 funds of funds, or by buying US shares directly. Those holdings are, in practice, heavily tilted towards the same few AI giants. That is not a reason to panic, but it is worth knowing what you own.

Nobody can say whether the AI boom is a bubble that pops next year, deflates slowly, or grows into its price the way the internet eventually did. History suggests the technology can be real and the valuations can still be wrong. Both things were true in 2000, and both may be true again.

This article is for general information and is not investment advice.

Sources