The AI ​​boom is eerily similar to the dot-com bust of the 2000s, but there are some important differences.

If last week’s multitrillion-dollar decline in major tech stocks sounded familiar, that’s because we were here before, the last time the hype about innovation ran headlong into economic reality.

As the market slumps due to investor concerns about soaring valuations for artificial intelligence (AI) companies, commentators are asking the same questions they asked during the dot-com crash 25 years ago.

Can technology really override basic economics?

This is an issue I discussed in my first teaching lecture at the University of Otago in August 2000. This was around the time when Internet stocks were crashing and hundreds of dot-coms were going bankrupt.

At the time, I argued that many Internet companies are “naked” because their business models are obvious to everyone. They spent huge amounts of money trying to attract customers, but had no surefire path to profit.

A generation later, the same logic is driving the AI ​​boom.

Same story across different metrics

In 2000, the Internet revolutionized commerce, promising that success would be measured in “attentions” and “clicks” rather than profits. These metrics are now “Tokens Processed” and “Model Queries.”

The language may have changed, but the belief that scale automatically translates to profit remains the same.

Just as we heard that the internet would eliminate middlemen and traditional intermediaries like retailers and brokers, there was a promise that AI would eliminate cognitive labor.

Both have encouraged investors to forgo losses in pursuit of long-term advantage.

At the height of the dot-com craze, companies such as online retailer eToys spent lavishly on marketing to attract customers. Currently, AI developers are investing billions of dollars in computing power, data, and energy, yet profitability remains low.

Nvidia’s multi-trillion dollar valuation, OpenAI continuing to lose money despite soaring revenues, and the influx of venture capital into AI startups all reflect the 1999 bubble.

Then, as now, spending will be mistaken for investment.

What the dot com crash has to teach us

In 2000, I proposed that Internet companies were building market-based assets such as brand value, customer relationships, and data that could only create real value if they created loyal and profitable customers.

The problem was that investors treated spending as evidence of growth and marketing as its own business model.

The AI ​​economy repeats this pattern.

Datasets, model architectures, and user ecosystems are treated as assets even if they are not yet generating positive returns.

Its value is based on the belief that monetization will eventually catch up with costs. The logic remains the same. It’s just that the story has changed.

The dot-com boom was driven by fragile startups backed by venture capital and public enthusiasm.

Today’s AI explosion is being led by powerful incumbents like Microsoft, Google, Amazon, and Nvidia, who may have years of losses in their pursuit of advantage. This reduces systemic risk but concentrates market power.

The destination of money has also changed. Internet companies used to spend cash on advertising. AI companies consume computing power and data.

Spending has shifted from marketing agencies to data centers, but the question still remains. Does it create real value or is it just an illusion of progress?

AI reaches even deeper than the internet. The web has changed the way we communicate and shop, and AI is shaping the way we think, learn, and make decisions.

If a crash were to occur, public trust in the technology itself would erode, potentially delaying innovation for years. Relatively low real interest rates and abundant capital have also helped fuel the current wave of technology investment.

Similar to the boom of the late 1990s, when favorable monetary policy supported rising valuations for high-tech companies, this cycle shows how macro-financial contexts can amplify technology optimism.

The revival of intangible mania

Despite these differences, the pattern of evaluation is well known. Investors are once again focusing on potential over performance.

In 2000, analysts justified their valuations by counting the number of users a company might someday monetize. In 2025, they will model “inference demand” and “data dominance.” Both are speculations about an imagined future.

Stories have become capital because markets reward belief over evidence. The danger is not technical failure, but economic distortion when storytelling outweighs ability to pay.

Even profitable companies can be caught in a downdraft.

In 2000, leaders like Yahoo! and eBay, despite long-term survival, lost most of their market value when the bubble burst. The same could happen with today’s AI giants.

There are still two lessons left. First, scalability without profitability is not a business model. Rather than mitigating losses, exponential growth can make them worse.

Each additional AI query has a real computational cost, so growth only matters if it leads to sustainable profits.

Second, intangible assets must create measurable value. Marketing, data, and algorithms are assets only if they generate sustainable cash flows and clear social benefits.

For policymakers, the implications are clear. It means funding AI projects that deliver tangible productivity and social benefits, rather than just fueling the hype.

AI will transform the way we work and think, but it will not erase the relationship between cost, value, and customer needs. Lasting value comes from providing real benefits to people.

The question now is whether the real productivity gains from AI will ultimately justify today’s valuations, just as the Internet ultimately did after painful revisions.

This article is republished from The Conversation under a Creative Commons license. Read the original article.