Tech History Repeats: The Structural Parallels Between AI and the Dot-Com Boom

Every financial expert remembers the dot-com bubble of the late 1990s. Back then, the internet sent the valuations of U.S. tech stocks skyrocketing, as investors of all kinds rushed to buy into “super-profitable” companies that promised to revolutionize everything. The result was that many of these firms were wildly overvalued and, when the bubble burst, hundreds went bankrupt while the survivors saw their valuations plummet.

So, why does this matter now?

We are witnessing a similar phenomenon, this time in artificial intelligence. While AI undoubtedly has far greater useful applications than the early internet, many of the same patterns of circular investment and related-party transactions are reappearing in today’s market. In 1999, AOL was selling advertising services to companies it had invested in. In 2024, NVIDIA sells GPUs to companies in which it has invested. The names have changed, but the playbook feels eerily familiar. Which raises the question: Is AI’s exponential growth truly organic, or is it the result of financial engineering between the same handful of players?

More than two decades after the dot-com bubble, the new shiny object in tech is AI. To be clear, this is not an apples-to-apples comparison. While the investment patterns in AI seem a bit alarming, AI’s real-world potential far exceeds that of the early internet. Still, looking back at historical parallels can help us understand where this might all be heading.

So how did AI become so big, so fast? The turning point was arguably OpenAI’s launch of ChatGPT in November 2022, the first AI product to truly capture the public’s imagination. Today, ChatGPT has roughly 700 million active users per week, making it one of the fastest-growing consumer technologies in history. At the same time, the COVID-era GPU shortage drove demand for high-performance electronics, sending waves of capital toward big tech firms. As the AI hype intensified, the so-called “Magnificent Seven” — Microsoft, Apple, NVIDIA, Meta, Amazon, Google, and Tesla — poured hundreds of billions into AI startups and infrastructure. A few standout deals illustrate the frenzy: Microsoft’s $13 billion investment in OpenAI, OpenAI’s $300 billion cloud deal with Oracle, and NVIDIA’s $100 billion investment in OpenAI. These transactions do not just signal optimism, they also hint at deep interconnections within the AI ecosystem, where everyone seems to own a piece of everyone else.

So, where exactly is the comparison to the dot-com era? In the late 1990s, one of the biggest reasons the tech bubble inflated, and eventually burst, was a form of financial engineering known as circular investment. This occurs when companies invest in or finance entities that later become their customers, creating a false sense of demand and inflated earnings. Telecom companies back then even engaged in capacity swaps, exchanging the right to use each other’s network capacity without actually exchanging cash. A famous example was Global Crossing, which repeatedly inflated its revenue through these swaps before collapsing into bankruptcy in 2002.

Fast-forward to today: AI companies are not swapping bandwidth, but they are recycling capital in ways that distort market signals. For instance, NVIDIA reportedly plans to invest $100 billion in OpenAI, with OpenAI in turn using NVIDIA’s chips to train its newest models — a near-perfect loop of money, investment, and demand.

This kind of financial engineering creates an illusion of growth: each party in the loop can record the others’ spending as genuine demand, booking the resulting sales and investments as real revenue or assets. The danger lies in the fact that these figures may not reflect true market demand, but rather self-reinforcing accounting between the same few players. If one link in this chain falters — say, OpenAI cuts back on GPU purchases or fails to raise new capital — NVIDIA’s projected demand could evaporate overnight, sending ripples through the entire system. It is the same structural fragility that fueled past bubbles: inflated earnings on paper masking a lack of independent demand beneath the surface.

You might ask, “If something is really off, would regulators like the SEC not have stepped in by now?” Unlike 1999, AI products today do have genuine real-world utility. Chatbots, copilots, and image generators have become household tools, making it much harder for regulators to separate legitimate growth from inflated valuations. Moreover, as long as these companies benefit collectively from their intertwined investments, there is little incentive to stop. Each new partnership props up the others, sustaining the illusion of endless demand — at least until the music stops.

And when the hype dies down, what happens next is what we call a bubble bursting: liquidity dries up, credit contracts, and inflated earnings unwind as the closed loop breaks. The dot-com bubble of 2000 burst when investor confidence evaporated after years of hype-driven equity issuance and cross-investment that never translated into sustainable profits. In AI, a comparable scenario could play out if capital inflows slow and the illusion of infinite compute demand collapses, leaving the industry’s valuations drastically exposed.

The risks here are far from theoretical. For publicly traded firms like NVIDIA or Oracle, these opaque financial relationships may expose retail investors to risks they do not fully understand. If the true nature of these transactions were widely known, some might think twice before buying in. The deeper these companies intertwine, the more vulnerable they become to a domino effect. If one major player stumbles, the shock could ripple through the entire ecosystem. And finally, there is liquidity risk: should financing conditions tighten, the same “hot potato” cash that has been circulating among AI firms could vanish overnight — exposing just how much of this demand was artificially created.

The dot-com bubble taught us that hype can lift a stock, but without genuine financial substance, it only takes one slip for that company to become a cautionary tale in a business textbook. The AI boom shares many similarities with that era, yet unlike the internet of the 1990s, AI has already proven its real-world value. The challenge lies in distinguishing sustainable innovation from excessive speculation.

History does not repeat itself in finance, but it does rhyme. So, if you choose to invest in AI today, ask yourself: are you funding the future, or just buying the hype?