In the days when artificial intelligence (AI) is becoming the brightest star in the financial market, a new report from the analysis firm MacroStrategy Partnership has startled global investors: the AI bubble is now 17 times larger than the DOT-com bubble that shook Wall Street in the early 2000s.
Analyst Julien Garran, the author of this opinion, likens the current AI phenomenon to a modern-day “gold rush” – but instead of gold, what is sought after is silicon, data and the illusion of limitless productivity.
According to Garran, the “17x” figure does not simply reflect the stock price, but the enormous scale of Capital pouring into the AI ecosystem – from chipmakers, data centers, software, to the thousands of new startups that are born every month. Combined, this level of investment far exceeds the DOT-com boom in both absolute value and cash flow velocity. In other words, the world is experiencing a new version of “internet mania,” but this time with many times more risk.
MarketWatch experts say MacroStrategy calculates the size of the bubble based on “ Capital misallocation” – the amount of Capital that exceeds the economy’s actual ability to generate profits. With interest rates remaining low for a decade, cheap money has driven investment funds to flock to AI as a new promised land. Anything labeled “AI” – from chips, servers, software to applications that are still in the idea stage – can Capital millions, even billions of dollars.
Behind the hype are big names: Nvidia, Microsoft, OpenAI, Amazon, Meta, Google. All are investing heavily in AI infrastructure – Nvidia alone has added more than $1 trillion to its market Capital in less than two years, thanks to the demand for GPUs to power large language models. Goldman Sachs estimates that global AI infrastructure spending alone could exceed $400 billion by 2025, equivalent to the GDP of a mid-sized country. Garran says the boom “cannot last long without demonstrating the real economic value of AI.”
Analysts’ biggest concern is that the Capital of AI is still too low compared to the amount of Capital being poured into it. Many businesses spend tens of millions of dollars to integrate AI into their processes, but have not seen significant productivity improvements. Large models like ChatGPT or Claude still consume huge amounts of electricity and data, while revenue from users and businesses is limited.
“The losers in the AI boom could be even bigger than in the DOT-com bubble,” University of Michigan investor Erik Gordon told Business Insider, because this time it’s not just tech stocks, but the entire physical supply chain – from energy, construction to semiconductors – that are betting on AI.
If the DOT-com bubble was a story of unprofitable websites, then the AI 2025 bubble is a story of unprofitable models. Both were based on the belief that technology would change everything, but both ignored the basic law of economics: real value must precede expectations. And if Garran is right, when the bubble bursts, the impact will be much more widespread.
However, there is one difference that makes investors still hopeful: AI is not an illusion, but a real technology trend. Many experts argue that comparing it to DOT-com is too pessimistic. The Internet also experienced a major collapse, but then strongly reborn, creating giants like Amazon or Google.
But the risk remains. As cheap Capital flows slow and interest rates remain high, projects that do not have real returns will be purged. Similar to the DOT-com boom 20 years ago, the market could see a wave of startups collapse, leading to a sharp correction in tech stocks. MacroStrategy estimates that if just 20% of the Capital invested in AI infrastructure becomes ineffective, the market could lose more than $3 trillion in value.
The core problem lies in the gap between technology and economics. AI technology is evolving faster than businesses can absorb it. Most organizations are still struggling to find ways to integrate AI models into real-world operations without compromising security, privacy, or internal processes. Meanwhile, large corporations are competing fiercely, pushing research and infrastructure costs to unprecedented levels.
The signs of overheating are already beginning to emerge: chip and AI software companies’ valuations have multiplied rapidly; venture Capital is back in full swing; and many publicly traded companies are adding “AI” to their product names to boost their stock prices. All of this is reminiscent of 1999, when any company that added a “.com” suffix could easily raise Capital .
The question is not whether there will be a bubble – but when and to what extent. Garran argues that the current AI bubble cannot last forever, but its bursting will not mean the collapse of the technology. “Like the internet in 2001, when the bubble dissipates, the real core will remain – and that is where the lasting value will be created.”
In that sense, the AI bubble can also be XEM as a “natural cleansing period.” The companies that follow the trends will be eliminated, while the companies with real technology foundations – like Nvidia, Microsoft, Google, and OpenAI – will continue to shape the new world.
Historically, every technological revolution has gone through a period of extreme euphoria before it reaches its true value. The 19th-century railroad bubble, the early 21st-century internet bubble—and now, the mid-2020s AI bubble—all represent overconfidence in the future. But that confidence, while painful in the short term, is the driving force for long-term progress.
