Appeared in Substack on November 22, 2025.
Artificial Intelligence will surely transform business, education and society in significant ways in years to come. However, the current exuberance about AI, especially in the financial world, is way over blown. In our 2004 book “Beware the Winners Curse” Hank Lucas and I provided compelling evidence for the 1990s fast rise of dot com and tech companies, and their eventual spectacular fall. The causes included hubris (invulnerability), herd mentality (fear of missing out – FOMO), voodoo financial arrangements (leading to high valuations without revenue) etc. These factors are playing out right now about artificial intelligence.
The level of optimism about AI is sky high. Not a day passes by without excited claims about all the amazing things that AI can do. The CEOs of AI companies have become rock stars. McKinsey is pumping up this excitement reporting that “Europe’s deep-tech engine could spur $1 trillion in economic growth”. This is just Europe; the projections for the U.S. and Asia are even more lofty. Turns out that the U.S. economic growth is 90% spurred on by data center and AI investments. We also know that 60-70% of the rise of the S&P is due to AI-based companies; this does not even capture the unlisted start-ups. Venture capitalists have spent over $200 Billion on AI in the past three years. There is almost a religious fervor about AI. This level of passion about AI far surpasses the exuberance in the dot com era.
There is no doubt that herd mentality has taken over AI. Every company is looking to adapt AI into its operations. Many universities are leaping on the AI bandwagon with little or no thought. Titles of several courses in engineering and business now have AI in their titles. Research proposals in fields as disparate as music and mathematics are about AI. Data center investments have tripled since 2022; whole host of companies are trying to find a niche in AI use before anyone else does. There is data center investment across the planet from Indonesia and Pakistan to UAE and the USA. Oracle, for example, is aggressively building data centers and cloud computing facilities and is apparently carrying more than $20 Billion in debt. Even Secretary of Commerce Ludnick’s family are supposedly getting into data center “construction” according to the New York Times. Companies are investing on mission rather than profitability. Every project involving AI gets funded, good ideas and bad ones, and executives have a tough time figuring out the difference. The herd mentality around AI smacks of the Tulip Mania in Holland in the 17th century. A recent JP Morgan Chase AI Capex report estimates that AI companies need $60 Billion in annual revenue to generate just 10% profit. But, no one wants to be left behind in the AI “gold rush”; there is a true fear of missing out on the AI “boom”.
Circular financing in the AI ecosystem has become incorrigible even though there are lessons to be learned from the dot com era. In the late 1990s, several tech companies used “vendor financing” to get around the fact that they could not easily sell their tech products in an open market. At that time, the most exciting emerging technology was optical switching and telecom service providers financed tech build out. For example, Williams Communications agreed to provide financing to Sycamore Networks to the tune of $100 million per year for four years, and Sycamore would provide Williams first dibs on it potentially ultra-fast optical switches. When Sycamore went IPO in October 1999, it had losses of $19.5 million on sales of only $11.3 million and yet the offer was oversubscribed and ended at a valuation of almost $3 Billion. There are many such stories including Chromatis being bought by Lucent for $4.5 Billion only to shut it down! Sycamore Networks crashed and burned, pivoted to telecom management systems and got sold to Marlin Equity Partners in October 2012 for just $18.75 million! What a drop in value!! This kind of financial behavior seems to be coming back in droves in the AI sector.
The AI network of companies is totally immersed in circular financing. Nvidia produces 80-90% of high performing chips that go into fast computers and data centers, followed by AMD. Nvidia recently committed to invest $300 Billion in OpenAI which in turn promised to buy almost all its chips from Nvidia; the rest from AMD. It is reported that OpenAI has bought 150 million shares in AMD, I presume to ensure good cost terms for the chip purchases. This circularity is boosting the share prices of Nvidia and AMD and the private valuation of OpenAI. Almost 2/3rd of the revenue of Nvidia comes from just four companies. Nvidia is now the most valued company in the world at over $4.2 trillion.
Microsoft owns 30% of OpenAI and thus OpenAI is obliged to use MS-Azure cloud computing for its data centers. It turns out that 50% of the data center cost for OpenAI is spent on Microsoft. A similar deal has been struck by OpenAI’s competitor Anthropic with Amazon. Amazon has invested a total of $8 Billion in Anthropic-Claude and in turn, Anthropic has spent over $2.66 billion on Amazon Web Services (AWS) for the entirety of 2024 and into September 2025, which exceeds its estimated revenue of $2.55 billion. Google has its own AI model called Gemini but has also invested in Anthropic-Claude to add competition to OpenAI’s GPT models and to pressure Anthropic to use the Google cloud services. OpenAI recently contracted to buy $300 Billion of data center cloud computing from Oracle which in turn will buy chips from Nvidia. Circular financing is also boosting the share prices of all these tech companies into the stratosphere.
The question is whether all this excitement is misplaced and are we looking at a house of cards? Is there too much hubris among the tech leaders, management consultants and academia about AI? Are there too many people investing in both the physical assts of AI and the shares of AI-based companies for fear of missing out? What if the deployment of AI slows down worldwide for regulatory and business reasons? There is already evidence that getting real productivity out of AI is going to take several years. What if data center investments stall? Are we going “all in” in terms of AI? Only time will tell. When the dot com era came tumbling down in 2000, the market lost 75% of its value; it took over 15 years to recover. Sure, several companies succeeded including Google, Amazon and Microsoft but their valuations took a big hit. Who can predict which companies will be left standing when the AI bubble bursts? Bank of England Financial Policy Committee reported that the valuation in the U.S. market is comparable to the peak of the dot com bubble. Even so, it seems difficult to moderate the religious zeal re AI.
Anand Anandalingam - Personal Web Site 