Wealth Elimination: The AI Bubble is about to burst
Investor Ray Dalio predicted the deflation of the artificial intelligence bubble. The reasons will be related not so much to the financial failure of companies operating in this field, as to the banal desire of investors to lock in profits. All signs of the retreat of the almost religious euphoria are already evident. Meanwhile, a downturn in the AI sector could hit the broader economy: primarily the American one, but eventually the global one as well. About how events will develop and whether to expect a new global financial crisis, see the Izvestia article.
Wealth or money
To understand the nature of the current crisis, Dalio suggests drawing a hard line between the concepts of "wealth" and "money." Wealth is the capitalization of assets, often existing solely on paper. A startup can receive a valuation of $1 billion after attracting a funding round of just $50 million. These figures get into reports and increase the virtual wealth of founders and investors, but they cannot be spent directly.
On the contrary, money is a real means of payment, the liquidity needed to service debts, pay taxes, cover losses, or refund funds to investors when closing positions.
Deflating a bubble is the process of converting wealth into money. It is launched when holders of paper assets urgently need real cash to fulfill their financial obligations. Especially when "wealth" does not generate "money" directly through profits.
The main problem of the AI sector by the summer of 2026 is a significant gap between capital expenditures and real revenue. This year, the technology giants — Microsoft, Alphabet, Meta (recognized as extremist in Russia), Amazon — planned to spend about $650 billion on the purchase of chips, servers and the construction of data centers.
Venture capital funds like Sequoia Capital calculated a year ago that in order to recoup infrastructure investments of this scale, the AI industry must generate over $1 trillion in revenue from end products. The reality turned out to be an order of magnitude more modest. The corporate sector is extremely reluctant to integrate language models due to the risks of data leakage and algorithm hallucinations. Revenue from the sale of AI subscriptions and specialized software is estimated in the tens, but far from hundreds of billions of dollars. Companies have built a digital infrastructure, whose products have not yet found a mass paying customer.
Energy barrier
The problems started not only because of weak demand, but also because of physical limitations. Generative networks require growing volumes of basic electricity generation from year to year. Here, geopolitics intervened in technological plans.
The blocking of the Strait of Hormuz in the spring of 2026 triggered global demand for all types of energy. The cost of oil has consolidated above $100 per barrel, and natural gas prices have skyrocketed in Europe and Asia. Although the United States has its own gas base, the American energy system was not ready for the simultaneous increase in the cost of resources and a sharp jump in consumption from new mega-data centers. Operating costs for cooling and powering server racks increased by 30-40% compared to the 2024 business plans, which made the operation of many data centers unprofitable.
At the same time, technology companies have faced resistance from society, which is not happy about the increase in electricity bills. Approval of new construction sites for data centers in the U.S. states of Virginia, Texas and Ohio is blocked by local municipalities. Server appetites lead to a shortage of capacity in local networks and an increase in tariffs for the population. In addition, cooling systems require millions of gallons of fresh water. Against the background of climate change, the withdrawal of water resources provokes open protests by farmers and local residents. The industry faced a "not in my backyard" effect that physically halted the expansion of AI infrastructure.
Dotcom Ghosts
Comparing the current situation with the dotcom crash of 2000 is self-evident. Investors draw direct parallels between router manufacturer Cisco (the symbol of the zero bubble) and Nvidia. Both companies sold "shovels during the gold rush," reached astronomical capitalization, and then faced a collapse in orders.
However, the structure of the current crisis is fundamentally different from the events of a quarter century ago. In 2000, the market was flooded with new startup companies without revenue and viable business models. In 2026, the main players in the AI race are ultra-successful corporations with a dominant position in their core niches (search engines, advertising, e—commerce). Their balance sheets have a huge margin of safety.
The deflation of the bubble will not affect the survival of bigtech as such, but the estimated multipliers of the stock price to earnings ratio. The market placed a premium on the exponential growth of AI in the quotes of these companies. The withdrawal of this premium means a drop in the capitalization of the technology sector by 30-40%. In absolute terms, shrinking the Nasdaq 100 index by a third will lead to the evaporation of $7–$8 trillion of virtual wealth. It will be a severe stock market crash, extremely painful for retail investors and pension funds, but unlike in 2008, it is unlikely to provoke a classic banking crisis and a series of defaults in the financial sector.
Cooling system
For the United States, the cooling of the AI boom poses the threat of a hard landing of the economy. Growth in the United States since 2024 has been based on two pillars: the wealth effect of the stock market and large-scale corporate capital investments in infrastructure.
If the stock market corrects by 30%, consumers feel poorer and sharply reduce spending. Retail sales are going into negative territory. At the same time, technology giants, having fixed losses from AI, are switching to austerity mode. Construction of new data centers will be frozen, and contracts for the supply of equipment, concrete, cables, and cooling systems will be canceled. The programs for hiring engineers will also be curtailed.
Against the backdrop of Middle Eastern inflation (expensive gasoline) and high Federal Reserve interest rates, the shutdown of the investment cycle of technology companies deprives the US economy of the last driver of growth. Stagflation is emerging: the economy freezes due to a lack of investment, and prices remain high due to disrupted global energy logistics. In an extreme scenario, the economy may fall into a long—lasting recession, up to a year, especially adding to the equation a sharp reduction in migration (which means not only a reduction in labor, but also consumption, and the collapse of the part of the economy that is focused on serving millions of migrants). But even a drop in growth rates to 0-0.5% will be extremely painful.
Then everywhere
The global economy will feel the impact through microelectronics supply chains. The reduction in orders for new accelerators from Nvidia and AMD will instantly hit TSMC's production plans in Taiwan. But the main victims will be memory chip manufacturers.
South Korea (Samsung, Hynix), whose economy has already been hit by the energy crisis, will face a whiplash effect. The huge reserves of HBM memory produced for the expected boom of data centers will be unclaimed. Chip prices will collapse, and the export earnings of Asian economies will decline sharply, triggering a recession in the Asia-Pacific region. From here, the crisis will spread along the chain to all countries of the world. With inflation still high, the arsenal of central banks will be limited, which threatens to further exacerbate the crisis. Finally, both the geopolitical and domestic political situation in many countries is highly unstable, which may encourage politicians and regulators to take erroneous steps. In general, the prospect of a new global recession looks like a likely scenario, although not a "baseline" one.
Ray Dalio's forecasts indicate the end of the stage of irrational optimism about the AI sector. The technology will certainly stay with us, but its financial prospects need to be reviewed. Artificial intelligence will cease to be a magic word that attracts trillions for free for a healthy living. After the shake-up, neural networks and the industries serving them will become a "mature" sector, where every dollar invested should have a clear justification for the return on investment. The adaptation of the global economy to this reality will take the next few years, simultaneously washing away the excess capital accumulated in recent years from the market.
Переведено сервисом «Яндекс Переводчик»
