
AI data centers are being built at a pace the world has never seen before. But while investments amount to hundreds of billions, the calculations don’t add up. Revenues are too small, hardware becomes obsolete quickly, and energy consumption is skyrocketing. More and more experts are now pointing to the risk of an economic bubble that could become one of the most costly in modern times.
In recent years, AI models have developed at a rapid pace. Enormous computing power is required to train and operate these, which in turn has led to a massive expansion of data centers. Forecasts from McKinsey show that demand could triple by 2030. Tech giants are therefore competing to build the largest and most advanced facilities.
The development has created a global race. Countries and companies are investing enormous amounts to avoid falling behind in technological competition. But at the same time, there are increasing questions about the economic sustainability of this expansion.
The driving force behind this development is the rapidly growing demand for AI services. Generative models require processors in the billions, sophisticated cooling, and continuous hardware updates. Tech companies therefore build data centers in advance, often without certainty that they will be filled with sufficient demand.
This strategy is risky. Costs quickly soar when buildings, chips, and power supply need to be financed simultaneously. In addition, the hardware becomes obsolete in just a few years, which means that investments must be repeated in a continuous cycle.
Harris Kupperman, founder of Praetorian Capital, has made calculations on the economics of AI data centers. The results are discouraging. According to him, depreciation of chips, networks, and buildings costs around 40 billion dollars annually, while revenues are only 15–20 billion. To cover costs, revenues would have to increase tenfold.
This means that even if the AI market continues to grow, it is very unlikely that it can reach a level corresponding to the investments made today. This reinforces the view that the current investments are not sustainable in the long term.
When revenues cannot cover the costs, a classic imbalance arises. Several analysts are now warning that AI is becoming a speculative bubble. History has shown that such bubbles can have serious consequences, not only for investors but also for societies facing resource shortages.
It is not just about the economy. Energy consumption is increasing rapidly, creating additional problems. If the expansion continues at the same pace, power grids and infrastructure risk being heavily strained, which is already becoming apparent in some countries.
One example is Alibaba, where chairman Joe Tsai has warned that data centers are being built without existing demand. Microsoft has already hit the brakes and paused planned projects in Ohio, signaling that even the largest players see the risks.
In Europe, Ireland has become a warning example. Data centers there now account for 23 percent of the country’s total electricity consumption. This has led to power shortages, increased fees, and heightened political tensions. The situation clearly shows how quickly overestablishment can create problems that go far beyond the technology itself.
The central question is whether AI will generate sufficient revenue to support the enormous investments. If development is not as profitable as expected, the world risks facing hundreds of billions in lost capital and an oversized infrastructure.
What determines the outcome is the pace at which AI can be commercialized and generate actual stable income. If that development is delayed, or if it never reaches the expected level, AI is likely to be remembered as one of the biggest technological bubbles in modern times.
https://www.realtid.se/it-tech/vansinnighetsbubbla-kalkylen-gar-inte-ihop-for-ai-datacenter
https://futurism.com/data-centers-financial-bubble
https://www.cbsnews.com/pittsburgh/news/microsoft-slowing-or-pausing-ai-data-center-projects-ohio






