Governments must exercise caution in these times of AI overinvestment

A significant proportion of investment in the Western world is currently going into the supply chain for providers of artificial intelligence (AI). In fact, without this capital expenditure, growth might well be minimal in many key geographies. According to Jason Furman, former chair of the United States’ Council of Economic Advisors, the “investment in information processing equipment and software” category of the United States’ national income accounts has been responsible for over 90 per cent of that economy’s growth in the first half of 2025. This is the statistical reflection of a real-world phenomenon: The vast diversion of resources to building data centres, both in the United States (US) and elsewhere. This trend is visible as well in sectoral results within the equity markets. Two of the three best-performing sectors in the S&P 500 are communications services and information technology —this is the “AI trade” that is consuming Wall Street’s attention. The best-performing sector, utilities, is also likely growing on the back of expectations regarding future demand growth, thanks to the power-hungry data centres being planned.

 

Within the sector, however, there are signs that reflect not just a classic boom-and-bust cycle building up, but also that significant systemic risks are being ignored. One sign of irrational exuberance is when valuations become detached from the reality of earnings or of reasonable projections. Investors begin to price a stock on the basis of a narrative rather than due to the execution of a new technology. What is particularly worrying is that signs of circularity are developing. Semiconductor companies like Nvidia are investing in companies that will buy their products, shoring up expectations of demand that then justify their book value to their investors. It recently announced plans to put $100 billion in OpenAI, alongside an equity stake, which will require the AI firm to make cash purchases of the chipmaker’s products. Some extended valuations and investments parallel the dotcom boom at the turn of the century. Back then, Warren Buffett would talk about how the ratio of US market capitalisation to gross domestic product was a danger signal; it had reached over 150 per cent. Today, it stands at over 200 per cent. More than half its gains this year have been driven by Nvidia, Meta, Microsoft, and Broadcom, according to research from Morningstar. The Economist has estimated that AI revenue to leading tech firms is about $50 billion a year, less than 2 per cent of the amount being invested annually.

 

The larger this investment is, the more is publicly traded wealth tied up in it. And the greater the dependence of economic growth on such investment, the greater is the systemic danger. When the possibility of a bubble is widely understood and yet the money keeps flowing, something is providing a sense of security — and, in this case, it is likely the belief that this is such an important sector that national treasures will provide a backstop if it all falls apart. Tech today is playing the role of driving growth and channelling investment that finance played prior to the 2008 crash. But governments today have far less fiscal space than they did two decades ago, and will struggle to bail out Big Tech even if they think it is justified and necessary. An equity wipeout and multiple stranded assets are a very real possibility. That said — just as after the dotcom boom — real assets will be built that add to national wealth. Nevertheless, for retail investors and cash-strapped developing-world governments like India’s, this should be a time of caution and not of exuberance.