Artificial intelligence can be revolutionary and overpriced at the same time. Those ideas are not enemies; they are practically roommates. Railways changed the 19th century, the internet changed the 21st, and both also inspired investment manias that left plenty of shareholders clutching beautifully designed empty bags.
That uncomfortable historical pattern is behind fresh anxiety over the AI boom. The International Monetary Fund has warned that a sharp fall in AI-related shares could create wider financial stress, while recent volatility in technology-heavy markets has reminded investors that gravity has not, regrettably, been discontinued.
The question is not whether AI will endure. It almost certainly will. The more useful question for households, investors and policymakers in Europe is whether the financial expectations piled on top of it have become too concentrated, too expensive and too dependent on cheap capital.
A real technology can still produce a bubble
Calling something a bubble is often mistaken for calling it a fraud. That is not what the term means. A bubble can form when a genuine innovation attracts so much money, so quickly, that investors begin pricing in a near-perfect future: enormous demand, durable margins, few competitors and no awkward surprises.
In the reference discussion, investor and author Ruchir Sharma describes four familiar warning signs: overvaluation, over-ownership, overinvestment and excessive borrowing. These are not a countdown clock, and bubbles are notoriously difficult to time. People warned about dot-com excesses years before the crash; being early in finance can feel suspiciously like being wrong, right up until it does not.
AI has some features that make the comparison tempting. The technology has become central to stock-market enthusiasm, particularly in the United States. At the same time, capital expenditure on data centres, chips, cloud capacity and power infrastructure has surged. The commercial payoff, however, remains uneven. Revenue is growing at major AI businesses, but whether it will justify the enormous outlays across the whole chain is another matter.
That distinction matters. A technology can make consumers and companies more productive even if many firms that race to build it deliver poor returns for shareholders. History is littered with useful inventions that were terrible investments at the peak of the excitement.
The bill is bigger than the chatbot
Much of the AI story is really an infrastructure story. Powerful models require vast computing capacity, advanced semiconductors, specialist networking equipment, data centres and startling quantities of electricity. Those assets are expensive, depreciate over time and cannot be conjured from a motivational LinkedIn post.
Large US technology groups including Microsoft, Amazon and Google have deep profits and strong balance sheets. That sets this cycle apart from the era of flimsy dot-com companies with no revenue and a business plan that amounted to “website, but faster.” Their capacity to absorb a downturn is substantial.
Yet resilience is not the same as immunity. Heavy AI spending can reduce free cash flow, lower future returns and make even excellent companies more vulnerable to a reassessment of growth expectations. Share prices do not need a company to fail in order to fall hard; they merely need reality to arrive five minutes late to the party.
Debt is a sharper issue for less established parts of the ecosystem. The video highlights AI computing provider CoreWeave, which had accumulated around $50 billion in debt while still reporting losses, according to the discussion. Its share and bond pressure became a vivid example of what happens when investors start demanding proof that spending can become sustainable cash generation.
There is also concern about circular financing: businesses investing in suppliers, customers or partners in ways that can make revenue growth look more robust while the cash ultimately travels around the same tightly connected network. Such arrangements are not automatically improper, but they deserve close scrutiny. Accounting can be patient; debt markets tend to be less romantic.
Why Europe cannot watch from the sidelines
Europe is not the epicentre of the listed AI trade in the way the United States is, but it is far from insulated. European pension funds, insurers, asset managers and retail investors often hold US equities directly or through global and US-focused index funds. When a small group of mega-cap technology stocks drives a large share of an index’s gains, passive diversification can become less diverse than its label suggests.
European companies are also part of the supply chain and the customer base. Semiconductor equipment, industrial automation, energy grids, cloud services, telecommunications and data-centre development all have European exposure. A sustained AI investment boom can support demand across these areas. A severe retrenchment could weaken it just as quickly.
There is a broader macroeconomic concern, too. If capital rushes into one fashionable sector, other productive investment can be squeezed out. Sharma’s argument is that a concentration of funding in AI may leave other users of the technology short of capital. It is a slightly absurd possibility: businesses may be offered miraculous tools while lacking the funds to deploy them properly.
Higher long-term interest rates add another complication. Governments in many advanced economies carry substantial debt burdens, and costly public borrowing can compete with private borrowers for funds. Data centres and computing projects thrive on large upfront investment; tighter financing conditions are therefore not background noise but a potential stress test.
A correction is not automatically a 2008-style crisis
The phrase “bubble burst” understandably summons images of the global financial crisis: bank failures, frozen credit markets and a queue outside every cash machine. That remains a much higher bar than a technology-stock correction.
A market fall becomes systemic when losses spread through highly leveraged institutions, opaque financial products and essential credit channels. The 2008 crisis was devastating not merely because US house prices declined, but because mortgage risk was deeply embedded across the banking system. Today’s AI investment landscape contains leverage and private-credit exposure, but the largest technology companies are generally profitable, cash-generative businesses rather than fragile start-ups built entirely on borrowed oxygen.
That does not make the risks trivial. A steep drop in highly concentrated equity markets could hit household wealth, business confidence and investment plans. It could expose stressed borrowers in data-centre infrastructure or private debt. It could also prompt a more general sell-off if investors abruptly decide that the same growth assumptions have been hiding inside far too many portfolios.
The likely range of outcomes is wide: from a painful but contained repricing to a broader financial event if leverage, liquidity pressures and weak confidence begin feeding one another. Regulators should be examining those connections rather than treating every AI concern as either hysteria or inevitable destiny.
What to watch when the hype meets the spreadsheet
Grand forecasts about AI’s future are plentiful. More revealing signals are usually rather less glamorous:
- Revenue versus capital expenditure: Are AI services producing durable income at a pace that can support infrastructure spending?
- Debt refinancing: Can highly leveraged computing and data-centre businesses raise fresh funds without punishing interest costs?
- Market breadth: Are gains spreading across industries, or resting on an increasingly tiny group of companies?
- Interest rates and bond yields: Higher yields raise the hurdle for long-dated, expensive growth projects.
- Corporate disclosure: Investors need clarity on related-party arrangements, commitments and the useful life of rapidly evolving hardware.
For individual investors, the least thrilling advice is often the most durable: check indirect exposure, understand what sits inside an index fund, and avoid confusing a popular theme with a balanced portfolio. Diversification will never go viral, but it has survived more market cycles than most buzzwords.
AI may reshape medicine, manufacturing, research and everyday office work. Its financial backers, however, still face the old-fashioned demand to turn ambition into returns. The technology can remain transformative even if some of today’s market assumptions discover that they were, shall we say, trained on a rather optimistic dataset.