Few questions in markets right now divide opinion as sharply as this one: has the rush into artificial intelligence created a bubble, or is it the early stage of a genuine industrial transformation? The argument has moved from social media into the pages of the world’s biggest business publications, and it deserves a calmer hearing than the headlines usually allow.
This article takes a balanced view, setting out what each side argues before drawing any conclusions. For readers who hold shares directly, through a pension, or through a fund, the answer matters less as a prediction than as a guide to what to watch over the next year.
What the dot-com comparison does and does not prove
History offers a warning, not a script. The dot-com bust did not prove the internet was a fad; it proved that correct ideas can still be bad investments at the wrong price, and that infrastructure built in a boom can sit underused for years before demand catches up. The parallel cuts both ways for AI. Even if the technology fulfils its promise, individual companies can fail, margins can compress as competition grows, and late-stage investors can be disappointed. Equally, calling every rise a bubble ignores how often transformative technologies look expensive before their biggest uses are understood.
What ordinary investors should watch
Four signals are more useful than daily commentary. First, revenue: are customers paying for AI products at a scale that justifies the spending on chips and data centres? Second, capital discipline: are the big spenders still funding investment from cash flow rather than debt? Third, breadth: does the rally spread beyond a handful of names, or does it become ever more concentrated? Fourth, expectations: when good results start producing falling share prices, the market is signalling that perfection was already priced in. Watching those indicators beats trying to call a top.
Readers following our recent coverage may also find this related report useful: Sixteen Thousand Students, One Clear Lesson on Phones.
A balanced way to hold a view
Diversification remains the unglamorous answer to an unresolvable argument. Investors who believe in AI’s long-term importance can hold exposure sized so that a sharp correction would be survivable rather than ruinous, avoid borrowing to chase momentum, and rebalance when one theme swells to dominate a portfolio. Those who are sceptical can still acknowledge that staying entirely out of the market’s largest growth engine carries its own risk. The debate is not really about whether AI matters — it plainly does — but about how much of its future success today’s prices have already spent.
The bear case: prices running ahead of proof
Sceptics, including commentators cited in recent Financial Times and New York Post business coverage, point first to valuation. A small group of AI-linked companies now accounts for an outsized share of major stock indexes, and some trade at multiples that assume years of flawless growth. Bears also highlight circular financing deals between chipmakers, cloud providers and AI labs, enormous capital spending on data centres whose returns are still unproven, and the speed with which private valuations have risen. Their historical reference point is the dot-com era, when a real technology — the internet — still produced a painful crash for investors who paid any price for exposure to it.
For wider background, see the resources and reporting collected by the Financial Times, which track the evidence behind the claims on both sides of this debate.
The bull case: earnings, demand and scarcity are real
Bulls answer that this cycle is different in one crucial respect: the leading companies are enormously profitable. Unlike most dot-com darlings, today’s AI leaders generate vast cash flows from existing businesses, and demand for computing power continues to outstrip supply. Supporters argue that AI is already producing measurable productivity gains in software, customer service, logistics and research, and that infrastructure spending resembles earlier build-outs of railways, electricity grids and broadband — expensive and occasionally excessive, but ultimately transformative. On this view, high valuations reflect scarcity of genuine AI capability, not pure fantasy.
