Is the AI bubble going to burst? Why the world is panicking about the future of artificial intelligence - The Independent

Want to bookmark your favourite articles and stories to read or reference later? Start your Independent Membership today.

Artificial intelligence companies’ valuations are plunging, and the world of AI looks set for a major shake-up – with some predicting that the current AI bubble might be coming to an end.

But the real process lurking behind the current sell-off is less about the dramatic predictions about artificial intelligence transforming or destroying the world. Instead, it has been brought about by a series of relatively down-to-Earth concerns about how the rollout of the future might actually happen.

Do those concerns really mean that the AI bubble – if we are in one – is actually going to burst, however? And what do they mean for normal people who are either worried or excited about how the world is about to change?

Here’s why people are both panicking and rejoicing about the future of the artificial intelligence market, after a week that has seen trillions of dollars lost in companies’ share prices and predictions of dramatic changes in predictions about how exactly the future might turn out.

Thursday and Friday of this week brought panic in the stock markets, with the US Nasdaq Composite index falling 5.5 per cent, after already trading down in previous days. The effect was even more dramatic in stocks that are focused on artificial intelligence: Nvidia, which makes the chips that power the AI revolution, dropped 9 per cent, for instance, and some even more specialist companies lost even more in value.

Things were already looking shaky for those companies that are heavily invested in AI. A week before the current drop, for instance, Google parent company Alphabet saw a dramatic sell-off after it announced results that showed a vast AI spend that was not translating into revenues, with the company’s shares falling 12 per cent in just one day and wiping $500 billion from its value.

Alphabet’s loss was second-biggest daily reduction in the value of the company in the history of the stock market. But the fact it is only the second is a reminder that this is not entirely unprecedented: the record is held by Nvidia, which lost $593 billion in January 2025, amid panic that Chinese AI companies were making new and more efficient models that could challenge the US domination of artificial intelligence and reduce demand for Nvidia’s specialised chips.

Those problems continued this week, however. And they were largely the result of the US midterms, which were influenced by increasing concern over the rollout of data centres, and a big victory for Democratic candidates who are not only more skeptical about those data centres but more keen on AI regulation more generally.

"There's going to be a lot of pushback about those data centres being geographically located in someone's backyard, straining the power grid, straining the water systems," said Bruce Zaro, managing director at Granite Wealth Management.

"Although there are a number of data centres already approved, there are still a lot to be permitted. There's going to be a difficult road to get those permitted forthcoming."

That meant that data centre operators were particularly hit by the fall in the stock markets, with Applied Digital, Cipher Digital and IREN all falling more than 14 per cent. But those data centres are also required to power artificial intelligence more generally – and so any challenge to their rollout will hit better known companies such as OpenAI and Anthropic, too.

Anthropic will be particularly concerned. It is currently marketing its own flotation on the public markets, which has been seen as a test of sentiment about AI generally, and any fall in investor sentiment could reduce the value or even throw off that IPO entirely.

What is clear, however, is that all of those concerns are not necessarily part of the grand claims about artificial intelligence, and the ways that it might restructure the economy or change daily life. Instead, they are about relatively concrete concerns: that companies’ heavy investment in AI might not pay off, and that they might not be able to roll out the big changes that they need to ensure that it actually works.

That means that it is actually harder to predict what happens next: it depends on the fine details of regulation, public sentiment and financial positions. Those complex concerns also mean that the effect of any problems in the AI market on normal people will probably be more complex, too.

There is one very obvious way that the current drop in the value of tech stocks and share prices more generally will affect normal people: any fall in the markets hits normal consumers who might have invested in the stock market. What’s more, so much of the world’s investments are now bound up in those AI companies and their continuing success that any substantial plunge could upset the markets more generally, bringing about a bursting of the bubble by itself.

Any of those troubles will not only affect the people who have invested in those companies, however. Building big AI companies is incredibly capital intensive – it requires funding the training work to build the systems, and the data centres required to keep them running – and so if the financial markets become less favourable to those technology companies then the rollout itself could quickly slow.

The same is true of the concerns about regulation and general worries about political pressure that have become very present after the US midterms. If newly elected American politicians really do restrict the ways that AI companies can work, including most notably how and where they can build data centres, then that too would slow the rollout of the technology.

That would affect normal people in a number of ways, far beyond its influence on the stock market. If artificial intelligence really is going to transform the economy, then any change in both the ease and cost of accessing those tools will clearly affect how exactly that transformation happens.

Notably, the latest worries are not primarily inspired by worries about the technology itself. In the past, the world has seen a number of “AI winters” – but those have generally been powered by slowing in the development of the technology, which then in turn brings reduced investment and lower public interest that only helps slow that development even more.

This time around, the AI companies at least continue to suggest that they are making major strides in the performance of their systems, and that they are not seeing any indication that might slow. Instead, the worries are more practical: that they might not be able to actually build the vast networks of computers they need to provide those systems, or to deliver the financial return on that technology that they have promised to the market.