A company can introduce increasingly advanced robots, attract the attention of investors, and still see its stock price plummet. That’s what’s happening in China, where enthusiasm for artificial intelligence is colliding with doubts about how much money the sector will actually make. The phenomenon has reignited the debate on the so-called AI bubble and the possibility that a similar correction could also occur in the United States and Europe. The issue closely concerns large technology companies, committed to investing billions in the development of new systems, but also those who hold shares or funds linked to the sector.
What’s happening in China
According to a Reuters analysis of October 26, 2026, the STAR 50, an index of the Shanghai Stock Exchange which brings together important innovative companies, including semiconductor manufacturers, lost around 40% in less than four months. The decline erased around 480 billion dollars of capitalisation, i.e. of stock market value, without directly taking money away from company coffers.
Among the companies involved is Unitree, known for its humanoid robots, whose shares have halved since their stock market debut in August. Chipmakers Moore Threads and MetaX have lost about half their value since July, while some optical component suppliers have seen declines of 60% from summer highs.
There was no single triggering event. After two years of gains fueled by enthusiasm for AI and expectations of public support, doubts about future earnings have emerged. Very high valuations, competition between manufacturers and the risk of restrictions on the use of Chinese equipment in US data centers also weighed on the stock.
To understand why there is talk of a possible bubble, just consider another number: despite the collapse, the STAR 50 was still worth around 100 times the annual profits of the companies represented, compared to a ratio of around 30 for the US Nasdaq 100. In other words, investors were willing to pay very high prices relative to companies’ actual earnings, betting on strong future growth.
We talk about a speculative bubble when share prices rise beyond what is justified by the economic results and realistic prospects of companies. If investor expectations are lowered, prices can fall rapidly, as happened in China. This decline, however, does not necessarily prove that the entire AI sector is overvalued.
How many billions are being invested in AI and why are there fears that the bubble will burst
According to Stanford University’s AI Index Report 2026, private investment in artificial intelligence reached approximately $285.9 billion in the United States in 2025, compared to $20.9 billion in Europe and $12.4 billion in China. The comparison does not include all public funding, which is especially relevant in China.
On a global scale, Goldman Sachs expects around $1 trillion in AI-related investments by 2026, mainly focused on infrastructure. The estimate concerns capital expenditure, i.e. those intended for the purchase or construction of assets to be used over time. It’s a different measure than the private investments surveyed by Stanford, so the data isn’t directly comparable.
Much of the resources are allocated to data centers, servers, specialized chips, networks and cooling systems. These infrastructures will have to pay for themselves over time through the services sold. If customers and businesses purchase fewer than expected, recovery of the sums used could take several years.
A concrete example comes from Microsoft. In the quarter ended September 2026, the group recorded $92.3 billion in revenue and $36.6 billion in net profit, as well as $45.2 billion in capital expenditure. This figure also includes the value of assets acquired through financial leasing, contracts that allow them to be used by paying over time. These are data relating to the entire group, not just the AI, and therefore do not allow the profitability of individual projects to be calculated, but show the extent of the expenses.
The issue is whether future revenues will be sufficient to justify the costs and stock market valuations. History offers a precedent: during the dot-com bubble, which burst in 2000, many Internet companies lost much of their value, but the Internet continued to spread. Despite the necessary differences, that story demonstrates that a technology can succeed even when some companies disappoint investors.
Could it also happen in the United States and Europe?
In the United States we are not seeing a generalized collapse comparable to that of China, at least not for now. According to Reuters, as of October 26, the Nasdaq had fallen by around 6% from the highs reached a few weeks earlier, recording a much smaller decline than the STAR 50, even though they are indices with different compositions.
Even the balance sheets of large US companies show how high technological investments are, although not all expenditures are directly attributable to AI. In quarterly results released on October 28, 2026, Google’s parent company Alphabet reported that Google Cloud revenue had more than doubled to $30.7 billion. In the same period, however, the group incurred 51.4 billion dollars in capital expenditure and recorded a negative free cash flow of 9.1 billion, meaning the liquidity generated by operating activities was not enough to cover investments. Other factors, including the build-up of inventories and the payment of a European fine, also affected cash flow. This doesn’t mean that Alphabet was at a loss: the quarter ended with a net profit of around $26.2 billion.
Compared to some of the younger companies caught up in China’s downturn, Alphabet and Microsoft can finance expansion as well through already profitable businesses. However, even this does not guarantee that the sums invested in AI will produce the desired returns or that prices will continue to rise.
The International Monetary Fund, in its October financial stability report, signals the risk of a correction in AI-related stocks if growth and profitability prospects prove less solid than expected. Any decline could be amplified by the debts incurred to finance new projects and by the concentration of investments in a few large companies.
The consequences could also reach Europe. According to an analysis published by the European Central Bank in August 2026, euro area families are exposed to around 440 billion euros to US technology stocks, especially through investment funds. Not all of these holdings are directly tied to AI, but a decline in large American companies could reduce the value of invested savings. A prolonged slowdown would also risk slowing down financing, projects and hiring in the sector.
The Chinese case demonstrates how quickly expectations can change, but it does not allow us to determine if or when a bubble will burst elsewhere.
