China has just delivered a $480 billion warning to the global artificial intelligence industry. American investors should take it seriously.
In less than four months, China’s STAR 50 index, dominated by technology and semiconductor companies, lost around 40% of its value. The sell-off wiped out approximately $480 billion in market capitalisation and dragged the broader CSI 300 index about 20% below its June peak.
Per Reuters, AI stocks had grown to represent 42% of the market cap and 57% of trading turnover across mainland Chinese and Hong Kong-listed shares.
The enthusiasm that drove these valuations has collided with concerns about demand, profitability and the restrictions facing Chinese technology exporters. Investors are discovering that government ambitions and technological progress do not guarantee rising share prices.
The obvious question is whether the United States is next. The answer lies in the financial structure of its AI boom, and that structure has serious weaknesses.
The biggest American technology companies are spending extraordinary sums to build the infrastructure required to power AI. Data centres, advanced chips, computing networks and electricity supplies require enormous upfront investment. Companies are betting that future demand will justify the expense.
Reuters reported earlier in July that five major American technology companies, Microsoft, Alphabet, Amazon, Meta and Oracle, were on course to spend more collectively on capital expenditure than they generated in free cash flow by 2027. Their annual operating cash flow was projected to rise by about $340 billion between 2025 and 2027, while capital expenditure was expected to increase by roughly $534 billion.
That is the problem in two numbers. Spending is accelerating faster than the additional operating cash these companies are generating. These businesses have substantial advantages over speculative start-ups. They possess established customers, profitable operations, valuable intellectual property and enormous distribution networks. Those advantages make them more resilient, but they do not make their investments automatically profitable.
A company can afford to spend billions and still destroy shareholder value if the returns fall short of expectations. Building the infrastructure is an expense. Making that infrastructure earn more than it costs is the business.
The AI industry has spent years selling investors a vision of extraordinary future productivity. The financial markets now need evidence that this productivity will translate into earnings at a scale that justifies the spending.
AI adoption can continue growing while AI stocks collapse. There is no contradiction.
Suppose businesses increasingly use AI to write software, analyse documents, automate customer service and improve productivity. That creates real economic value. But if competition drives prices down, customers capture much of that value instead of the companies providing the technology.
Cheaper models and open-source alternatives intensify the pressure. They make AI more accessible, but they also make it harder for individual providers to charge premium prices indefinitely. Companies must keep spending on computing power, research and infrastructure while competing to sell services that customers increasingly expect to become cheaper.
Investors, meanwhile, have priced many AI businesses around expectations of rapid growth and substantial future profits. If those profits arrive more slowly than expected, valuations can fall sharply even while the technology improves.
That is how a technological success becomes an investment disaster.
The dot-com crash demonstrated the distinction. The internet transformed commerce, communications and everyday life, but the transformation did not rescue every company that attracted money during the boom. Some businesses failed because their models were weak. Others were overtaken by competitors. Many investors paid prices that the eventual profits could never justify.
AI does not need to fail for its bubble to burst. Its economics simply need to disappoint the people financing it.
The danger extends beyond share prices. In its October 2026 financial stability assessment, the International Monetary Fund examined 14 publicly disclosed AI financing transactions since 2023. Together, these involved approximately $300 billion in equity investment and $650 billion in debt, linked to $1.4 trillion in future revenue commitments.
These arrangements connect investors, AI companies, infrastructure providers and major technology firms through investments, borrowing and promises of future purchases. If a major recipient of funding struggles, the consequences can spread through the same network that helped finance its expansion. Equity investments lose value, debt becomes harder to service and expected revenue fails to materialise.
This creates a dangerous feedback loop. Lower valuations make financing more expensive. More expensive financing forces companies to cut spending. Those cuts reduce expected revenue for suppliers and infrastructure operators, weakening the businesses that depend on continued expansion.A correction would hit semiconductor manufacturers, data-centre operators, construction contractors and other suppliers. It would also expose lenders and investors who assumed that the AI build-out would continue expanding at its current pace.
The IMF has explicitly warned that a correction in AI-related assets could affect the wider economy. This is no longer simply a question of whether technology shares are overpriced. It is a question of how much economic activity depends on the investment boom continuing.
Meanwhile China and the United States do not have identical technology industries or financial markets. The Chinese sell-off is not a timetable for an American crash. But America has built its own concentration of risk. A handful of technology companies dominate AI investment, and their spending plans influence chipmakers, cloud providers, energy projects and financial markets. Their valuations depend heavily on expectations about how much AI will earn in the future.
The United States has stronger established technology businesses capable of financing investment from existing operations. That gives it a cushion.
Yet even profitable companies can make poor capital-allocation decisions, and even financially strong firms can suffer major share-price declines when expectations become detached from returns. The scale of the investment makes the consequences important. If companies suddenly decide that they have built too much capacity, they will cancel projects, delay data centres and reduce orders for equipment. Suppliers will lose revenue. Investors will reassess valuations. Businesses that borrowed heavily to serve the boom will face the consequences of falling demand.
The United States does not need to reproduce China’s market crash exactly. Its own correction could take a different form, with different companies suffering at different times. The underlying vulnerability is the same: capital is being committed today on the assumption that tomorrow’s AI economy will deliver exceptional returns.
The question is not whether AI has value. It already does. The question is whether the value it creates will be large enough, and arrive quickly enough, to justify the money being poured into it. China’s market has shown what happens when investors lose confidence in the gap between expectations and commercial reality. American technology companies now face the same fundamental test. Their products must generate enough revenue and profit to justify their infrastructure costs, their financing arrangements and the valuations attached to their future growth.
Investors can believe in AI and still be wrong about the price of an AI company. They can correctly predict that a technology will transform the world and still lose money by paying too much for it. That is the lesson from China’s $480 billion collapse. The technology does not have to fail for the bubble to burst. The numbers only have to stop adding up.
Be First to Comment