AI chip sell-off deepens as investors question who will fund the boom
Semiconductor shares have fallen sharply across Asia, Europe and US pre-market trading as investors reassess the cost of AI infrastructure, Nvidia’s growing financial exposure to its customers and the competitive threat emerging from China.
A sharp sell-off in semiconductor shares has raised a question that has hovered over the artificial intelligence industry for months: how long can companies continue spending hundreds of billions of dollars on computing infrastructure before investors demand clearer evidence of financial returns?
South Korea was at the centre of the market decline on Tuesday, 28 July 2026. The Kospi closed approximately 10.8% lower after triggering a market-wide circuit breaker. Samsung Electronics fell 13.4%, while SK Hynix, a major supplier of the high-bandwidth memory used in AI systems, lost 14.7%. Japan’s Nikkei and Taiwan’s Taiex also declined as investors reduced their exposure to semiconductor companies.
The selling spread beyond Asia. European semiconductor shares opened lower, with ASML falling approximately 2%, while US Nasdaq futures declined ahead of the opening bell. Nvidia, Micron Technology, Applied Materials and US-listed shares of TSMC and SK Hynix were all lower in pre-market trading.
The market movement does not demonstrate that demand for artificial intelligence has collapsed. It shows that investors are becoming less willing to treat every increase in AI infrastructure spending as automatically positive.
Key takeaways
South Korea’s Kospi fell approximately 10.8%, with Samsung Electronics and SK Hynix among the largest casualties.
Investors are questioning whether the revenues produced by AI services will justify the enormous cost of chips, data centres and electricity.
A reported Nvidia financing guarantee for an OpenAI data-centre project has intensified concerns about financial circularity within the industry.
China’s progress in AI models and semiconductor manufacturing is challenging assumptions about the permanent dominance of existing suppliers.
Upcoming earnings from major US technology companies will provide an important test of whether AI spending is translating into revenue and profit.
Why Nvidia’s reported OpenAI financing matters
One immediate source of anxiety is a Wall Street Journal report that Nvidia is discussing a financial guarantee of approximately $250 billion for an OpenAI data-centre project being developed by SoftBank subsidiary SB Energy in southern Ohio.
The proposed 10-gigawatt facility could cost more than $500 billion, including the chips installed inside it. The reported $250 billion guarantee would support lease and construction financing, while Nvidia is also said to be considering financing as much as $350 billion of OpenAI chip purchases.
These figures are not completed investments. Reuters said it could not independently verify the report, while Nvidia, OpenAI and the US Commerce Department had not responded to requests for comment at the time of publication.
Nevertheless, the report unsettled investors because it illustrates how the AI infrastructure market may be becoming increasingly dependent on its largest suppliers helping customers finance future purchases.
Nvidia would benefit if OpenAI built a larger computing estate filled with Nvidia hardware. However, guaranteeing financing connected to those purchases would expose Nvidia to a different kind of risk. It would no longer be operating only as the supplier of chips. It would also be using the strength of its balance sheet to help sustain demand for them.
This is sometimes described as circular financing, although the term can oversimplify a wide range of legitimate commercial arrangements. The important issue is whether demand is being supported primarily by end-user revenue or increasingly by financing relationships between chipmakers, cloud providers, model developers and data-centre operators.
Nvidia’s shares fell 5% on Monday following the financing report. AMD declined 5.2%, while Micron Technology lost 2.3%.
The AI industry is encountering a credit problem
The first phase of the generative AI boom was largely funded by the considerable cash flows of Microsoft, Alphabet, Amazon and Meta. Those companies could build data centres and purchase processors without relying entirely on outside lenders.
The next stage is more demanding. Frontier model developers require access to increasingly large quantities of computing capacity, but companies such as OpenAI remain privately held and unprofitable. They must therefore rely on partnerships, leases, equity investments, supplier financing and debt to support their expansion.
Reuters Breakingviews reported that analysts at Morgan Stanley expect hyperscaler investment to reach approximately $875 billion in 2026. It also noted that free cash flow is expected to weaken at several major technology companies as infrastructure commitments increase. These are analyst estimates rather than confirmed future expenditure, but they help explain why investors are scrutinising the source of the industry’s capital more closely.
The central concern is not simply that AI infrastructure is expensive. It is that the facilities currently being proposed may take years to complete, while the economics of models and computing hardware are changing rapidly.
A data centre designed around current assumptions about model size, electricity consumption and chip demand could face a different market by the time it becomes operational.
China has added a second source of pressure
Financing is only part of the story. Investors are also reassessing the competitive position of established semiconductor companies as China develops lower-cost AI models, domestic memory capacity and its own chipmaking equipment.
The sell-off followed reports that China had begun manufacturing domestically developed deep-ultraviolet lithography systems. Such equipment is used during semiconductor production and has historically been dominated by a small group of international suppliers.
The claim originated in reporting by The Information and has not been independently established through detailed public technical documentation. It should therefore be treated as a reported development rather than proof that China has matched the most advanced systems produced by companies such as ASML.
Even so, the report affected market expectations. Investors are not required to believe that Chinese systems are already technologically equivalent to conclude that stronger competition could reduce prices, compress margins or weaken the long-term demand assumptions built into semiconductor valuations.
The successful market debut of Chinese memory-chip company CXMT also demonstrated the capital available to support China’s domestic semiconductor industry. CXMT raised at least $8.6 billion in Shanghai, although its shares became volatile after an exceptional first-day rise.
What the sell-off does not prove
The scale of Tuesday’s declines may encourage comparisons with the bursting of previous technology bubbles, but one trading session cannot establish that the AI investment cycle has ended.
Demand for high-bandwidth memory, advanced processors and data-centre equipment remains substantial. Major cloud providers are still expanding their AI services, while businesses continue to adopt models for software development, research, customer operations and automation.
Some of the selling may also reflect the structure of the market. Semiconductor shares had risen rapidly, leaving investors heavily exposed to a relatively small group of companies. When expectations change, crowded positions and leveraged investments can amplify movements in both directions.
Morningstar equity analyst Jing Jie Yu described the Asian decline as a likely knee-jerk reaction and argued that the competitive position of global semiconductor leaders was unlikely to be immediately displaced.
The more defensible conclusion is that investors are changing the question they ask about AI. For much of the boom, the primary question was how quickly infrastructure could be constructed. The new question is whether the companies funding it can generate sufficient returns.
What happens next
The next evidence will come from the companies spending the money.
Amazon, Meta, Apple and Microsoft are due to report earnings this week. Investors will examine cloud revenue, capital expenditure, margins and management guidance for evidence that AI products are producing returns proportionate to their infrastructure costs.
The market will also watch for any response from Nvidia, OpenAI or SoftBank concerning the reported Ohio data-centre financing. Confirmation, modification or denial of the reported arrangements could materially change how investors assess Nvidia’s financial exposure.
The US Federal Reserve’s interest-rate decision will provide another test. Higher borrowing costs would make the debt-funded construction of data centres more expensive and could place further pressure on projects with long development periods and uncertain revenue.
The semiconductor sell-off does not show that artificial intelligence has stopped growing. It shows that the financial assumptions supporting that growth are being challenged.
Investors are increasingly examining who is providing the capital, who carries the risk and whether revenue from AI products can keep pace with the cost of chips, data centres and energy. At the same time, stronger Chinese competition is undermining the assumption that today’s market leaders will retain their position indefinitely.
The AI boom has not necessarily reached its end. It has entered a more demanding stage, in which technological ambition is no longer enough. The companies building the infrastructure must now demonstrate how it will pay for itself.
Frequently asked questions
Why are AI chip shares falling?
AI chip shares are falling because investors are reconsidering the sustainability of enormous infrastructure spending, the use of debt and supplier financing, elevated semiconductor valuations and increasing competition from China. The decline does not necessarily indicate that demand for AI has disappeared.
How much did Samsung Electronics and SK Hynix fall?
Samsung Electronics closed 13.4% lower on 28 July 2026, while SK Hynix declined 14.7%. Their losses contributed to a fall of approximately 10.8% in South Korea’s Kospi index.
Is Nvidia investing $250 billion directly in OpenAI?
No completed $250 billion investment has been confirmed. Nvidia is reportedly discussing a financing guarantee connected to an OpenAI data-centre project. Reuters said it could not independently verify the report, and the companies had not commented when its article was published.
Does the sell-off mean the AI boom is over?
Not necessarily. The sell-off reflects changing investor expectations and greater scrutiny of spending, financing and competition. Major technology companies are still investing heavily in AI, but markets increasingly expect evidence that this expenditure will generate sustainable revenue and profit.
sources
Reuters, Asian chip stocks slide as China competition fears rattle AI trade, 28 July 2026.
Reuters, Nasdaq futures drop on AI chip worries ahead of pivotal earnings, 28 July 2026.
Reuters, Nvidia in talks with OpenAI to guarantee $250 billion financing for data center, WSJ reports, 26 July 2026.
Associated Press, South Korea’s Kospi share index plunges 11% on selling of chipmaking stocks, 28 July 2026.
Reuters Breakingviews, Cloudmaxxing sucks Nvidia into dangerous game, 27 July 2026. Analysis source used only for clearly attributed financial interpretation.