Questions raised over continuation of AI boom

Amid the frenzy on Wall Street, the big questions being raised in financial circles are whether the AI-based bubble is about to burst, what factors may trigger its collapse and what will be the consequences if that happens. The past weeks have been marked by increasing volatility: a surge one day followed by a significant fall, and then another surge. The South Korean market, which is highly dependent on the chipmaking giants, Samsung and SK Hynix, major suppliers of chips used in AI data centres, is one of the most graphic expressions of this process. Earlier this month, Samsung shares plunged by as much as 10 percent in a single day. But the share slide did not come on the back of bad results. Rather, it came after the company had reported a third straight quarter of record profits. Its forecast for the second quarter operating profit was $58.4 billion, up 19-fold from the figure a year ago and exceeding the profit for all of 2025. The dramatic fall in the share price, in the words of the Financial Times (FT), was “a sign of investor concerns about returns on massive AI investments.” The volatility continued last week. The South Korean Kospi index plunged 16 percent on Monday, rebounded by more than 6 percent on Wednesday, then fell again by 6 percent on Thursday. So far this year, the stock exchange has been closed 37 times because of volatile trading compared to three times last year. The memory chip industry is notoriously volatile, going through boom-bust cycles. But the rollout of the AI data centres is a new factor. Samsung shares have fallen by a third since their June high, SK Hynix has dropped 40 percent and shares of the US-based chipmaking firm Micron have dropped more than 30 percent. Reporting on the fall in semiconductor stocks yesterday, the FT cited remarks by Kwon Seok-joon, a professor at a Seoul university who warned of a significant oversupply problem if AI demand faltered. “Chipmakers are making these plans on the assumption that demand from AI centres will remain strong for the next two to three years. But memory demand will fall if returns on AI investments fail to meet expectations.” It is not just chipmakers that are dependent on the continuation of the AI boom, but Wall Street and major global equity markets. It is estimated that about 60 percent of the gains in the S&P 500 index have been driven by a handful of AI-related tech giants, including companies such as Nvidia, Microsoft, Meta and Broadcom. The 10 top companies in the index account for 41 percent of the S&P’s market capitalisation. This high degree of concentration and its role in driving the market higher is the cause of increasing concern in sections of the financial press. In a column published at the weekend, entitled “The next crash: why this time might not be different,” FT economics commentator Martin Wolf began by recalling a statement by Irving Fisher, whom he described as “one of the greatest American economists,” in the autumn of 1929. According to Fisher: “Stock prices have reached what looks like a permanently high plateau.” As Wolf noted, it turned out to be “one of the most incorrect forecasts ever made,” because in short order the US and global markets were hit by the “Great Crash” followed by the “Great Depression.” “Why might this story be relevant today?” he continued. “The answer is that the valuation of US stocks is even higher than in September 1929.” Boosters of AI, as always, maintain that “this time is different” pointing to the massive transformations flowing from the technology. There is no doubt that AI is a transformative technology, as Wolf noted, but “the history of investment surges underpinned by profound innovations does not show that the latter guarantee huge profits.” In fact, that history shows that “over-investment, destructive competition, waves of bankruptcies and then painful consolidation are standard features of such episodes, from the railway booms of the 19th century to the internet boom of the 1990s.” As for Wall Street, Wolf concluded that with the domination of a handful of AI-linked stocks, “the current extraordinary valuations of the market depend on the continuation of the AI boom.” There are significant questions being raised about that prospect because the key issue within the capitalist profit system is not whether AI can boost the productivity of labour, which it undoubtedly can, but what is the rate of return generated by trillions of dollars laid out by the hyperscalers on AI data centres. So far, they are loss making ventures. The expectation is that massive profits will be made as AI is taken up across industry. But such success depends on the capacity of the firms taking it up to use AI to reduce costs, largely through a major elimination of labour. It is by no means guaranteed that the US firms presently leading the development of AI, such as Anthropic and OpenAI, will emerge as the winners in the ruthless fight for markets and profit. Their business model assumes that they will be able to stay at the head of AI development and thereby accrue the lion’s share of the profits. However, this assumption is already under challenge with the development of AI firms in China which can produce results comparable with their powerful US rivals at a much lower cost. AI users may turn to Chinese models as being adequate for their purposes, even though they are not top of the range, because they are cheaper to run. This dynamic was seen on Friday when Chinese start-up firm Moonshot released a new model, Kimi K3, which it said performed as well as leading models from OpenAI and Anthropic on some key tasks. Its release contributed to the market selloff on Friday. Drawing out the wider implications of this development, the New York Times noted: “The release of a free, open-source Chinese AI model that could rival the performance of costly computing-intensive systems from Silicon Valley’s best-funded companies has reignited concerns over whether the industry’s enormous spending spree on data centres is justified.”