For several days now, the stock markets of South Korea and the United States have kept falling, hitting first and foremost the tech giants tied to artificial intelligence. The panic has no single clear cause, but most experts believe investors have begun to doubt AI’s prospects and the profits it might generate. Meduza explains how AI developers and chipmakers have already suffered, and why companies are starting to rehire the workers they laid off because of AI.
The sell-off began on June 23, 2026, with the collapse of shares in South Korea’s Samsung and SK Hynix, the world’s two largest makers of memory chips. They fell more than 12%, dragging down the entire market: the Kospi index plunged 10%, and trading had to be halted for 20 minutes to curb the panic, CNN reports.
The decline spread to the United States, Reuters notes. The Nasdaq and S&P 500 fell 2.2% and 1.4%, their worst showing in more than a week. Micron shares lost 13%, Nvidia more than 4% (wiping out $200 billion in market value), and Intel and AMD between 5.8% and 9.4%. Even Elon Musk took a hit: shares in SpaceX, which went public in mid-June, dropped from $225 to $156, and according to Bloomberg, he is no longer a trillionaire.
The Panic Has No Single Cause. But AI Spending Worries Investors More and More
The most widespread theory is that investors are alarmed by the constantly rising spending on AI infrastructure, which is financed largely through debt. This is indirectly confirmed by the fact that the blow landed first on semiconductor makers and AI developers. The Kospi index is up 90% this year, CNN notes and even slight doubts among traders about whether that growth will continue can push them to dump their shares.
“Some of the recent AI-related news is raising questions about all this spending, capital investment, and capacity build-out in the semiconductor industry,” a senior manager at the US investment firm Globalt told Reuters. On the evening of June 22, Goldman Sachs analysts warned clients that the market had become “more vulnerable to any news that calls into question” optimistic estimates of future profits, Politico confirms. Among additional factors, analysts surveyed by Reuters cite uncertainty over the US-Iran deal and expectations of a Fed rate hike.
Many Companies Are Rethinking AI’s Efficiency
For now, few expect a significant drop in demand for AI, Politico writes. But some companies among the first to adopt the technology are already limiting its use, the Financial Times notes: AI costs too much and is used too often. Walmart and Uber are shifting employees to cheaper models, and even Amazon and Meta have introduced restrictions.
“Consumers and businesses were taught that AI is cheap or even free, but that is definitely not the case,” Kosti Perrikos of Deloitte told the FT. The situation is compounded by autonomous AI agents, which are hungry for money and energy, and by the shift among market leaders, Anthropic and OpenAI, toward token-based billing. According to Goldman Sachs, by 2030 the use of agents will drive a 24-fold increase in token consumption and worsen the chip shortage.
Companies That Rushed to Replace People With AI Are Rehiring Them
A new trend is emerging in the labor market: companies that hurried to replace some of their staff with artificial intelligence are gradually bringing people back, usually within 6 to 12 months, Forbes notes. According to Forrester, 55% of these executives will regret the decision within 18 months, and per Robert Half, nearly a third of companies have already taken back some of their workers.
The problem is not only cost. In many areas, AI still falls short of humans: it handles standard requests but fails to recognize when human intervention is needed. It can write ad copy but doesn’t grasp customer needs the way a seasoned marketer does; it can process transactions but won’t spot anomalies the way an experienced finance professional would, Forbes journalists observe.
Many companies have realized AI’s inefficiency in certain areas, along with the ever-rising costs of that inefficiency. This doesn’t mean the bubble has started to burst. But all the players, and above all the developers themselves, will have to set aside endless scaling for a while and think about optimization.
