In its latest annual report, the Bank for International Settlements (BIS) warned that the current wave of investment around artificial intelligence (AI) carries significant economic risks. The Basel‑based institution — concerns of which were reported by Reuters — is not primarily questioning the scale of spending, but what may happen when that spending reaches its peak.
What the market looks like now: key figures
The BIS notes that momentum remains extraordinary: in the second quarter of 2026, shares of US semiconductor manufacturers rose by a record 75 percent. That rally has been driven in part by ever‑larger investment plans from the major cloud providers, which in turn have created supply bottlenecks and chip shortages due to competitive pressures.
The report states that the top five investors’ AI spending this year approaches $1,000 billion, and Goldman Sachs estimates cumulative spending could reach $7,600 billion by 2031. Some market participants reject the “bubble” label: Masayoshi Son, head of SoftBank, was quoted as saying last week that calling AI a bubble is wrong because, in his view, the technology is only at the beginning.
At the same time, some investors are becoming less likely to automatically view the rally as a bubble. A recent Deutsche Bank quarterly client survey found that sentiment of bubble risk for the group known as the “Magnificent Seven” is at its lowest since 2021, although perceived risks remain higher across the broader US technology sector.
Many semiconductor companies are showing real financial gains: Micron Technology, currently valued at $1,250 billion, has seen its share price rise to more than three times its March level, while revenue expectations have increased at a similar pace. As a result, Micron’s forward price/earnings ratio has stayed roughly flat and at eight times is less than half what it was two years ago. Valuations for Broadcom and Qualcomm also remain moderate in historical terms.
BIS’s main concerns: sustainability, capacity and demand
The BIS’s principal worry is the sustainability of the investment tempo. Intense competition among a few players could push large amounts of capital into projects with uncertain returns. If returns disappoint, spending could be sharply curtailed, turning the current investment boom into a prolonged downturn.
The analysis also singles out capacity bottlenecks in energy production, electrical grids and memory chips as particular risks that could hamper AI expansion.
Perhaps the most striking warning is that AI could ultimately “eat itself.” If the technology replaces human labour to the extent its proponents claim, a growing share of income could flow away from wages toward further AI investment. That shift could reduce workers’ share of national income toward zero, leaving fewer people able to buy the output of the economy. As the BIS puts it, productivity gains may not be halted by technological limits, but by a lack of demand that would justify additional capacity expansion.
Why this matters
The BIS’s warning matters because the size and pace of AI investment can materially affect global growth, labour markets and financial stability. Unsustainably rapid investment could create vulnerabilities not only for tech firms but for the wider economy if a reversal occurs.
The BIS does not oppose technological progress; rather, it stresses that policymakers and investors need to account for capacity constraints, return uncertainty and the risk that falling demand could undermine further investment.
