The AI investment boom that supported the global economy in 2025 is itself becoming a source of macrofinancial risks, according to the Bank for International Settlements’ annual report.
The report cites debt-funded AI infrastructure, stretched company valuations, the rise of private credit, and opaque deals among hyperscalers, chip manufacturers and labs.
AI aided growth but heightened vulnerabilities
According to BIS, in 2025 the global economy proved resilient to tariff and geopolitical shocks. Optimism around artificial intelligence was one factor: it supported capital spending, trade in intermediate goods, and accommodative financial conditions.
However, in 2026 the set of risks broadened. The bank’s analysts identified four pressure points:
- the risk of persistent inflation;
- doubts over the sustainability of AI investment;
- rising financial vulnerabilities;
- deteriorating fiscal positions.
“Optimism around AI may not persist, despite the promise of future productivity gains,” the report said.
BIS estimates the current capex surge may prove unsustainable if production hits bottlenecks. The bank cited electricity, advanced semiconductors and networking equipment.
Capex outpaces cash flow
The report estimates AI-related capital expenditures by the five largest hyperscalers at more than $1 trillion in 2025–2026 combined. According to BIS, these commitments already exceed earnings and free cash flow at some companies, forcing them to tap debt financing.
“Disappointment with returns could trigger a sudden pullback in financing and turn the capex boom into a protracted investment slump with potential consequences for financial conditions,” the analysts warned.
The bank compared the current cycle with earlier periods of technological overheating: the canal mania of the 1830s, the British railway mania of the 1840s, the electrification euphoria of the late 1920s, and the dot-com boom of the late 1990s. A common feature, BIS said, was a genuine technological breakthrough that attracted more capital than subsequent commercial results justified.
Circular financing emerged as a distinct risk
According to the report’s authors, credit markets, private credit, and contractors building data centers, power infrastructure and related facilities could prove vulnerable. If hyperscalers slow or halt aggressive capex, borrowers across the supply chain could lose the revenue needed to service their debt.
Another vulnerability is opaque private deals within the AI sector. This refers to circular financing in which hyperscalers or chipmakers take stakes in AI labs and cloud providers, which in turn assume multi-year commitments to buy chips or computing resources.
“The terms of such deals are typically poorly disclosed, creating a risk of rehypothecation of the same asset,” the document said.
BIS also noted that a correction in the United States could quickly spread worldwide. U.S. equities account for about 64% of the MSCI Global index, so a repricing of AI companies could affect household wealth, consumption and global financial conditions.
Memory market shows the scale of demand
In parallel, the memory-chip market has become a separate indicator of stress in the AI supply chain. On June 24, U.S. semiconductor manufacturer Micron reported record revenue for the third quarter of fiscal 2026 — $41.46 billion. In prepared remarks for the investor call, the company said it had signed 16 strategic customer agreements. Such contracts are typically five years — from 2026 through the end of 2030; automotive agreements are generally three years.
According to Micron, for agreements already in place, including those signed after quarter-end, RPO is about $100 billion. The company emphasized that this figure is not equal to all expected future revenue. The manufacturer also expects $22 billion in customer deposits and related financial obligations under signed agreements. Of that, about $18 billion will be cash deposits.
“We do not presently have clarity on when [memory] supply can catch up to growing demand,” said Micron CEO Sanjay Mehrotra.
In the quarter, DRAM generated a record $31.3 billion for Micron, or 76% of total revenue. The company linked the increase to tight industry conditions, a favorable sales mix and rising prices.
At the same time, rising chip prices have already become the subject of litigation. On June 25, a class-action antitrust lawsuit was filed in the Northern District of California against Samsung Electronics, Samsung Semiconductor, SK Hynix, SK Hynix America and Micron Technology.
According to MLex, the plaintiffs allege that the three largest DRAM producers coordinated to restrict supply and raise prices. The claims also mention a shift in priorities toward more expensive AI products, including HBM memory.
Earlier, rising demand for AI infrastructure has become a problem for tech giants due to spending on memory, electricity and data centers.
In November 2025, media reported plans by Micron to invest $9.6 billion in AI memory chip production in Japan.
