The investment boom surrounding artificial intelligence, which supported the global economy in 2025, is now becoming a source of macro-financial risks. This is highlighted in the annual report from the Bank for International Settlements (BIS).

The report discusses debt financing for AI infrastructure, inflated company valuations, increased private lending, and opaque transactions among hyperscalers, chip manufacturers, and laboratories.

AI Boosted Growth but Increased Vulnerabilities

According to the BIS, the global economy demonstrated resilience to tariff and geopolitical shocks in 2025. One contributing factor was the optimism surrounding artificial intelligence, which bolstered capital expenditures, trade in intermediate goods, and favorable financial conditions.

However, by 2026, the set of risks had expanded. Bank analysts identified four pressure points:

  • the threat of persistent inflation;
  • concerns about the sustainability of AI investments;
  • increased financial vulnerabilities;
  • deteriorating fiscal positions.

“Optimism around AI may not be sustained, despite the promise of future productivity growth,” the report states.

The BIS estimates that the current growth in capital expenditures could prove unsustainable if production encounters bottlenecks, including electricity, advanced semiconductors, and networking equipment.

Capital Expenditures Outpace Cash Flow

The report estimates that capital expenditures on AI by the five largest hyperscalers will exceed $1 trillion in 2025-2026 (a combined figure for two years). According to the BIS, these commitments already outstrip profits and free cash flow for some companies, forcing them to seek debt financing.

Growth of AI capital expenditures by hyperscalers compared to previous investment booms. Source: BIS Annual Economic Report 2026.

“Disappointment in returns could trigger a sudden pullback in financing and turn the capital expenditure boom into a prolonged investment slump with potential consequences for financial conditions,” analysts warned.

The bank compared the current cycle to previous technological overheating episodes: 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 of these episodes, according to the BIS, was a genuine technological breakthrough that attracted more capital than could later be justified by commercial results.

Circular Financing Emerges as a Separate Risk

The report's authors believe that credit markets, private lending, and contractors building data centers, energy facilities, and related infrastructure could be vulnerable. If hyperscalers slow down or halt aggressive capital expenditures, borrowers throughout the supply chain may lose the revenue needed to service their debt.

Another vulnerability is the opaque private deals within the AI sector. This involves circular financing, where hyperscalers or chip manufacturers acquire stakes in AI laboratories and cloud providers, which in turn take on long-term commitments to purchase chips or computing resources.

“The terms of such deals are often poorly disclosed, creating the risk of re-pledging the same asset,” the document states.

The BIS also reminded that a correction in the U.S. could quickly spread worldwide. American stocks account for about 64% of the MSCI Global Index, so a reevaluation of AI companies could impact household wealth, consumption, and global financial conditions.

Memory Market Reflects Demand Scale

Meanwhile, a separate indicator of tension in the AI supply chain has been the memory chip market. On June 24, American semiconductor manufacturer Micron reported record revenue for the third quarter of fiscal 2026—$41.46 billion. In prepared remarks for an investor call, the company stated that it had secured 16 strategic customer agreements. Typically, these contracts are set for five years—from 2026 to the end of 2030; automotive agreements usually have a three-year term.

According to Micron, for the agreements already signed, including those finalized after the quarter ended, the remaining performance obligation (RPO) stands at about $100 billion. The company emphasized that this figure does not represent the total expected future revenue. Micron also anticipates $22 billion in customer deposits and related financial obligations from already signed agreements, with approximately $18 billion attributed to cash deposits.

“We currently have no clarity on when memory [chip] supply will catch up with growing demand,” stated Micron CEO Sanjay Mehrotra.

In the reporting quarter, DRAM generated a record $31.3 billion for Micron, accounting for 76% of total revenue. The company linked this growth to tight industry conditions, favorable sales structure, and rising prices.

Additionally, the increase in chip prices has already become the subject of a lawsuit. 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, plaintiffs allege that the three largest DRAM manufacturers coordinated to restrict supply and raise prices. The claims also mention a shift in priorities towards more expensive AI products, including HBM memory.

Earlier, it was reported that the rising demand for AI infrastructure has posed challenges for tech giants due to costs associated with memory, electricity, and data centers.

In November 2025, it was reported that Micron planned to invest $9.6 billion in memory chip production for AI in Japan.