US major banks' loan commitments to artificial intelligence-related companies have reached approximately 25% of their Tier 1 capital, raising concerns that aggressive AI investments by US mega-cap tech firms could spill over into a global financial risk.
Tier 1 capital, the sum of a bank's paid-in capital and retained earnings, is a key indicator of loss absorption capacity.
According to the Bank of Korea and the Federal Reserve Bank of Chicago on Tuesday, the outstanding balance of AI-related corporate loans at major US banks stood at $150 billion (approximately 227 trillion won) as of the end of last year. This represents 0.8% of major US banks' total assets. Of this, loans for data center construction totaled $14.9 billion (approximately 22 trillion won).
While the outstanding loan balance may appear modest, the picture changes when looking at loan commitments. AI-related industry loan commitments at major US banks reached $450 billion (approximately 683 trillion won) last year, an 80% surge from a decade ago. This represents about 25% of Tier 1 capital, a key metric used by the Bank for International Settlements (BIS) to assess bank soundness. Given that borrowing levels continue to grow—with Anthropic recently pursuing a $36 billion loan to purchase AI chips—the soundness ratio is likely to have deteriorated further.
Notably, loan commitments to high-risk borrowers rated B or below have reached $50 billion (approximately 75 trillion won). This amounts to roughly a quarter of the $191 billion in loans US major banks have committed to software-related companies.
The concern is that if AI bubble fears materialize, large-scale loans could boomerang back as non-performing assets. A decline in investment and demand in downstream industries such as AI software development, along with weakening energy and semiconductor demand from improvements in AI model efficiency, could increase the likelihood of these loans turning sour.
The Chicago Fed noted that "the loan balance, at 0.8% of total assets, is unlikely to have a meaningful impact on soundness," while pointing out that "loan commitments approach 25% of Tier 1 capital, raising concerns about large-scale losses if risks materialize." It added, "Loans to companies rated B or below in particular have a high probability of non-repayment if high interest rates persist," and "considering exposures to non-bank financial institutions and private credit lending, where data is difficult to obtain, risks could spread across the entire financial system."
Additionally, prolonged Middle East conflicts have been cited as a risk factor, raising concerns over potential delivery delays from major Southeast Asian countries that serve as key suppliers of AI equipment to the US. Southeast Asia has emerged as a critical supplier of servers and networking equipment needed for AI infrastructure construction following US efforts to contain China. In fact, US imports of goods from Southeast Asia rose 46% year-on-year last year. However, since Southeast Asia is heavily dependent on energy imports from the Middle East, a prolonged Middle East war could force factory shutdowns due to power shortages, potentially affecting US supply chains.
Meanwhile, an analysis showed that AI investment contributes about 40% to US economic growth. According to the Federal Reserve Bank of St. Louis, US real gross domestic product (GDP) growth from the first to third quarters of last year was 2.51% (annualized), with AI-related investment contributing 0.97 percentage points. This means AI-related investment accounted for approximately 39% of US growth during the period. In detail, information processing equipment investment drove 0.42 percentage points of growth, software 0.35 percentage points, research and development (R&D) 0.13 percentage points, and data centers 0.07 percentage points.
This exceeds the contribution of dot-com bubble-related investment to US economic growth in the early 2000s. In 2000, US real GDP growth was 2.94%, with dot-com bubble-related investment contributing 0.81 percentage points (28%). The analysis suggests that aggressive large-scale AI investments by the top 5% of mega-cap US firms are strongly boosting the growth rate.
The St. Louis Fed explained that "investments by big tech companies are assessed to have made the most decisive contribution to the US economy maintaining solid growth last year and this year."
