Financing the AI Boom: From Cash Flows to Debt

AI Investment is Shifting from Internal Cash Flows to Debt

Artificial Intelligence (AI) investment is surging in both nominal terms and as a share of GDP, necessitating a fundamental shift in how IT firms fund their growth. While leading technology companies historically relied on highly profitable operating cash flows to finance expansion, the sheer scale of current infrastructure needs—specifically data centers, servers, and power stations—has exceeded the limits of internal funding. Consequently, firms are increasingly turning to external debt, including corporate bonds and loans, to align financing maturities with the long-term economic life of AI assets.

The Macroeconomic Impact of AI Infrastructure

AI-related investment has become a material driver of US GDP growth since 2022. This growth is primarily concentrated in two areas: the construction of data centers and the expansion of IT manufacturing facilities (such as semiconductor plants), the latter of which was partially spurred by the US CHIPS and Science Act.

Key macroeconomic data points include:

GDP Share: By mid-2025, expenditures on IT manufacturing facilities and data centers reached 1% of GDP. Total IT-related investment, including software and other equipment, rose to 5% of GDP, surpassing the peak seen during the dot-com boom of 2000.
Growth Contribution: Since 2022, semiconductor manufacturing and data center expenditures have contributed an average of 0.4 percentage points to US GDP growth.
Future Projections: Annual spending on data centers alone is projected to increase by $100 billion to $225 billion over the next five years, potentially raising data center spending to between 0.8% and 1.3% of GDP.

The Rapid Rise of Private Credit in AI Financing

Because AI projects involve high construction risks, power availability concerns, and tenant concentration, they often fall outside the scope of traditional bank and bond financing. This has led to a surge in private credit—non-bank credit extended by specialized investment funds through directly negotiated deals.

Private credit has become a critical mechanism for AI funding due to its bespoke covenant structures and faster execution. The growth in this sector is stark:

Volume Growth: Outstanding direct loans to AI-related companies have grown from near zero in 2010 to over $200 billion today.
Market Share: AI-related loans now account for nearly 8% of total outstanding private credit volumes, up from less than 1%.
Future Outlook: Based on projected investment growth, outstanding private credit to AI firms could reach $300 billion to $600 billion by 2030.

Despite the volume increase, the average private credit fund's exposure remains modest, with AI loans representing about 5% of total volumes for the average fund.

Financial Stability Risks and the "Valuation Schism"

The transition to debt-based financing introduces systemic vulnerabilities. If the expected returns on AI investments fail to materialize, the resulting leverage could amplify financial shocks. There are specific concerns regarding "circular financing" within the AI ecosystem and the use of off-balance-sheet structures that may mask the true extent of leverage.

A significant tension exists between debt and equity markets, described as a valuation schism:

Debt Market Pricing: Loan spreads for AI firms are similar to those of non-AI firms (approximately 6.1 to 6.2 percentage points), suggesting lenders view AI loans as having average risk.
Equity Market Pricing: AI equity valuations are exceptionally high, implying outsized future returns.

This discrepancy suggests that either lenders are underestimating the risks of AI infrastructure or equity markets are overestimating the future cash flows AI will generate.

Historical Context and Potential for Correction

While the current boom is significant, it is smaller relative to GDP than several historical investment bubbles. For example, the AI boom (at ~1% of GDP) is similar to the US shale boom of the mid-2010s and half the size of the dot-com boom's IT investment. It is significantly smaller than the commercial property boom in 1980s Japan or the mining boom in 2010s Australia.

However, historical data indicates that the end of such booms often correlates with a GDP growth slowdown of more than 1 percentage point on average. The dot-com crash provides a cautionary example where a relatively small boom relative to GDP still caused a significant contraction. If a decline in AI investment is accompanied by a sharp stock market correction and a credit market spillover from hidden leverage, the macroeconomic impact could be substantial.