Financing the AI Boom: From Cash Flows to Debt – BIS Bulletin 120

Financing the AI Boom: From Cash Flows to Debt – BIS Bulletin 120

Key Takeaways

  • AI‑related investment now accounts for a substantial share of US GDP growth, reaching about 5% of GDP and contributing roughly 0.4 percentage points to quarterly growth.
  • The scale of AI infrastructure spending is pushing firms to replace internal cash‑flow financing with external debt, with private‑credit funds emerging as a fast‑growing lender.
  • Macro‑financial stability risks are moderate, but the sustainability of the boom hinges on AI firms delivering the high earnings that equity markets currently price in; a mismatch between debt pricing and equity valuations could trigger sharp corrections.

1. AI Investment Is Becoming a Major Driver of US Economic Growth

  • Investment composition: AI‑related spending includes data‑centre construction, data‑centre equipment (≈ 3 × construction cost), IT‑manufacturing facilities (chips, hardware), and broader IT hardware/software upgrades.
  • GDP share: By mid‑2025, combined data‑centre and IT‑manufacturing investment equaled ~1 % of US GDP; total IT‑related investment (including other equipment and software) rose to ~5 % of GDP, surpassing the peak of the 2000 dot‑com boom.
  • Growth contribution: Over the three years following 2022, AI‑related capital expenditures added an average of 0.4 pp to quarterly GDP growth. Total IT investment has accounted for almost half of recent GDP growth, cushioning the impact of trade‑tariff shocks.
  • Future outlook: Forecasts suggest annual data‑centre spending could increase by $100‑$225 bn over the next five years, lifting data‑centre investment to 0.8‑1.3 % of GDP (up from 0.5 % today).

"AI‑related investment has emerged as an important driver of GDP growth in the United States. From a negligible contribution before 2022, expenditures on semiconductor manufacturing facilities and data centres have contributed on average 0.4 percentage points to GDP growth over the subsequent three years." – BIS Bulletin

2. The Financing Shift: From Internal Cash Flows to Debt

  • Historical financing: Leading AI firms (e.g., Alphabet, Amazon, Meta, Microsoft, Oracle) have traditionally financed investments with low leverage, relying on strong operating cash flows.
  • Current pressure: Capital expenditures have outpaced free cash flow, forcing firms to seek external funding. Equity issuance is constrained by volatile AI valuations and narrow issuance windows.
  • Debt uptake: Companies are turning to corporate bonds, leasing, and loans to match the long life of data‑centre assets. Record bond issuances have been reported, but construction‑ and power‑risk profiles sometimes push financing outside traditional bank channels.
  • Private credit’s role:
    • Private‑credit funds (non‑bank lenders) have grown from $100 bn in assets in 2010 to >$2.2 tn in 2024.
    • Outstanding loans to AI‑related firms have risen from near zero to >$200 bn, representing ~8 % of total private‑credit loan volume.
    • Average AI‑sector loan size is $169 mn (vs $90 mn for other sectors); loan maturities (≈ 4.7 years) and rate spreads (≈ 6.2 pp) are comparable to non‑AI loans.
    • About 20 % of private‑credit funds now have AI exposure, up from 5 % in 2010, though AI loans still represent only ~5 % of an average fund’s portfolio.

"Private credit funds are mostly closed‑end structures that lock in institutional capital for the life cycle of their loan portfolios (around four to eight years), mitigating liquidity and maturity transformation risks." – BIS Bulletin

3. Financial‑Stability Implications

  • Higher leverage: AI firms moving from cash‑flow to debt increase corporate leverage, amplifying potential shocks to both firms and financial intermediaries.
  • Risk of hidden leverage: Some financing structures may keep leverage off‑balance‑sheet, but the risk does not disappear.
  • Debt‑equity disconnect: Private‑credit spreads for AI loans are similar to those for non‑AI borrowers, suggesting lenders view AI risk as average, while equity markets price AI firms at highly elevated multiples. This divergence could lead to simultaneous equity and debt market corrections if AI returns fall short.
  • Historical context: The AI boom’s size (~1 % of GDP) is comparable to the US shale boom and half the magnitude of the 1990s dot‑com surge. Past investment booms have often been followed by GDP slowdowns of >1 pp, but they rarely produce lasting higher growth rates.

"If a decline in AI investment were to come with a significant stock‑market correction, negative spillovers could be larger than previous booms suggest." – BIS Bulletin

4. Outlook and Open Questions

  • Productivity gains: Commenters on Hacker News ask whether AI investment will translate into macro‑productivity improvements, similar to past automation waves.
  • Profitability evidence: Some users note a lack of clear profit‑growth examples among AI‑heavy firms (e.g., Duolingo’s flat earnings despite AI adoption), highlighting the uncertainty around AI’s bottom‑line impact.
  • Financing developments: Questions remain about upcoming AI‑related IPOs (e.g., Anthropic) and how they will affect the debt‑equity balance.
  • Scenario range: The BIS report presents “medium” and “high” demand scenarios; community members wonder whether a more pessimistic baseline is missing.

5. Conclusion

AI‑related capital spending is reshaping the US economy, now representing a sizable share of GDP and prompting a fundamental financing shift from internal cash flows to external debt. Private credit has become a pivotal source of funding, but the rapid build‑up of leverage, combined with a stark gap between debt pricing and equity valuations, creates a moderate but real risk to financial stability. The ultimate macroeconomic payoff will depend on whether AI firms can deliver the high earnings that markets expect; otherwise, the sector could experience a sharp correction with broader spillovers.

Sources