Who Pays for AI: Financing the Next Investment Cycle

AI is driving one of the largest capital-spending cycles in decades, prompting questions about how the investment will be financed and what risks and opportunities it may create across credit markets. PIMCO’s Lotfi Karoui, Goldman Sachs’ Amanda Lynam and Barclays’ Brad Rogoff discussed the topic at Bloomberg’s Future of Fixed Income 2026 conference; the article provides no specific financing figures or market reaction.

Analysis

The investable issue is not AI demand; it is who ultimately bears the funding and execution risk. If cash-rich platform companies fund buildout internally, the credit effect may remain modest while suppliers, power providers and infrastructure owners capture incremental financing demand. If debt, leases, joint ventures or customer-backed structures absorb more of the spend, headline corporate leverage could understate the risk transferred to bondholders and private-credit investors. A further-order constraint is power and grid access: delayed energization can leave data-center assets producing interest expense before revenue, while shifting bargaining power toward utilities and equipment providers.

Near term, conference commentary alone is not a catalyst for broad credit exposure. Over 1–3 months, track issuer capex guidance against free cash flow, debt issuance, lease commitments, project cancellations and data-center utilization; spreads may react more to funding mix and returns on invested capital than to capex totals. Over 6–18 months, poor utilization or slower monetization could pressure highly levered project structures and lower-quality borrowers, while investment-grade issuers may retain financing flexibility. The contrarian risk is treating AI capex as uniformly credit-positive: spend can support suppliers while weakening the credit profile of the party funding it. No issuer-level trade is justified without deal terms, leverage and valuation data.

Key Decisions for Investors

• No trade on the event itself. Avoid extrapolating aggregate AI investment into a blanket long on credit; first separate internally funded hyperscaler capex from debt-, lease- and project-financed exposure.
• Set an alert for material revisions to free-cash-flow conversion, debt issuance, lease obligations, project financing terms or utilization at data-center and power-infrastructure borrowers. These are more decision-useful than capex announcements alone.
• If financing shifts materially toward debt-funded projects, assess a relative-value position favoring higher-rated, senior claims over subordinated or highly levered data-center/project credit; require verified leverage, contractual protections and spread compensation before entry.
• Falsify the credit-risk thesis if major borrowers sustain investment-grade metrics and cash generation while projects reach utilization on schedule; strengthen it if funding needs rise faster than cash flow, spreads widen selectively, or construction/power delays push out revenue.