JPMorgan’s Mulvihill on AI Concentration Concerns

JPMorgan Asset Management strategist Aaron Mulvihill said credit demand and originations for the AI and data-center buildout have not slowed, but lenders are becoming more selective. AI-related borrowing must generate returns sufficient to justify its financing cost as investors reassess risks across public and private markets.

Analysis

The key signal is not a credit-demand collapse; it is a higher hurdle rate for projects whose returns depend on uncertain AI utilization. That creates a likely split between cash-rich cloud platforms able to fund infrastructure internally and developers or operators reliant on repeated external financing. If marginal projects are delayed, the pressure can travel upstream with a lag to data-center construction, power equipment, and AI hardware orders; it need not imply weaker demand for compute at the largest platforms.

The contrarian risk is treating selective underwriting as proof that the buildout is overfunded. Tighter credit could instead screen out low-return capacity while leaving contracted, well-capitalized projects intact. Over the next 1–3 months, watch financing spreads, loan terms, and project-level preleasing or contracted power. Over 6–18 months, the test is whether delays or cancellations reach supplier orders and hyperscaler capex plans. The article provides no deal terms, borrower exposures, or independently verified project returns, so it does not support a broad short on AI or credit.

Key Decisions for Investors

• Avoid a thematic short on AI infrastructure based on this signal alone. Treat it as a financing-quality distinction, not evidence that aggregate credit availability or AI demand has already rolled over.
• Prefer a relative-value watch: favor cash-generative cloud platforms over externally financed data-center development exposure if lending terms tighten or project cancellations rise. Do not size the pair until borrower-level leverage, refinancing schedules, and contracted utilization are verified.
• Monitor private-credit and infrastructure-debt spreads, covenants, and new-issue terms over the next 1–3 months. A sustained widening alongside weaker preleasing would strengthen the case for reducing exposure to leveraged developers and related construction suppliers.
• Falsify the cautious thesis if financing terms remain stable and projects continue to secure long-duration customer commitments, power capacity, and funding without material delays; that would indicate lender selectivity is not yet constraining the buildout.