Pacific Investment Management Co. Shifts Its AI Debt Strategy

Key Takeaways

• Pacific Investment Management Co. is becoming more selective after providing billions of dollars for AI infrastructure.
• Hyperscaler and data-center borrowing is expanding as projected capital spending approaches $690 billion in 2026.
• Secured financing may offer stronger protections, but technology, refinancing, and customer-concentration risks remain.

Pacific Investment Management Co., widely known as Pimco, helped fuel the artificial intelligence infrastructure boom with billions of dollars in lending. Now the investment manager is repositioning as a flood of new borrowing tests how much exposure credit markets can comfortably absorb.

The shift does not amount to a wholesale retreat from AI. Instead, Pimco is emphasizing selectivity, stronger collateral, and financing structures that can provide claims on physical assets, leases, or contracted cash flows. In a market increasingly divided among investment-grade hyperscalers, data-center landlords, and higher-yield neocloud providers, those distinctions matter.

The sheer scale helps explain the caution. In its September credit-market assessment, Pimco said consensus capital-expenditure projections for the five largest hyperscalers were approaching $690 billion in 2026 and $870 billion in 2027. Hyperscaler debt issuance had reached about $136 billion year to date in 2026, with another $58 billion tied to data centers.

The volume of debt required to fund this growth is substantial. More than $5 trillion may be required through 2030 for AI data centers, chips, power, cooling, and network infrastructure, according to estimates cited by Pimco. More than 40% of that total could require debt-market financing (source). What happens when lenders stop treating every AI-linked project as essentially the same trade?

AI financing is no longer confined to conventional corporate bonds from companies with enormous balance sheets. The market now includes asset-backed securities, commercial mortgage-backed securities, private credit, and secured infrastructure loans. Rule 144A private placements also give issuers access to institutional capital without using the fully registered public-bond market.

This widening financing chain mirrors a rapid buildout. Dell'Oro Group reported that global data-center capital expenditure increased 59% year over year in Q3 2025, marking an eighth consecutive quarter of double-digit growth. AI systems and accelerated servers were central drivers. Separately, Synergy Research Group estimated hyperscaler infrastructure spending at approximately $142 billion in Q3 2025, nearly triple the year-earlier level, while global hyperscaler capacity additions rose 170% per quarter.

For lenders, the borrower's position in that chain can shape the risk. Microsoft, Amazon Web Services, and Google Cloud can support investment with large businesses and established cash flows. Data-center landlords depend more heavily on lease terms, tenant quality, and refinancing conditions. Neocloud providers such as CoreWeave may offer exposure to fast-growing demand, but often with higher credit risk and greater dependence on a relatively concentrated set of customers and suppliers. NVIDIA-backed infrastructure adds another layer to the ecosystem.

Collateral can help, though it does not remove the underlying uncertainty. An asset-backed transaction may give lenders claims on servers, buildings, leases, or contracted payments. Yet specialized computing equipment can become obsolete quickly, while data-center properties may require expensive upgrades to accommodate denser racks, new cooling systems, or greater power demand. Physical security is useful. Economic durability is harder to assess.

Then there is timing. AI infrastructure is being financed today against expectations for demand that may unfold over many years. Loan maturities, customer contracts, and useful equipment life do not necessarily line up neatly. If a major tenant declines to renew, utilization weakens, or refinancing becomes more expensive, apparently well-protected debt can behave differently than investors expected.

For enterprise technology leaders, Pimco's repositioning is a signal that infrastructure funding may become more discriminating rather than simply disappearing. Projects with credible tenants, durable contracts, available power, and flexible facilities could retain access to capital on better terms. More speculative developments may face higher borrowing costs or stricter covenants. The AI buildout is continuing, but debt investors are starting to ask tougher questions about who owns the risk when the bill comes due.