The private credit market is confronting a systemic risk test under multiple, overlapping pressures. A wave of redemptions continues to spread, Wall Street is accelerating the launch of credit default swap (CDS) products to short private credit, and regulators are adopting a stricter stance. The Federal Reserve and the Treasury Department have already intervened, revealing the fragility of this asset class at an unprecedented pace.
The full-blown crisis in private credit stems from a deeper structural fracture: the underlying logic supporting its rapid expansion—corporate financing demand driven by the AI data center construction boom—is now facing a severe backlash from reality.
According to the latest data from Sightline Climate, of the approximately 16 gigawatts of data center capacity planned for operation in the US by 2026, an estimated 30% to 50% are expected to face delays or cancellations. Currently, only about 5 gigawatts are under active construction. This large-scale stagnation on the supply side directly impacts the pool of private credit assets that rely on AI infrastructure financing demand to sustain growth.
Simultaneously, Wall Street is positioning for the next crisis—the introduction of private credit CDS instruments signals that institutional investors have begun pricing in systemic stress for this market. The intervention by the Federal Reserve and the Treasury further confirms regulators' high alertness regarding potential risk spillovers.
Analysts point out that the underperformance in AI infrastructure project execution is pressuring the quality of underlying assets, liquidity mismatches are triggering redemption waves, the launch of shorting tools like CDS provides a new negative feedback mechanism for the market, and regulatory involvement further reinforces the market's pricing of systemic risk. The convergence of these multiple pressures is creating a "perfect storm" in the private credit market.
The AI Infrastructure Bubble: The Foundation of Private Credit Expansion is Shaking
The rapid growth of the private credit market in recent years has been deeply tied to the financing boom for AI infrastructure construction. However, this foundational narrative is now facing severe real-world challenges.
According to the "2026 Data Center Outlook" report released by Sightline Climate, the US plans to add at least 16 gigawatts of new data center capacity this year across 140 projects, but only 5 gigawatts are currently under construction.
The report clearly states that 30% to 50% of projects slated for 2026 are expected to face delays, primarily driven by power supply bottlenecks (25% of projects have not yet disclosed power supply solutions), increasingly effective community opposition, and risks of grid equipment shortages.
This gap is set to widen further in the coming years. Reportedly, of the 21.5 gigawatts of planned capacity announced for 2027, only about 6.3 gigawatts are actually under construction. For the period from 2028 to 2032, the vast majority of planned projects have not even broken ground—out of 37 gigawatts of planned infrastructure, only 4.5 gigawatts have commenced construction, with many projects lacking clear completion timelines.
Canaccord Genuity analyst George Gianarikas characterized this situation as "the US data center boom hitting a powerful wall of logistical friction." He warned that unless domestic manufacturing and grid integration see fundamental acceleration, "the digital expansion wave of the late 2020s faces a series of unfulfilled promises."
One of the core bottlenecks constraining data center construction is a severe shortage of key electrical equipment. Supply constraints for transformers, switchgear, and energy storage devices are not only due to surging demand from AI data centers but are also compounded by grid expansion needs driven by the adoption of electric vehicles and heat pumps.
According to Bloomberg reporting, domestic US manufacturing capacity cannot keep up with demand growth, forcing builders to rely heavily on imports. Philippe Piron, CEO of GE Vernova's Electrification business unit, pointed out that before 2020, delivery lead times for high-power transformers were typically 24 to 30 months, while AI companies often demand delivery within 18 months. The surge in demand has pushed lead times out to as long as five years, with some companies even having to refurbish old transformers from decommissioned power plants as emergency replacements.
This predicament reflects a deep-seated structural issue of hollowed-out US manufacturing. Over the past decade, a series of US government policies aimed at reshoring manufacturing have not yet yielded substantial capacity increases, and companies still heavily depend on Chinese supply chains—creating a direct contradiction with the US strategic goal of competing with China in the AI field.
The Funding Gap: $5 Trillion Needed, Government May Need to Backstop Over $1 Trillion
Pressures on the capital front are equally significant. The massive capital expenditures of hyperscale cloud computing companies still fall short relative to the overall financing requirement.
According to JPMorgan analysis, the total capital required to fully support this AI cycle is no less than $5 trillion. Even accounting for the substantial capital expenditures and debt financing of tech giants, the US government may need to fill a funding gap exceeding $1 trillion.
This financing pressure is precisely the core context for the private credit market's deep involvement in AI infrastructure. However, as risks of project delays and cancellations rise, the quality of private credit assets reliant on these projects' cash flows for debt repayment is facing increasing scrutiny.
Shreeti Kapa, an Executive Director at Goldman Sachs, cited a consensus from an investor dinner in a recent report, stating, "There simply isn't enough compute power. Every participant faces severe compute constraints—from chip fabrication plants to data center siting permits, to power, memory, and labor. The bottlenecks are real and will persist for a considerable time."
Comments