JPMorgan flags $29 billion in data-center lending as banks eye “AI debt” more cautiously

JPMorgan Chase is holding a sizable portfolio of commercial real estate loans linked to data centers, underscoring how major banks remain exposed to the infrastructure behind artificial intelligence and the power-hungry computing buildout it drives. A new market report points to JPMorgan’s position in that lending niche and frames it as a pressure point as some lenders grow reluctant to add what the article characterizes as “AI debt.”

The report says JPMorgan holds $29 billion of loans in a commercial property category that includes data centers. It also states that this category has grown to represent 12% of JPMorgan’s commercial real estate loan exposure, illustrating that while data-center lending is still a segment rather than a dominant share, it is large enough to matter materially inside the bank’s broader commercial book.

The caution referenced in the article comes as lenders assess the risk of stretching leverage to finance AI-related projects. In the data-center business, the economics can depend on long-term power availability, network connectivity, construction timelines, and the pace at which demand converts into contracted capacity. When financing terms tighten or underwriting standards become more conservative, banks may slow new commitments or reprice risk.

JPMorgan is not described in the post as abandoning the sector. Instead, the emphasis is on scale and exposure as lending decisions come under greater scrutiny. The article’s framing suggests that, even for a large, diversified lender, the combination of higher costs and rapid capacity buildouts can change how banks evaluate new deals.

For JPMorgan, this is also a question of portfolio management across commercial real estate, where banks must balance growth against concentrations, property-level risk, and the potential for refinancing stress if market conditions shift. Data centers are often treated as a comparatively resilient asset class when demand is strong, but they can still face volatility tied to construction completion risk and tenant commitment timing.

The report does not provide additional details on underwriting standards, loan maturities, delinquency rates, or specific segments of data-center exposure such as single-tenant versus multi-tenant facilities. It also does not specify whether the $29 billion figure reflects gross loans outstanding, commitments, or another measure of exposure, beyond saying it relates to the data-center-inclusive commercial property category.

JPMorgan did not disclose further granularity in the cited market post, such as how much of the portfolio is tied directly to AI-related tenants, how much is secured by longer-dated leases, or what assumptions it uses for power and demand. Those are key points that typically determine how banks model losses and recovery prospects in commercial real estate downturns.

What to watch next is whether JPMorgan’s future commentary or filings show changes in the pace of data-center lending, tighter constraints on leverage, or revised risk metrics. For the broader market, investors will likely focus on whether other lenders follow JPMorgan’s lead in keeping exposure steady or move toward smaller targets as they reassess funding for AI-linked infrastructure.