- PGIM anchored a $500 million collateralized loan obligation issued by Allstate that limits AI-related collateral to 15%, the first CLO to include such an explicit restriction.
- PGIM scores every loan in a pool for its connection to the AI industry, catching indirect exposures like utilities that traditional industry concentration limits tend to miss.
- With Bank of America projecting CLO exposure to chip and data center loans could reach $100 billion, AI concentration caps may become a standard feature of loan pools.
PGIM, the asset management arm of Prudential Financial, anchored a $500 million CLO that caps AI exposure at 15%, according to Bloomberg, which cited people familiar with the matter. Allstate issued the collateralized loan obligation in late September, and BNP Paribas arranged the deal.
It is the first CLO to carry an explicit limit on collateral tied to AI, a sign that credit investors are starting to treat the technology as a concentration risk in its own right.
Get Smarter about what matters in CRE
Stay ahead of trends in commercial real estate with CRE Daily – the free newsletter delivering everything you need to start your day in just 5-minutes
Why Industry Buckets Fall Short
CLOs have long used industry limits to keep loan pools diversified. PGIM argues those buckets miss how widely AI financing has spread across credit and equity markets.
Data center and chip debt carry the clearest risk if the AI boom stalls. However, borrowers with little direct technology exposure could also suffer losses. Utilities that run electrical grids, for example, could take a hit. That could happen if planned data center expansion fails to materialize.
The Details
Under PGIM’s framework, each of the hundreds of loans in a CLO receives an AI connection score. The score reflects how closely each loan relates to the AI industry.
Once the pool’s AI score exceeds 15%, the CLO cannot add more AI-related debt. Edwin Wilches, co-head of securitized products at PGIM Credit, described the framework.
Wilches said PGIM is working with other managers on deals with similar restrictions. None have reached the final stage yet. He estimates AI-related debt makes up roughly 2% to 3% of the $1.4 trillion CLO market. At that level, he said, it remains too small to raise concern.
Credit Markets Get Jumpy on AI
Debt investors have grown more cautious about AI exposure. Bloomberg reported that junk-bond risk premiums reached a five-month high in late September.
Those concerns grew after Oracle told a developer it could defer rent payments. Oracle plans to lease a massive data center in New Mexico. The developer could face delays on the project.
The episode shows how risk from the data center buildout can spread beyond developers. It can move through tenant leases and into the credit stack.
Why It Matters
For CRE lenders and data center developers, the cap sends a clear signal. Institutional credit buyers are starting to treat AI concentration as a distinct risk.
Bank of America strategists estimated in September 2026 that CLO exposure to chip and data center term loans could reach $100 billion. That compares with roughly $4.5 billion today.
Such growth would make AI a major driver of asset-backed issuance. It could also make diversification harder. Big Tech continues to spend record sums on AI development, Wilches said.
Equity investors searching for an AI immunity trade face a similar question.
What’s Next
Watch for PGIM’s framework to appear in other managers’ deals. Wilches said several managers are already working on similar structures.
If AI-linked loan volume approaches Bank of America’s $100 billion estimate, more CLO investors could demand explicit caps. They may rely less on traditional industry limits.
