Endowments Are Taking AI Equity and Shunning AI Debt

What's New

Allocators want AI upside through equity and private infrastructure, and they are pulling their diversifying capital out of loans that fund the buildout. Jason Safran, Chief Investment Officer at Texas Christian University, argues this in a panel at Bloomberg Invest in Dallas. TCU has owned data center equity since 2020 and is adding power and transmission. It has moved its diversifying sleeve away from direct lending and triple net leases on data centers. LPs should map how much AI credit risk hides in the parts of the portfolio meant to protect them.

Why It Matters

The view runs against the pitch that infrastructure credit is the safe way into AI. Data center lenders and private credit funds raising AI capital sit across the table, and CoreWeave bonds trading at mid-teens yields show the repricing has started. Fort Worth's Derrick Dagnan adds that AI debt issuance is crowding out fixed income globally. Partners Capital's Olivia Newell notes AI now reaches credit, real assets and venture. TCU's existing data center and venture stakes benefit if AI equity keeps its upside.

What To Watch

• Data center lending spreads in private markets. Safran expects capital markets to pace the buildout by raising the cost of capital for individual players, and private loans should follow public bonds.

• Follow-on fund sizes from GPs with aging portfolios. Dagnan expects many to come back with smaller funds to prove their thesis again, a reset he says the 2019 to 2021 vintages need.

• Sales below marks and continuation vehicles. Safran sees impatient LPs pushing GPs to sell under carrying value, which he says has left private equity better marked than 2 years ago.

Memorable Quotes

• "We know what we own. We really deeply understand what we own." Newell treats exposure mapping as the first defense when AI touches every asset class.

• "So take your medicine and the medicine is transacting." Dagnan wants GPs to sell aging assets and accept bad vintages.

• "I'm not saying those are bad investments." Safran concedes that AI-linked direct lending may perform well, which leaves his case resting on concentration risk alone.

• "I don't want to take the other side of that bet in our diversifying strategies." Equity captures AI upside, and the safety sleeve has to stay outside the trade.

The Wrap

Safran's split works while AI borrowing needs keep rising faster than lenders reprice. It also depends on his alternatives, such as mortgage servicing REITs, reinsurance and senior housing, staying uncorrelated in a higher-for-longer rate world. It breaks if AI equity marks fall first, since equity absorbs losses before secured lenders. Safran already worries that expectations for AI-linked companies run very high. The marks on AI-linked private companies over the next several quarters, set by the frontier bellwethers, will test it.