
WSJ's Take On the Week The Risks Behind the Expected $5.3T AI Data Center Funding Boom
17 snips
Jan 25, 2026 Greg Peters, co-CIO for public fixed income at PGIM, a bond-market and credit specialist. He discusses the $5.3 trillion AI data-center financing surge and why much of it may hit bond markets. Short construction loans vs long unsecured hyperscaler debt come up. He highlights winner-take-all risks in hyperscaler buildouts and his top near-term worry: economic overheating.
AI Snips
Chapters
Transcript
Episode notes
Scale Of The AI Data-Center Finance Wave
- The AI data-center buildout requires roughly $5.3 trillion of financing, with about $4 trillion expected to hit the bond market.
- Greg Peters warns investors to be careful because this is an eye-popping supply wave that must be absorbed soon.
Prefer Shorter, Collateralized Loans
- Favor shorter-tenor, collateralized construction-style financing over long unsecured bonds when lending into data-center builds.
- Greg Peters prefers five-year in structures with strong collateral and backstops to mitigate refinancing and technology risks.
Mismatch Between Debt Tenor And Tech Risk
- Hyperscalers have been issuing long-term unsecured investment-grade debt despite high uncertainty about tech and useful-life of racks and chips.
- Peters questions issuing 40-year unsecured debt given rapid technology and refinancing unknowns.
