The Daily Brief The conjurings of data center finance
30 snips
Mar 2, 2026 An exploration of how data centers are financed, from construction loans to GPU-driven split financing and US securitization trends. A look at distinct business models like colocation versus compute-as-a-service and why that matters for lenders. A comparison of India’s bank-led funding model and its systemic risks. A rundown of SEBI’s new mutual fund rules, lifecycle funds, overlap caps, and tightened investor exit loads.
AI Snips
Chapters
Transcript
Episode notes
Compute Operators Use Revenue Locks Not Long Bonds
- Compute-as-a-service operators struggle to issue long-term bonds because GPUs depreciate in 2–3 years, so they rely on equity, private credit, advance payments, and long take-or contracts.
- Some have issued non-investment grade bonds and use advance payments to create revenue predictability.
India Uses LRD Loans Not Securitization
- India fuses construction and long-term financing: operators typically use bank LRD loans that pledge future rental income instead of tapping securitized capital markets.
- This concentrates risk in banks, forces reliance on parent equity infusions, and limits recycling of capital seen in the US.
SEBI Overhauls Mutual Fund Categorization
- SEBI updated mutual fund rules, replacing the 2017 framework to address thematic funds, REITs/InvITs, and the need for goal oriented products.
- The 2017 single-scheme-per-category rule brought order but had become outdated as new asset classes and thematic inflows exploded.