MIT/CRE Meet the Visionaries

Building for AI: Capital, Power, and Risk in Next-Generation Data Center Development

18 snips
May 6, 2026
Hossein Fateh, founder and CEO of Cloud Capital and CloudHQ, has spent 25+ years building data centers worldwide. He explores the shift from cloud to AI, the real estate differences between training and inference facilities, and the financing risks of massive AI projects. The conversation also covers power constraints, liquid cooling, renewables, global expansion, water use, and why data centers matter as critical infrastructure.
Ask episode
AI Snips
Chapters
Transcript
Episode notes
INSIGHT

Training And Inference Need Different Real Estate

  • AI creates distinct real estate demands because training can locate broadly while inference often needs proximity to users.
  • Inference may ultimately become five to ten times larger than training, making location increasingly decisive.
ADVICE

Match Data Center Location To Latency

  • Match facility locations to the application’s latency sensitivity rather than treating every AI workload alike.
  • Distinguish edge content delivery from inference concentrated in major cloud hubs such as Ashburn, Chicago, Dallas, and San Jose.
ADVICE

Underwrite AI Tenants For Residual Risk

  • Underwrite non-investment-grade AI tenants conservatively because enormous construction costs create difficult financing exposure.
  • Seek residual guarantees or reduce debt over the lease term to protect against renewal and obsolescence risk.
Get the Snipd Podcast app to discover more snips from this episode
Get the app