Milk Road AI The Massive AI Build-Out No One Is Pricing In Yet w/ Salim Ismail
23 snips
Nov 19, 2025 
Salim Ismail, founder of OpenExO and former CEO of Singularity University, dives into the AI landscape's future and physical infrastructure needs. He debunks the notion of AI as a bubble, discussing how model efficiency won't solve the compute demands. Salim highlights the transformative role of on-premises AI for sensitive industries and the need for new workflows in organizations. He envisions a future of AI-driven education and healthcare, while cautioning about outdated governance structures potentially hindering progress.
AI Snips
Chapters
Books
Transcript
Episode notes
AI Is Not A Broad Market Bubble
- Salim Ismail argues AI is not a bubble because only a few companies dominate hype while the real market requires massive, sustained infrastructure buildout.
- He warns we must 'tile the earth with data centers' and sustain investment for years to meet demand.
Efficiency Won't Stop Compute Demand
- Model efficiency improves rapidly but demand for compute grows far faster due to many new applications and deployments.
- Even 40x efficiency gains are dwarfed by an estimated thousandfold increase in models and workloads to train and run.
Build AI-Native Units, Not Just Wrappers
- Reengineer business processes to be AI-native instead of merely automating human-centric workflows.
- Create parallel AI-native units in legacy companies and slowly deprecate old processes to scale transformation safely.
