Possible The grid(lock) slowing AI down
46 snips
Apr 15, 2026 They debate AI moving from apps into everyday devices and why hardware presence alone won’t make a winner. They dig into the $650B data center buildout, its transformer and electrical bottlenecks, and geopolitical supply risks. They explore how memory, personalization, and repeated use create lasting stickiness. They spotlight a challenge that uses AI to rebuild trust in institutions.
AI Snips
Chapters
Transcript
Episode notes
Grand Canyon Photo and Gemini on 800 Million Devices
- Aria Finger notes Samsung aims to run Google's Gemini on 800 million devices by end of year.
- She gives consumer examples like AI removing people from a Grand Canyon photo and real-time phone features boosting awareness from 30% to 80%.
Hardware Presence Alone Won't Decide AI Winners
- Hardware presence alone doesn't determine AI value capture.
- Reid Hoffman explains hours of user interaction, iteration, and depth (phones, co-pilots) drive model improvement more than mere device distribution.
Different Network Effects Explain Stickiness
- Strong versus weak network effects matter for device-installed models.
- Hoffman contrasts weak effects (multiple messaging apps coexisting) with stickiness from personalization and memory in services like CarPlay.