"World of DaaS" OpenAI CSO Jason Kwon - power, policy and the race to scale
30 snips
Sep 23, 2025 Jason Kwon, Chief Strategy Officer at OpenAI, dives into the pressing issues of compute bottlenecks and power infrastructure challenges in the AI sector. He discusses user expectations versus actual AI capabilities, exploring the complexities of training versus inference energy demands. The conversation touches on the geopolitical stakes of AI, particularly in relation to talent acquisition. Kwon also highlights the importance of policy in driving AI adoption and the need for organizations to adapt for effective integration.
AI Snips
Chapters
Transcript
Episode notes
Compute Is The Strategic Bottleneck
- At the industry and country level, compute (capital + power) is the primary bottleneck for AI progress.
- More compute enables broader experiments, faster research, and better test-time reasoning performance.
Bet Selection Beats Raw Capital
- At an organization level, bet selection and how you use compute matter more than raw capital.
- Organizational structure, taste, and sustained research choices determine who converts compute into wins.
Power Grids Weren't Built For AI
- Power planning worldwide didn't account for rapid AI growth, creating future capacity gaps.
- Adding tens of gigawatts is significant and site approvals plus construction take many months.