Milk Road AI Inside the AI Cold War: Why Compute, Power, and Chips Decide Everything w/ Patrick & Duncan
11 snips
Nov 28, 2025 Dive into the AI Cold War, where Google surges and Meta eyes TPUs to challenge NVIDIA. The podcast explores the battle between specialized TPUs and flexible GPUs, decoding financial ecosystems behind major players. They discuss how AI boosts productivity and the challenges in data center power supply. With a focus on the US-China race, they analyze China's rapid growth and lag in frontier models. Lastly, there's a forecast that while AI adoption is gradual, its integration into workflows is inevitable.
AI Snips
Chapters
Transcript
Episode notes
Google's TPU Push Threatens GPU Dominance
- Google pairing Gemini 3 with its TPU stack challenges NVIDIA's GPU dominance by offering specialized hardware and software integration.
- Specialized chips win on efficiency for inference while GPUs retain flexibility and ecosystem strength.
Flexibility Versus Specialization Trade-Off
- GPUs trade flexibility for broad applicability while TPUs and other specialized chips bet on architectural specialization for efficiency gains.
- If models and workloads stay similar, custom silicon can deliver sustained performance and cost advantages.
Rebuild Workflows To Capture AI Gains
- Reconstruct workflows from scratch to unlock the full productivity gains from new AI models instead of layering them onto old processes.
- Startups should restructure teams so individuals hold more context and leverage agents to reduce coordination overhead.