Gavin Baker argues the AI investment cycle is fundamentally different from the 2000s internet bubble, pointing to full utilization of GPUs versus the 97% of unused fiber from that era. He notes that while trillions are being spent on data centers, usage is surging and valuations are more reasonable, with tech giants' massive cash flows providing a buffer. Baker identifies Google as Nvidia's primary competitor due to its TPU chips and sees infrastructure as a safer bet than predicting application-layer winners. He discusses how AI will impact SaaS, suggesting that legacy companies can compete with AI-native startups by running new products at breakeven. The conversation also covers the economics of frontier models, the competitive landscape for custom chips, and the shift toward outcome-based pricing models. Baker concludes by comparing the AI cycle to past tech booms and shares his view on robotics, highlighting Tesla's lead in humanoid robots.