Why Kain's $20,000 Mac Studio Never Paid for Itself: Uneasy Money

Summary

Erik Voorhees discusses AI geopolitics and why he prioritizes principles like freedom and privacy over US nationalism, arguing that America does not automatically deserve to win AI. He points to Bitcoin as the model for permissionless opt-out and says decentralized AI inference is feasible while decentralized training is not yet competitive. The conversation then turns to the economics of local AI: Kain's $20,000 Mac Studio never paid back, and Erik argues large-scale data-center inference is structurally cheaper.

• Erik says he is not a nationalist and cares more about principles than about the US winning AI.
• He argues the US is becoming more authoritarian while China has become more market-oriented.
• Bitcoin is framed as proof that people can opt out of a broken system without permission.
• Decentralized inference is possible, but decentralized training is not yet competitive.
• Kain says his $20,000 Mac Studio had a poor and worsening payback curve for local models.
• Erik argues local AI inference is not about cost savings because data-center inference is structurally more efficient.
• The key stated investment implication is that cloud/data-center inference has the durable economic advantage over local AI hardware.