Episode #61: Powering the Machine: Altman, Milei, and the Energy Behind AI’s Future

00:00 – Stewart Alsop III opens with news of Sam Altman’s $25B OpenAI data center plan in Patagonia, tied to a U.S.–Argentina $20B currency swap and Trump’s backing of Milei.
05:00 – They unpack Argentina’s political turmoil, corruption scandals, and the U.S. effort to counter China’s influence over lithium and rare earths.
10:00 – Discussion turns to AI infrastructure logistics, how data centers need massive power, and Altman’s ties to U.S. energy interests, including solar, nuclear, and SMR reactors.
15:00 – Stewart II compares this boom to the 1990s dark fiber overbuild, warning of overcapacity and shifting ownership in infrastructure cycles.
20:00 – They analyze OpenAI’s 800M users, inference costs, and Sora’s energy demand, considering how infrastructure strain shapes AI access.
25:00 – The talk shifts to Unitree robots, physical AI, and Stewart II’s investment in Chef Robotics, linking automation to industrial change.
30:00 – Closing with reflections on distributed systems, uptime, Google’s architecture, and the evolution from AltaVista to TikTok as symbols of scalable intelligence.

Key Insights

1. AI expansion is reshaping geopolitics. Stewart Alsop III and Stewart Alsop II frame Sam Altman’s $25 billion OpenAI data center project in Patagonia as more than a business move—it’s a geopolitical play. By pairing it with a U.S.–Argentina $20 billion currency swap, the initiative strengthens U.S. influence in South America while countering China’s earlier economic foothold through currency deals and lithium investments.
2. Energy is the new frontier for AI infrastructure. The conversation underscores that AI growth isn’t limited by hardware alone but by power. Data centers require enormous, stable energy supplies, and Altman’s reported interest in solar, battery storage, and small modular nuclear reactors (SMRs) reflects how energy independence has become central to national AI strategies.
3. Overbuilding echoes the dot-com era. Stewart II draws parallels to the 1990s dark fiber boom, when telecom firms massively overbuilt capacity that sat unused for years. The hosts suggest today’s $400-billion-plus data center race—by OpenAI, Microsoft, Oracle, and others—may follow a similar arc, where hype precedes utility and ownership eventually shifts to new players.
4. Chip scarcity defines the AI arms race. They emphasize how Nvidia’s limited GPU supply and OpenAI’s deal with AMD to secure more chips illustrate the bottlenecks in AI scalability. Control over advanced semiconductors now carries the same strategic weight as oil once did.
5. Inference cost and access inequality. With OpenAI serving roughly 800 million users but only 20 million paid, the discussion highlights how computational costs shape user experience. Free users get constrained performance because inference—running models at scale—consumes vast, expensive compute power.
6. Physical AI remains in its infancy. Stewart II’s firsthand experience with the Unitree robot shows how humanoid robotics are still more experimental than autonomous. Yet, his investment in Chef Robotics signals that real commercial progress is happening in less glamorous, industrial automation.
7. Distributed systems are the hidden backbone of AI. The pair close by tracing the lineage from Google’s early distributed architecture to today’s global platforms like TikTok and Instagram. These systems represent decades of evolution toward high-availability computing—proof that scaling intelligence depends as much on resilient infrastructure as on smarter models.