200GW Hiding in Grid, Sodium Batteries 10x Cheaper, Wave-Powered Datacenters w/ Ramez Naam | EP #280

Energy scarcity, specifically regarding grid infrastructure, represents the primary bottleneck for the rapid scaling of AI compute. While power generation capacity is expanding, the inability to connect large-scale data centers to the grid creates a multi-year delay that threatens to stall progress. To address this, hyperscalers are increasingly adopting behind-the-meter power solutions, such as natural gas turbines and solar-plus-battery systems, alongside regulatory shifts that incentivize "interruptible loads." Long-term energy abundance relies on the successful commercialization of small modular reactors, fusion technology, and potentially off-world or ocean-based data centers. Although AI development currently follows a log-linear relationship with compute, achieving future intelligence milestones requires a massive expansion of energy production and a fundamental shift in how power is distributed and managed across global infrastructure.