Get your free personalized podcast brief

We scan new podcasts and send you the top 5 insights daily.

While securing a gigawatt of power for a large AI campus is difficult, smaller pockets of 10-100 megawatts are often available on the grid. Crusoe's modular 'Spark' data centers are designed to tap into this unused capacity, allowing for much faster deployment of inference infrastructure by bypassing the immense challenge of large-scale grid interconnection.

Related Insights

Crusoe Cloud is partnering with Tesla co-founder JB Straubel's Redwood Materials to use second-life EV batteries for power. By pairing these recycled batteries with solar, they can run a fully off-grid AI data center 24/7 at a lower price than grid power in Northern Virginia, a major data center hub.

Contrary to the popular "off-grid" narrative, hyperscale AI data centers will likely adopt a hybrid power architecture. This involves being grid-tied while using captive generation, storage, and demand response as a bridge solution to overcome utility interconnection delays and ensure stability.

To overcome energy bottlenecks, political opposition, and grid reliability issues, AI data center developers are building their own dedicated, 'behind-the-meter' power plants. This strategy, typically using natural gas, ensures a stable power supply for their massive operations without relying on the public grid.

AI companies are building their own power plants due to slow utility responses. They overbuild for reliability, and this excess capacity will eventually be sold back to the grid, transforming them into desirable sources of cheap, local energy for communities within five years.

Contrary to the common focus on chip manufacturing, the immediate bottleneck for building new AI data centers is energy. Factors like power availability, grid interconnects, and high-voltage equipment are the true constraints, forcing companies to explore solutions like on-site power generation.

The massive power demands of AI will force hyperscalers to abandon their reliance on the public grid. They will build dedicated, co-located power plants, likely small modular nuclear reactors. This "Bring Your Own Energy" approach ensures speed to power and creates opportunities to sell excess energy back to communities.

According to Poolside's CEO, the primary constraint in scaling AI is not chips or energy, but the 18-24 month lead time for building powered data centers. Poolside's strategy is to vertically integrate by manufacturing modular electrical, cooling, and compute 'skids' off-site, which can be trucked in and deployed incrementally.

Vast amounts of solar power are curtailed or unused due to grid limitations and overbuilding. Rune's modular data centers treat this "excess" power not as waste, but as a stranded asset, converting it directly into AI compute at the source. This reframes renewable energy sites as future compute hubs.

Crusoe's CEO explains their core strategy isn't just finding stranded energy, but actively developing new power sources alongside their AI factories. By building out power capacity to meet peak demand, they create an abundance of energy that can also benefit the surrounding grid, turning a potential liability into an asset.

The "across the meter" concept involves co-locating power generation with a data center and a grid interconnection. This allows the data center to consume the power it needs, draw from the grid to cover shortfalls, and, crucially, supply its excess generated power back to the grid. This transforms a major power consumer into a source of energy abundance for the local community.