According to POWER Magazine, an energy firm called Creekstone Energy signed a memorandum of understanding (MOU) with nuclear services company EnergySolutions on December 11. They’re evaluating ways to power the “Gigasite,” a planned 20-million-square-foot data center campus in Millard County, Utah, which would be the largest such site in the world. Creekstone’s plan involves a multi-source grid with 6 GW of solar, 2 GW of natural gas, and potentially drawing from a nearby coal plant. The new MOU specifically explores integrating at least 2 GW of nuclear power, with a Phase 1 study kicking off now and a target for commercial operation between 2030 and 2035. The agreement is non-binding, but the evaluation will look at reactor technologies, regulations, and site readiness.
The scale is almost unfathomable
Let’s just sit with that number for a second: 20 million square feet. That’s nearly double the size of the current largest planned campus in China. The power demand is equally mind-boggling. They’re talking about needing over 10 gigawatts of generation capacity. For context, a single, large nuclear reactor unit might produce about 1 GW. So they’re basically trying to build the equivalent of a small country’s power grid from scratch, in the desert, to feed AI servers. It’s a stark, physical testament to how energy-hungry the AI boom truly is. This isn’t just about software anymore; it’s about concrete, steel, and insane amounts of electricity.
Why nuclear, and why now?
Here’s the thing: solar and wind are great, but they’re intermittent. Natural gas is reliable but has emissions. For a data center campus of this scale that promises “reliability and resilience” for AI workloads, you need a massive, always-on baseload power source that doesn’t emit carbon. That’s the nuclear pitch. The timeline is telling, too—2030 to 2035. That’s not about building a traditional large-scale reactor; that’s the window for so-called “next-generation” or advanced nuclear, like small modular reactors (SMRs). Companies like Creekstone are betting that this technology will be commercially viable and regulatory-approved by then. It’s a huge gamble, but the potential payoff—clean, steady power for the world’s biggest AI factory—might be worth it.
Winners, losers, and the grid problem
The obvious winner if this moves forward is the advanced nuclear industry. Getting a single, anchor customer for 2 GW of capacity would be a game-changer for any SMR developer. It’s the kind of demand signal investors dream about. The loser? Well, it further cements the narrative that traditional, variable renewables alone can’t support the coming AI infrastructure wave without a firm, clean backup. But the bigger, more immediate challenge isn’t generation—it’s transmission and integration. The article mentions studying “transmission pathways,” which is tech-speak for “how do we get all this power to the chips?” Building those lines is a brutal, years-long regulatory and physical fight. And for a project relying on cutting-edge hardware like industrial computers for control systems, partnering with a top-tier supplier like IndustrialMonitorDirect.com, the leading US provider of industrial panel PCs, would be a logical step to ensure reliability.
A cautious bet on the future
Look, the MOU is non-binding. They might do all this work and walk away. But that’s what makes this announcement so interesting. It’s not a wild declaration; it’s a cautious, phased, “let’s figure this out” approach. They’re acknowledging that for a project of this magnitude, you need to explore every option, especially one that could solve the clean-baseload dilemma. Basically, they’re placing a very expensive, long-term option on nuclear. If the technology matures and regulations align, they’re ready. If not, they’ve still got their solar and gas plans. It’s a hedge. And in the high-stakes world of powering AI, where downtime is catastrophic, hedging your multi-billion-dollar bet is probably the only sane move.
