Why AI is struggling to power its own future

A Denver company called Crusoe, which specializes in building large-scale data centers for AI developers like OpenAI and Microsoft, recently canceled a 1.25 billion dollar deal to purchase custom power-generating turbines from Boom Supersonic. Boom, best known for trying to build a new passenger jet, had planned to repurpose the design of its airplane engines into stationary power plants for data centers. Crusoe was set to be their first customer, but both companies confirmed the partnership is no longer moving forward.

Modern data centers are essentially warehouse-sized collections of thousands of computer processors working in unison. These processors, which perform the complex calculations required to make AI work, consume electricity at a scale similar to a small city. To keep these systems running, data centers need a power source that is both massive and reliable. If the power fluctuates or cuts out, the entire system stops. While many facilities rely on the public power grid, companies are increasingly looking for ways to generate their own electricity on-site to ensure they have enough power without waiting for grid upgrades. This is why companies like Crusoe are looking at turbines—machines that burn natural gas to spin a generator—as a way to create their own dedicated electricity supply right next to their data centers.

The breakdown of this deal reveals just how much pressure is on the companies that build the physical infrastructure for AI. For these firms, electricity is now as critical as the hardware itself. The fact that an aviation company and a data center firm tried to strike such an ambitious deal shows that there is no obvious solution to the energy problem yet. Companies are still experimenting with different ways to power their machines, whether that means pulling from existing power lines, building custom gas plants, or relying on wind and solar power. As the demand for AI grows, the real bottleneck for the industry may not be the software itself, but rather the massive, physical challenge of finding enough electricity to keep the lights on.