Imagine a future where your lights flicker, not because of a storm, but because an AI data center down the street is sucking up all the power. Sounds like science fiction, right? Wrong. We're already seeing the beginning of a massive energy crunch driven by the explosive growth of Artificial Intelligence, and it's hitting the US power grid harder and faster than anyone predicted.
According to a recent BloombergNEF forecast (December 1, 2025), data centers are projected to demand a staggering 106 gigawatts (GW) of electricity by 2035. That's a 36% increase from their forecast just seven months prior! To put that in perspective, one gigawatt can power roughly 750,000 homes. So, we're talking about needing the equivalent of powering nearly 80 million homes just to keep these AI brains humming.
This isn't just about more data centers being built; it's about the sheer size of these new facilities. Almost a quarter of the 150 new data center projects added to BloombergNEF’s tracker in the last year boast a capacity exceeding 500 megawatts. That's more than double the share from the previous year! These aren't your average server farms; they're energy-hungry behemoths. Think of it like this: instead of building a small coffee shop, everyone is suddenly building a massive coffee roasting plant.
But here's where it gets controversial... This rapid increase in demand is slamming head-on into the limitations of our existing power grids. In the PJM Interconnection (a regional transmission organization coordinating the movement of wholesale electricity in all or parts of 13 states and the District of Columbia), BloombergNEF anticipates data center capacity reaching 31GW by 2030. That's almost the same as the 28.7GW of new generation capacity that the Energy Information Administration expects to come online during the same period. In other words, data centers could gobble up almost all of the new power being generated! And this is the part most people miss: areas like the Electric Reliability Council of Texas (ERCOT), which manages the electric grid for most of the state, could see reserve margins (the buffer between supply and demand) dip into dangerous territory after 2028. While they can handle the immediate surge, long-term supply may lag behind. Some argue this is a wake-up call for massive investments in renewable energy and grid upgrades. Others worry about the potential for blackouts and rising electricity costs.
This situation presents a critical inflection point for the US. How do we accommodate the ever-growing energy demands of AI without compromising the reliability of our power grids or sending electricity prices through the roof? Is it even possible?
Adding another layer of complexity, the location of these data centers is also evolving. Northern Virginia, once the undisputed king of data center locations, is nearing saturation. New projects are now migrating south and west, towards central and southern Virginia, and Georgia, where expansion is extending beyond the Atlanta metropolitan area due to land and power constraints. Texas stands apart; Developers are repurposing former crypto-mining sites (which themselves consumed vast amounts of energy) into AI data centers, strategically located closer to population centers and vital fiber optic networks.
Fiber optics are the unsung heroes enabling this data center sprawl. They act as the digital backbone, allowing hyperscale campuses to operate efficiently even when located far from urban cores. By providing low-latency performance, these fiber networks empower developers to build these massive campuses in suburban and exurban zones, usually within 30 miles of major cities. Virginia pioneered this trend, leveraging its existing infrastructure and robust fiber backbone. Now, Georgia and Ohio are following suit, drawn by the promise of capturing the next wave of digital demand.
The big question is: who should bear the cost of upgrading the grid to support this AI-driven revolution? Should it be taxpayers, the data center companies themselves, or a combination of both? And what happens if we can't keep up with the demand? Will innovation be stifled? Will certain regions become more attractive than others based solely on power availability? What impact will this have on electricity prices for everyday consumers? Let us know your thoughts in the comments below! (BloombergNEF clients can access the full report here: https://www.bnef.com/insights/38111)
Related Content
Stay up-to-date with the latest insights from BloombergNEF. Sign up for their free newsletter.
