The hills surrounding Sugarloaf in northern Pennsylvania still resemble the rural landscape that most Americans picture when they hear the term “farmland.” Apple trees, deep-red barns, and peaceful roads with pickup trucks passing by perhaps once every few minutes. John Zola believed his forty acres would remain that way for many years. A basketball court for the grandchildren, a tiny orchard, and enough room to construct houses for his three grown children.
They proposed to run a massive 500-kilovolt power line across the property, along with maps and metal stakes. The century-old apple trees would be far below the 240-foot-tall steel towers. Zola referred to the project as “hell.” However, the line meant much more to the utility companies and tech companies that made the request: electricity for artificial intelligence.
| Topic | Key Information |
|---|---|
| Subject | Global expansion of AI data centers and land acquisition |
| Major Players | Amazon, Microsoft, Google, Meta, OpenAI |
| Infrastructure | Hyperscale AI data centers and supercomputing clusters |
| Land Use | Former farmland, industrial land, and rural areas |
| Power Sources | Natural gas plants, nuclear projects, renewable power deals |
| Energy Demand | AI servers use up to 10x the energy of traditional search systems |
| Investment Scale | Tech firms projected to spend hundreds of billions on AI infrastructure |
| Water Use | Data centers consume large volumes of water for cooling systems |
| Community Impact | Rising electricity demand, land disputes, and new power lines |
| Reference | https://www.cnet.com/tech/services-and-software/ai-data-centers-are-coming-for-your-land-water-and-power |
Something that increasingly resembles a resource rush has been sparked by the explosion of artificial intelligence. This time, it was for land, electricity, and water—the raw materials needed to operate massive data centers—rather than for gold or oil. Once primarily competing in software, tech companies are now negotiating with power plants, purchasing farmland, and lobbying governments for energy infrastructure.
A typical office building with servers hidden in a basement is not what a modern AI data center looks like. The scale is immediately apparent when you enter one. Brightly lit hallways are lined with rows of black server racks. Chilled air is forced over thousands of processors by fans that roar nonstop. The machines are fed by red, blue, and yellow cables that hang in thick bundles like exposed roots.
Tens of kilowatts of power can be consumed by each rack, which is significantly more than the equipment found in conventional computing facilities. The demand for electricity increases dramatically when you multiply that by thousands of racks operating around the clock. A midsize city’s worth of electricity will be consumed by some of the biggest AI data centers currently under construction.
Recently, Amazon started constructing a massive AI facility on about 1,200 acres of Indiana farmland. When the project is completed, it may use more than two gigawatts of electricity, which is about the same amount as a million households. Similar initiatives are being pursued by Microsoft, Google, and Meta, who frequently locate their data centers close to sizable power plants or hydroelectric dams.
Investors appear certain that the next ten years of technology will be shaped by this infrastructure race.
The amount spent is astounding. According to analysts, in the upcoming years, large tech companies may invest hundreds of billions of dollars in developing AI infrastructure. According to some projections, the world will spend trillions of dollars on AI data centers by 2030. However, the land is only one part of the puzzle.
Data centers already spend a lot of money on electricity, and AI workloads use a lot more power than previous computing tasks. For weeks, thousands of specialized chips may need to run concurrently in order to train large language models. Because those chips are hot, they require large cooling systems, which use more energy.
Retired natural gas plants are being brought back to life. Some businesses are entering into contracts with nuclear power providers that span decades. Others are constructing their own gas-fired generators right next to data centers, thereby establishing AI-focused private power plants.
Large server campuses are being built on former rural and desert land in Texas and Nevada. Meta is building a facility in Louisiana that will encompass an area comparable to portions of Manhattan. The buildings themselves, which are low concrete structures encircled by security fences, have an almost anonymous exterior, but within are some of the most potent computing clusters ever constructed.
Jensen Huang, the CEO of Nvidia, has started referring to these establishments as “AI factories,” a term that encapsulates the industrial scope of what is taking place. Instead of steel beams or automobiles, the products are trained neural networks that can produce language, images, and code.
However, factories need resources. And that’s where the narrative gets trickier.
Local communities have started to rebel. Concerns among the locals include growing electricity bills, water usage, and the loss of farmland. Environmental organizations caution that the increase in electricity consumption may impede the shift to greener energy.
Another concern is whether the industry is expanding too rapidly. In private, some analysts speculate that the current enthusiasm for AI might wane, leaving behind costly infrastructure and significant energy commitments. The outcome of that gamble is still unknown.
However, one thing is starting to become clear. The center of gravity of the technology sector is changing. Companies that used to thrive on cloud software and code are now thinking in terms of megawatts, transmission lines, and acres.