An elderly woman in rural Ohio has refused to sell her land for a proposed artificial intelligence data center, even though the offer was worth millions of dollars. The family owns about 1,200 acres and has lived on the land for generations, raising children and working the soil through decades of farming. A major artificial intelligence company approached several landowners in Mason County, including the Huddleston family, to buy part of their property for a data center campus.
The company offered $60,000 per acre for the woman’s land and $48,000 an acre for her daughter’s portion, totaling over $26 million, according to Fortune. Although the family was initially tempted by the offer, they ultimately decided against selling. The daughter said fear played a role, especially the possibility of eminent domain and pressure from unnamed developers.
She and her mother pulled out of the deal, saying they didn’t believe the promised jobs would come to pass. The family also expressed concern about water shortages and groundwater contamination near similar data centers in other states. The woman said that money could not replace the deep connection she had to the land.
Despite their refusal, the project has moved forward with land from other willing sellers. A zoning request has been approved to rezone more than 2,000 acres of farmland for the data center. The facility would be built next to the family’s farm, not on it.
Local officials claim the project will bring hundreds of full-time jobs and over a thousand construction positions. But the family remains unconvinced, saying they don’t want or need the money. The Huddlestons’ resistance is part of a broader conflict across rural America.
Farmers and ranchers are organizing opposition to data center projects, with some communities holding town halls to express their concerns. In Nevada, a similar conflict has emerged over power usage for AI data centers, according to a report from CBS News. A company called Tract plans to build two campuses that together would use more than two gigawatts of power.
This amount is nearly a third of the total generating capacity in the state, according to utility officials. The utility company has argued that such large power users must pay for their own infrastructure and not shift costs to local residents. The issue stems from agreements between the utility and Tract’s local partner, Reno Power, which set rules for electricity delivery.
Tract has asked for private arbitration to resolve the issue, which the utility says is meant to avoid public oversight. Nevada’s regulations require large customers to go through public regulators to determine their energy needs. The utility company has sued to prevent the arbitration, saying it had no choice but to act.
Nevada’s power companies are legally required to serve anyone who asks for electricity in their area. The utility has already invested over $127 million in infrastructure and committed nearly $1 billion more to support all customers. The rise of AI has made data centers and energy providers increasingly dependent on each other.
Meanwhile, in South Korea, companies are working to reduce their reliance on Nvidia chips for AI infrastructure, UPI reported. Nvidia still dominates the global market for AI training, with most large language models built using its chips and software. But the focus is shifting from training models to running them, which brings new challenges in cost and energy use.
SK Telecom and another company called Rebellions are developing an infrastructure system using homegrown chips. They have already deployed Rebellions’ chips in systems supporting SK Telecom’s A. X K1 AI model. The company sees this as part of a broader plan to build an AI factory that includes data centers, chips, networks, and software.
SK Telecom is combining different types of processors to make its AI operations more efficient. The company’s AI business is beginning to generate revenue, with profits rising significantly in recent quarters. SK Telecom plans to build 5 gigawatts of AI data center capacity starting in 2029, with a goal of reaching 15 gigawatts over time.
Industry experts expect the competition in AI infrastructure to grow beyond just Nvidia and South Korean chips. Data center operators will likely choose different processors for different tasks, using GPUs for training and other chips for running AI services. This shift means that companies will continue to depend on Nvidia for large model training, but also use local chips for inference.
IMAGE: Mindseyes / Pixabay