When multiple companies are involved in one project, it can be unclear who is responsible for problems that arise. The Lake Mariner Data Center under construction in Somerset, New York, became the center of attention after a fire broke out in early June. The fire occurred in a building that was not yet finished, and it revealed how little the local fire department knew about the site. Firefighters found no working alarm system, no suppression system, and three dead hydrants at the scene. Safety documents that firefighters were legally allowed to see had apparently burned in the fire itself.
Steve Matisz, chief of the Barker Fire Department, said his team entered the building without much information, according to Ars Technica. The smoke from chemicals they couldn’t identify was thick and dangerous, as safety sheets were destroyed in the blaze. He said the situation had been difficult to handle, and he was unsure how to interpret claims that safety documents burned in the fire.
The site sits on land that was once a coal mine near Lake Ontario. This location is part of a large AI data center project in New York, one that includes many different companies. TeraWulf owns and operates the data center on land it leases from a company owned by its own CEO. Fluidstack, a UK-based AI company, will run the center. Google holds warrants for a future 14 percent equity stake in Fluidstack and has agreed to guarantee its lease payments. Anthropic is among the AI companies that will use the computing power provided by this facility.
After the fire, TeraWulf said it had taken steps to improve safety at the site. These included installing Knox boxes, adding more hydrants, and preparing safety data sheet “go-bags”.
In mid-August, the fire chief said a representative from Knox had visited and was planning to work on expanding the program. Despite the fire, construction at the site continued without any injuries reported.
The Lake Mariner incident shows how complex it is to assign responsibility in large AI projects. Legal, operational, financial, and reputational responsibilities are often spread across many companies. When a data center causes problems like high energy use or safety issues, it can be hard to say who is at fault.
Questions about how many jobs the project will create have also come up in Somerset. Somerset officials wanted more answers from TeraWulf after the fire, but the company did not respond to community requests. Town supervisor Jeffrey Dewart asked for a meeting, but TeraWulf’s communications team did not reply. At town board meetings, residents questioned what Somerset was getting in return for hosting one of the state’s largest AI data center projects.
According to a 2024 planning board presentation, TeraWulf’s project manager said the full buildout would bring between 35 and 40 jobs. That number is much lower than the 165 permanent jobs and $85 million in capital investment promised in 2019. The original application was made by Somerset Operating Company, which is owned by TeraWulf CEO Paul Prager.
Prager did not comment on the difference in job numbers. The New York Power Authority reviews how well the site meets its commitments each year. If the project doesn’t meet its goals, the benefits it received can be changed.
New York Governor Kathy Hochul recently placed a moratorium on new hyperscaler data center development. She has recognized that there is a gap between the promises made by these projects and the actual results.
Pilar Thomas, who once worked in government energy roles and now teaches tribal energy law, said it’s honest to say that many people are needed just to build these projects. She added that building a large data center can take hundreds of workers for months or even years.
When a data center uses a lot of electricity, the costs often affect more than just the company that caused them. Leah Stokes, a political scientist who studies energy and utility policy, said large loads increase the cost for everyone. She explained that when a data center runs 24 hours a day, 365 days a year, it puts pressure on the whole power system.
This situation shows how quickly large tech projects can create problems that go beyond one company’s control.