TeraWulf의 Lake Mariner 데이터센터 화재로 AI 컴퓨팅 플릿 전반에 안전 검토 및 운영 중단이 발생했다.

When firefighters from the Barker Fire Department arrived at TeraWulf’s Lake Mariner data center campus in Somerset, New York, in early June, they encountered a scenario no first responder wants to face. There was no functioning fire alarm, no suppression system, and three hydrants that did not work. Additionally, according to TeraWulf, the safety data sheets that would have identified the chemicals producing the heavy black smoke filling the unfinished building were destroyed in the fire itself.

Fire Chief Steve Matisz stated that his crew entered the building “kind of blind.” He expressed uncertainty regarding the company’s claim that the safety documentation was consumed by the flames. “It’s been a difficult situation,” Matisz said.

The Lake Mariner campus occupies the site of a retired coal plant in upstate New York. TeraWulf began operating the facility in 2022, initially focusing on Bitcoin mining. The company has since aggressively pivoted the site toward artificial intelligence and high-performance computing workloads, targeting a combined capacity of 500 to 750 MW across multiple buildings by the end of 2026. Current tenants include Core42 and Fluidstack, both of which are linked to technology operations backed by Google credit.

The fire originated in a structure that remained under construction. Following the incident, discussions reportedly centered on installing additional hydrants and implementing stricter storage protocols for safety data sheets. Despite these operational shortcomings, no regulatory follow-ups or investigative actions have been publicly reported in connection with the blaze.

TeraWulf’s transition from Bitcoin mining to AI and HPC leasing reflects a broader industry strategy adopted by numerous former cryptocurrency miners. Traded on the Nasdaq under the ticker WULF, the company’s ability to secure tenants like Core42 and Fluidstack underscores the commercial viability of repurposing legacy crypto-mining infrastructure for expanded compute workloads.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy .