TeraWulf’s AI hosting revenue tops bitcoin mining at $21M

TeraWulf reported $21 million in high-performance computing (HPC) hosting revenue in the first quarter of 2026, surpassing roughly $13 million from bitcoin mining and bringing total quarterly revenue to about $34 million.

The company has repurposed facilities and grid capacity originally built for crypto mining to serve AI and cloud customers. Chief Executive Paul Prager described the quarter as ‘the first period where HPC leasing is meaningfully reflected in our financials.’ Chief Financial Officer Patrick Fleury characterized the company as ‘a business in transition’ and wrote that management expects revenue to be increasingly tied to ‘stable, contracted’ compute agreements.

TeraWulf recorded a net loss of $427.6 million for the quarter, compared with a loss of $61.4 million in the year-ago period. Operating expenses approached $200 million. Management attributed higher costs to investments in data center expansion and AI-related infrastructure and to impairment charges tied to scaling back parts of its bitcoin mining operations.

At quarter end TeraWulf had 60 megawatts of operational HPC capacity at its Lake Mariner facility in New York, where AI cloud provider Core42 is a long-term lessee. The company is developing additional capacity at Lake Mariner and coordinating deployments with customers including Fluidstack and Google.

Outside New York, the Hawesville, Kentucky, project carries about 480 megawatts of grid-connected power. A Maryland site could expand to roughly 1 gigawatt pending regulatory approval. TeraWulf previously disclosed more than $12.8 billion in long-term AI and HPC contracts tied to 522 megawatts of critical IT capacity.

The company plans to add between 250 and 500 megawatts of new contracted capacity each year and reported approximately $3.1 billion in cash and restricted cash at quarter end.

Demand for AI compute has increased the need for large, reliable supplies of electricity, cooling and ready-built data center space. The International Energy Agency projects electricity consumption for data centers worldwide could nearly double to roughly 945 terawatt-hours by 2030. American Electric Power forecast about 63 gigawatts of incremental electricity load by 2030 driven largely by data center demand.

Morgan Stanley analysts Stephen Byrd and James Faucette noted growing willingness among some AI customers to pay premiums for rapid access to power, often described as ‘time to power’ premia.

TeraWulf reiterated its intent to prioritize AI hosting over legacy bitcoin mining operations and to scale the business toward recurring, contracted revenue streams.