How Former Bitcoin Miners Are Powering the AI Boom

Over the past year, a clear trend has emerged in the U.S. data center sector as former cryptocurrency mining operators redirect their assets toward AI and high-performance computing, leveraging existing power infrastructure, cooling systems, and real estate to meet a rapidly expanding market.

This trend is not just a pivot to the latest market narrative; it’s driven by business logic. Earlier this month, Microsoft signed a $9.7 billion, five-year agreement with former Bitcoin miner IREN to lease about 200 MW of GPU powered data center capacity for AI workloads. Around the same time, AWS announced a partnership with Cipher Mining to lease AI focused data center infrastructure, highlighting how mining operators are applying their expertise in large-scale, power intensive infrastructure to capture surging AI demand. These are just two examples of companies that once ran warehouses of ASIC miners now converting those same megawatts of power and cooling into AI factories.

Why Miners Are Well Positioned to Build AI Data Centers

The underlying logic is simple: Bitcoin mining and AI workloads both rely on extreme power density, efficient cooling, and scalable infrastructure. As demand for AI computing continues to rise, miners find themselves sitting on what the industry needs most, megawatts of ready-to-use power capacity.

• Power Infrastructure Already in Place: Most mining campuses already control large interconnection rights, substation capacity, and grid access, which is usually the hardest and slowest part of any new data center project. IREN’s Co-CEO Daniel Roberts highlighted how this existing infrastructure gives the company a head start in serving AI demand:

“We control the entire stack from the substation all the way down to the GPU,” he said, emphasizing IREN’s vertically integrated model.

• Energy Cost Advantage: Most miners operate in low-cost, high-availability regions like Texas, often under long-term PPAs or self-generation setups. These same agreements make AI hosting far more profitable per megawatt than mining.
• Facilities Built for Density and Scale: Mining halls are engineered for extreme power density and continuous thermal load, meaning they can be upgraded to liquid-cooled GPU systems with minimal redesign. During IREN’s latest earnings call, Co-CEO Daniel Roberts explained just how straightforward this transition can be:

“The vast majority of the work is removing ASICs, removing the racks that the ASICs sit on, and replacing those with standard data center racks … That is relatively minimal. As we’ve discussed before, it’s a matter of weeks to do that conversion.”

• Operational and Permitting Head Start: Sites are already zoned, permitted, and automated with monitoring and energy curtailment systems integrated into local grids. What once enabled flexible mining now enables dynamic AI workload management.

Who’s Making The Shift

• IREN Limited (IREN): Having begun as a bitcoin miner, IREN is actively pivoting to GPU-dominated AI cloud infrastructure. In 2025, the company announced expansion of its AI Cloud business and paused further bitcoin hashrate growth to focus on high-density AI data centres.
• Cipher Mining Inc. (CIFR): A U.S. based mining operator which in September 2025 signed a 168 MW, 10-year AI-hosting agreement and shortly thereafter a 15-year, $5.5 billion lease deal for 300 MW with a major public cloud provider (AWS) aimed at AI workloads.
• Riot Platforms Inc. (RIOT): Also in late 2025, the company repositioned its business, announcing development of ~112 MW of data centre capacity at its Corsicana, Texas campus and signalling a shift from pure bitcoin mining to data centre/HPC hosting.
• TeraWulf Inc. (WULF): In October 2025, TeraWulf struck a strategic partnership with Fluidstack to develop and deliver 168 MW of critical IT load at its campus in Abernathy, Texas campus
• Applied Digital (APLD): Once a large-scale miner, the company’s new “Polaris Forge” sites are purpose built for high-density GPU systems rather than ASICs.

Why This Pivot Makes Financial Sense

The economics are compelling. One estimate shows miners converting existing sites for AI hosting can earn $1-4 million per MW annually, or over $10 million/MW if operating GPUs directly. In contrast, greenfield AI builds cost $6-8 million per MW just for shell and infrastructure. Miners already own the expensive pieces like interconnections, substations, cooling, and land, so retrofit costs are minimal and timelines short. IREN’s co-CEO Daniel Roberts noted that converting ASIC halls to GPU racks takes only weeks, not years. The result is far higher revenue yield per MW with dramatically lower incremental CapEx.

What’s Next?

As AI demand accelerates, more projects are emerging from the repurposing of crypto mining sites into AI infrastructure. Those controlling power and land are emerging as the new backbone of the AI economy.

For a deeper look at the U.S. data center landscape, explore Aterio’s latest U.S. Data Center Report for state-by-state insights, provider comparisons, and project pipeline analysis, or schedule a call to learn how our data can enhance your investment strategy.