While the headlines of the artificial intelligence boom are dominated by the software giants and the architects of large language models, a more understated group of winners has emerged from the wreckage of the cryptocurrency crash. A handful of crypto miners turned AI stars are leveraging a surprising legacy asset to secure their place in the new economy: massive utility power contracts.
For companies like TeraWulf, Applied Digital, Iren, Core Scientific, and Cipher Digital, the transition from minting Bitcoin to hosting high-performance computing (HPC) clusters is not just a change in business model, but a strategic pivot based on scarcity. In the race to build AI infrastructure, the primary bottleneck is no longer just the availability of GPUs, but the ability to plug them into a power grid that is increasingly strained.
This “legacy wattage” has allowed these firms to bypass the years-long queues typically associated with securing industrial-scale electricity. By repurposing sites originally designed for the energy-intensive process of cryptocurrency mining, these companies have become indispensable partners for hyperscale tenants and cloud providers desperate for immediate capacity.
The financial rewards of this pivot have been stark. According to Nick Giles, a senior research analyst at B. Riley Securities, the cumulative market capitalization of 11 leading former crypto-mining firms surged from approximately $2.1 billion in late 2022 to roughly $48.5 billion today. Giles noted that the firms have been “more successful than anyone — including themselves — could have anticipated.”
The operational gap: Mining vs. AI infrastructure
Despite the shared require for electricity, the technical requirements for an AI data center are exponentially more demanding than those of a crypto-mining farm. While mining facilities are designed for raw power consumption, top-tier AI facilities require extreme reliability and sophisticated climate control.
AI clusters generate tremendous waste heat that requires vast, expensive cooling systems to manage. While a Bitcoin miner can tolerate brief outages without catastrophic failure, AI tenants demand near-zero downtime, necessitating redundant backup generators and high-end power conditioning equipment. This transition requires a massive capital expenditure that far exceeds the costs of the mining operations these companies previously managed.
Brian Dobson, a managing director at Clear Street, suggests that the industry is entering a critical transition phase. “The theme of 2025 was that everyone would get out and get a contract with a major customer,” Dobson said. “2026 is the year of expanding on those contracts and, more importantly, executing on those deals.”
High-stakes execution and the risk of penalties
The market has already priced in the success of these pivots, but the actual construction of these facilities remains a volatile process. Many of these firms are now managing multibillion-dollar projects that are susceptible to supply chain disruptions and grid interconnection delays. If these companies miss their deadlines, the financial consequences could be severe.
A clear example of this risk can be found in the public filings of TeraWulf. In a filing with the Securities and Exchange Commission, the Maryland-based company detailed a deal with Fluidstack and Google to expand a campus in Western New York to 360 megawatts. The filing revealed that if the project is delayed by 180 days or more, Fluidstack would have the right to terminate the lease, and a $3.2 billion rent guarantee from Google could be voided.
Such penalties could fundamentally alter the borrowing costs for these projects. For the Lake Mariner project, losing the Google backstop would likely increase refinancing costs, potentially destabilizing the project’s economics. As Giles observed, 180 days may seem like a reasonable window, but in the world of “mega-construction,” such timelines can vanish quickly.
Comparative Capacity and Partnerships
| Company | Planned/Active Capacity | Key Mentioned Partners |
|---|---|---|
| Applied Digital | 5 Gigawatts | Oracle, CoreWeave |
| Cipher Digital | 4.1 Gigawatts | Amazon, Google/Fluidstack |
| TeraWulf | 360 Megawatts (NY site) | Google/Fluidstack |
The lending market’s verdict
Despite the construction risks, the debt markets have shown increasing confidence in these firms, offering lower interest rates as they secure partnerships with “investment-grade” tenants. The creditworthiness of the tenant often dictates the cost of the loan for the data center developer.
For instance, Cipher Digital secured a $2 billion loan in February at a 6.1% rate for a Texas facility being built for Amazon. In contrast, a $1.4 billion loan for a project involving Fluidstack and Google in November 2025 carried a higher rate of 7.1%. Similarly, Applied Digital borrowed $2.15 billion at 6.8% for an Oracle project in March, a significantly lower rate than the 9.3% it received for a $2.35 billion loan for a facility being developed for CoreWeave.
This disparity highlights the importance of the “hyperscale” label. When a company like Amazon or Oracle guarantees the revenue stream, the risk to the lender drops, allowing the developer to scale faster and more cheaply.
A safety net of scarcity
There is a prevailing sentiment among analysts that the sheer scarcity of power may protect these firms even if they stumble. Because the demand for AI computing power is so high, tenants may be more forgiving of delays than they would be in a saturated market.
Paul Golding, a senior digital infrastructure analyst at Macquarie, noted that customers have generally not changed terms despite some infrastructure delays. He explained that the scarcity of power has created a dynamic where seeking an alternative site would likely accept far longer than simply waiting for a delayed project to be completed.
For the former crypto miners, the immediate future depends on their ability to transform “legacy wattage” into operational, cooled, and stable data centers. The next major checkpoint for the industry will be the 2026 delivery cycle, where the ability to execute on these massive contracts will determine which firms remain stars and which are undone by the complexities of physical infrastructure.
Disclaimer: This article is for informational purposes only and does not constitute financial, investment, or legal advice.
We invite readers to share their perspectives on the AI infrastructure race in the comments below or via our social channels.
