Key Highlights

  • Six former Bitcoin miners — Cipher, Riot, HIVE, Marathon, Hut 8, and IREN — are all repositioning toward AI infrastructure.
  • Hyperscale partnerships with AWS, Microsoft, and NVIDIA have become the key differentiator among these companies.
  • Leased and owned power capacity is emerging as the critical bottleneck determining which miners can scale fastest.

A defining theme across the former Bitcoin mining sector this year has been a collective pivot toward AI infrastructure, as companies that built their businesses around cryptocurrency mining increasingly reposition themselves as providers of power capacity and computing infrastructure for artificial intelligence workloads. Six names in particular — Cipher, Riot, HIVE, Marathon, Hut 8, and IREN — illustrate both the shared logic and the differentiated execution paths behind this broader industry shift.

The common thread across all six companies is a recognition that the specialized infrastructure built for Bitcoin mining — substantial power procurement, cooling systems, and data center real estate — translates reasonably well to the demands of AI computing, provided companies can secure the right hyperscale customer relationships and upgrade their facilities accordingly. That has led each company to pursue partnerships with major cloud and AI compute providers.

Cipher has anchored its transition around a 15-year lease agreement with Amazon Web Services, supporting $11.4 billion in contracted hyperscale revenue across a 700-megawatt portfolio. Hut 8, meanwhile, has built its narrative around a $9.8 billion Beacon Point AI data center lease alongside a substantial $4.25 billion institutional debt raise to fund expansion. IREN has secured perhaps the most diversified set of hyperscale relationships, with confirmed deals worth $9.7 billion with Microsoft and $3.4 billion with NVIDIA.

Other companies in the group are still building out their infrastructure ambitions. HIVE is targeting a dramatic scale-up of its 320-megawatt Toronto AI Gigafactory, aiming to grow contracted revenue from $35 million to $660 million by 2028, while Marathon recently announced a Texas land acquisition expected to nearly double its power capacity to roughly 4.8 gigawatts by 2028 — though that announcement has been met with skepticism from at least one major analyst regarding execution risk.

Across all six names, power capacity has emerged as the critical bottleneck and, increasingly, the primary metric investors use to differentiate between companies in this space. As the sector matures, the companies that can most efficiently convert secured power and leased capacity into signed, revenue-generating hyperscale contracts are likely to see the greatest re-rating as their AI infrastructure businesses scale beyond their original crypto mining roots.