📰 Key Highlights

Bitcoin miner TeraWulf announced a major partnership with Anthropic, agreeing to lease and build a dedicated AI data center at its Justified Data campus in Hawesville, Kentucky. The facility is designed for 401 MW of critical IT capacity, with initial operations expected in H2 2027 and full buildout by early 2028.

At the same time, TeraWulf announced the sale of its 50.1% stake in the Abernathy AI data center joint venture in Texas to an investment group led by its partner Fluidstack. The deal is expected to recover roughly $450 million in invested capital, all of which will be reinvested into the company’s wholly-owned AI infrastructure projects.

Behind this wave of miners pivoting into AI is a sustained market pressure: AI compute demand keeps outstripping available supply. Training and running large AI models requires high-performance chips, advanced cooling systems, and abundant, stable power, which has sent valuations soaring for campuses with rich power resources. Bitcoin miners already own much of the infrastructure needed for high-energy compute — grid connections, power agreements, and more — making them natural candidates for the transition.

That said, this pivot doesn’t come cheap. Blocksbridge Consulting estimates that publicly listed Bitcoin miners building out AI infrastructure may collectively need around $50 billion in capital in the near term, far exceeding what traditional mining facilities require. Around the same time, HIVE Digital signed a three-year, $220 million GPU cloud infrastructure deal to provide compute for AI startup Cohere, while IREN acquired Spanish data center developer Nostrum Group, securing roughly 490 MW of confirmed grid connection capacity and formally entering the European AI market.

💬 JudyAI Lab Perspective

TeraWulf’s deal with Anthropic signals that the trend of Bitcoin miners transitioning into AI compute providers is moving from concept to large-scale, long-term infrastructure commitments — and it’s worth keeping a close eye on.

This case reveals a key resource-reallocation logic: the AI compute shortage isn’t just a chip problem — it’s a power and campus problem. Miners already hold the hardest-to-acquire assets, including grid connections and large-scale power agreements, so their pivot costs are far lower than developers starting from scratch. The fact that Anthropic chose to have TeraWulf build a dedicated 401 MW facility rather than handle it entirely in-house also reflects how large AI companies are actively seeking compute partners instead of trying to absorb all infrastructure buildout themselves. We estimate that listed miners collectively need around $50 billion in capital to push into AI infrastructure, a scale that far exceeds traditional mining — this isn’t dipping a toe in the water, it’s industry-level capital reallocation. The parallel moves by HIVE and IREN further confirm this is a collective pivot, not an isolated one.

For builders tracking the AI infrastructure landscape, now is a good time to start watching which non-tech-background companies are emerging as new compute suppliers — the map of this market is being redrawn fast.

📅 Source Info

🔗 Further Reading