Why Bitcoin Miners are Perfect for AI?

There is an interesting change in Bitcoin mining. The ASIC chip manufacturers have been constructing large warehouses over the years. They have now become some of the hottest landlords in AI. And no, it is not just a coincidence or money-grabbing. Bitcoin miners are literally in a position to satisfy the growing computing power that AI needs.

The Problem with the Infrastructure

It is inconceivably hard to create a new AI data facility. This is not money restraining it, nor is it technology. It is power. The process of acquiring new grid hookups and required permits takes years. It could take years to acquire land in the substation area, secure long-term power purchase contracts, build cooling plants, and obtain zoning permission. 

Bitcoin mining companies have already done this to a great extent. Most mining campuses already control significant interconnection rights, substation capacity, and grid access. Those are usually the most challenging and slowest parts of any new data center project. 

Their choice of location was determined by cheap, reliable electricity and high capacity. They invented cooling systems capable of handling high thermal loads 24 hours a day. They are automated in terms of scale monitoring and management. In other words, they created what hyperscalers in AI are desperately missing.

Mining companies are realizing that they can be successful in the new economy of AI infrastructure. They already have experience in power procurement, thermal management, and the operation of bare-metal servers.

The Conversion is Faster Than You Think

The most widespread myth is that it would require years to turn a Bitcoin miner into an AI-ready data center. Actually, it can take less time than erecting a new facility. The simplest part of the work is removing the old Bitcoin mining machines (ASICs) and racking them down. Installing regular data center racks in their place could take less time.

The mining halls are made to be exceptionally power-dense and are thermally loaded. It is possible to redesign them into liquid-cooled GPU systems with a fairly minor redesign. The dense rack design, high power delivery, powerful cooling, and the requirements for operating thousands of Bitcoin miners in one location are the same features that modern AI hardware requires. 

The economics are tempting. Estimates show that miners can monetize between $1 million and $4 million per megawatt per year to transform current locations to suit AI. By contrast, building a shell and infrastructure for new AI-hosting locations would require $6 to $8 million per megawatt per year. The costly equipment is already in the miner’s possession.

What the Numbers Say?

The business is changing, and the companies are adapting. An example of this repositioning is Core Scientific, which does not just position itself as a Bitcoin miner, but as a developer of high-powered data centers that can support AI and other intensive computing work. The proof lies in the deals that they are signing. They have formed a partnership with CoreWeave to provide approximately 590 megawatts of capacity in 6 locations. That is $10.2 billion of guaranteed income over the life of these contracts. 

MARA Holdings has identified its power-concentrated locations as customer-facing, which require consistent, large-scale access to energy. It is now developing data centers for AI- and HPC-specific workloads. It collaborated with Starwood Digital Ventures in the development of AI data centers with the potential of up to 2.5 gigawatts of IT capacity.

On the other hand, Bitdeer has entered AI and high-performance computing and continues to grow its mining operation. This combination of high mining potential and AI development suggests that management believes in the industry’s rising AI potential.

One Site, Two Income Streams

The majority of these places are not only crypto mining places. They are computer infrastructure resources that can mine Bitcoin, run AI workloads, or do both simultaneously, depending on which yields the highest payoff per megawatt.

Fred Thiel, chairman and CEO of MARA, has expressed the opinion that Bitcoin mining doesn’t compete with AI for power. It just uses the same facilities differently. The mining systems can possess dynamically varying energy usage. Therefore, together with AI activities, they can offer a flexible and smooth power pattern. Miners can easily power down to allow other users access to the grid during peak demand or surging AI workloads.

This may be termed as a hybrid data center, where Bitcoin mining is used as a scalable load, and AI colocation is used as a source of contracted base revenue, which sites like Webopedia describe as a data center that combines physical, virtual, and cloud computing platforms to efficiently handle workloads of all types. 

Those companies that have embraced this model may be more resilient than pure-play mining and less capital-intensive than building greenfield AI data facilities.

The post Why Bitcoin Miners are Perfect for AI? appeared first on The Market Periodical.