According to DCD, cryptocurrency mining company CleanSpark has announced a partnership with immersion cooling provider Submer to create an integrated infrastructure platform for AI data centers. The collaboration, announced this week, will see CleanSpark focus on selecting, developing, building, and operating AI-focused campuses while Submer provides its cooling technology and modular data center expertise. CleanSpark CEO Matt Schultz stated the partnership positions them to deliver “AI capacity at gigawatt-scale” by combining CleanSpark’s energy portfolio with Submer’s liquid cooling expertise. The company, which describes itself as “America’s Bitcoin Miner,” currently operates around 33 Bitcoin mining sites across Wyoming, Tennessee, Mississippi, and Georgia with over 1GW of capacity under contract. This strategic move represents a significant shift in the crypto mining industry’s adaptation to new market opportunities.
Table of Contents
The Great Crypto-to-AI Infrastructure Migration
The CleanSpark-Submer partnership exemplifies a broader industry trend where cryptocurrency mining operations are repurposing their substantial infrastructure investments toward artificial intelligence workloads. These companies possess exactly what the AI boom desperately needs: massive power capacity, established grid connections, and experience operating energy-intensive computing operations at scale. The transition makes economic sense – as cryptocurrency mining profitability faces volatility, AI compute demand shows no signs of slowing. What’s particularly strategic about this move is that these companies aren’t just leasing out space; they’re leveraging their core competencies in power management and large-scale data center operations to become full-service AI infrastructure providers.
Why Immersion Cooling Changes the Game
Submer’s immersion cooling technology represents a critical enabler for this pivot. Traditional air-cooled data centers struggle with the extreme power densities of AI training clusters, which can consume 50-100kW per rack compared to 5-15kW for conventional servers. Immersion cooling allows for much higher power densities while reducing energy consumption for cooling by up to 95%. This efficiency gain is particularly valuable for companies like CleanSpark that have optimized their operations around power generation economics. The partnership suggests they’re not just retrofitting existing facilities but building purpose-designed AI infrastructure from the ground up, which could give them significant operational advantages over competitors attempting to adapt conventional data centers.
Strategic Risks in the Crypto-to-AI Transition
While the pivot appears logical, several challenges could undermine this strategy. First, the operational requirements for AI data centers differ significantly from Bitcoin mining operations. AI workloads demand different networking architectures, storage subsystems, and reliability standards. Second, these companies face competition from established cloud providers and specialized AI infrastructure firms with deeper expertise in machine learning workflows. Third, the capital expenditure required for this transition is substantial – while they may repurpose power infrastructure, the computing hardware and cooling systems represent massive new investments. Finally, there’s timing risk: if they’re too late to market, they may miss the current AI infrastructure gold rush; if too early, they could face underutilized capacity during the industry’s maturation.
The Emerging AI Infrastructure Ecosystem
CleanSpark joins a growing list of crypto miners making this strategic shift, including Hive Digital Technologies, Northern Data, and Iris Energy. What’s interesting is how each company is approaching the opportunity differently. Some are focusing on modular data center deployments that can be rapidly scaled, while others like CleanSpark appear to be pursuing larger campus-style developments. The modular approach offered by Submer’s technology provides flexibility but may face scalability limitations compared to traditional data center designs. The success of these ventures will depend on their ability to not just provide space and power, but to deliver the full stack of services that AI companies need – including specialized networking, storage, and management tools that traditional cloud providers have spent years developing.
Long-term Implications for AI Infrastructure
If successful, this crypto-to-AI migration could fundamentally reshape the AI infrastructure landscape. These companies bring a different economic model to the table – one optimized around power costs rather than traditional data center real estate economics. This could put downward pressure on AI compute pricing and create new capacity in regions that traditional providers have underserved. However, the sustainability of this model depends on continued AI demand growth and these companies’ ability to compete on reliability and performance, not just price. The next 12-18 months will be critical as we see whether these converted mining operations can deliver the uptime and performance that enterprise AI workloads require, or if they remain niche providers for specific use cases.
