For years, the race to build ever-larger AI models and amass computing muscle dominated headlines, yet few predicted that energy supply would emerge as the sector’s defining constraint. As AI systems train and process huge volumes of data, their appetite for electricity ramps up exponentially. In the US, grid connections for new high-load projects often require multi-year waits, turning access to power into a rare and coveted resource.​
This gridlock has given rise to an unanticipated group of industry power brokers: cryptocurrency miners. Their experience securing significant energy contracts and their pre-existing data center footprints have made them natural allies for AI’s compute-intensive ambitions. Companies like Iris Energy (IREN) have experienced dramatic stock rallies after converting some operations from Bitcoin mining to high-performance AI data centers, locking in lucrative contracts with hyperscalers such as Microsoft and showcasing the untapped value of existing power agreements.​
Not all players are rushing headlong into the AI gold rush. Journalist Emma Rose highlights Bitdeer Technologies Group, steered by industry veteran Wu Jihan, which controls nearly 3 gigawatts of global power resources. Instead of abandoning crypto altogether, Bitdeer is taking a measured route, maintaining profitable mining while gradually upgrading select sites for AI workloads, balancing the risk and reward inherent in pivoting to a new frontier.​
Bitdeer’s VP of Global Data Center Business, Wang Wenguang, sheds light on the practical challenges: US power shortages stem from stressed, aging grids rather than regulatory delays. Crypto miners, having secured much of the available capacity, now hold the keys AI firms need to unlock further expansion, leading to a wave of collaboration across the sector.​
Wang is keen to point out another misconception: AI inference, not just model training, requires hefty power investments, especially as applications scale across industries. This surge has forced miners to invest in advanced cooling tech and rethink their resource strategies to keep pace with evolving demand.​
AI-focused data centers differ dramatically from traditional internet server farms, morphing into sophisticated, power-optimized electrical projects. Infrastructure upgrades, redundancy, and cooling become top priorities to handle thousands of power-hungry GPUs—far beyond the needs of Bitcoin ASICs.​
In this arms race, the winner is often the one who controls access to reliable energy. Just as crypto miners once hunted for cheap electricity to maximize profits, AI firms now find themselves negotiating with these same operators to secure grid connections swiftly, turning timelines of years into months.​
The transformation isn’t plug-and-play. It demands high-capacity upgrades: miners swap out ASICs for GPUs, overhaul cooling and power systems, and tighten security—all steps accelerated by their ownership of valuable energy assets, reducing conversion times compared to starting from scratch.​
Bitdeer’s Wang notes that AI data center sites are driven first by access to cheap, abundant power, often in remote regions where land costs are modest. This trend is evident across the US, where such partnerships are setting the tone for the next wave of infrastructure projects.​
The energy demands of modern AI, from chip supply to physical space, exceed those of almost any other digital industry. The scale and density required for advanced AI inference make efficient power provisioning the ultimate deciding factor in success.​
AI data centers in America increasingly rely on crypto miner partnerships and innovative, infrastructure-led designs. With heightened scrutiny on data residency and privacy, these sites are set to play an outsized role in global technology geopolitics, something Bitdeer monitors closely as it refines its own strategy.​
Bitdeer leverages both its existing global power portfolio, spanning the US, Canada, Norway, Ethiopia, and Bhutan and its engineering know-how to execute data center upgrades swiftly. Its expansion plans include converting mining sites to AI clusters in Singapore and Malaysia, plus a bid to raise up to $460 million via convertible notes for ongoing development.​
