GPUs and RAM Are in Short Supply, but the Real Bottleneck for AI Is Electricians (www.nextplatform.com)

TeraWulf has shifted its focus from Bitcoin mining to becoming a Heavy Performance Computing (HPC) and AI infrastructure provider at its Lake Mariner datacenter in New York, expanding the facility to 750 megawatts of capacity by 2025. The site, which employs advanced liquid cooling technologies from Schneider Electric, highlights a growing trend toward optimizing datacenters for AI workloads, requiring significant economic investments—estimated at $7 million to $10 million per megawatt for AI infrastructure, compared to $500,000 per megawatt for Bitcoin mining. The shift underscores a crucial challenge for the AI/ML community: the scarcity of skilled electricians is now a bottleneck for datacenter development, as seen with TeraWulf's reliance on up to 800 electricians at any given time. With a blend of green energy sources and a robust power infrastructure, TeraWulf is well-positioned to support AI's growing demands, but the need for electrical contractors reveals the broader constraints facing the industry as it seeks to scale rapidly. As AI deployments surge, the community must address these logistical challenges to maintain momentum in technological advancements.