Bitcoin miners are quietly becoming the power brokers of AI infrastructure as grid delays push project timelines past a decade
Building a data center used to be complicated. Now it’s an exercise in patience that would humble a geologist. Bernstein, the research firm, reports that clearing the current 338 GW pipeline of planned data center projects now takes an estimated 12 years, up from 10 years just a month ago.
That two-year jump in a single month tells you something important: the bottleneck isn’t capital, and it isn’t ambition. It’s the physical reality of connecting enormous power loads to grids that weren’t designed for them.
Why the pipeline keeps growing while projects stand still #
Global data center capacity sits at roughly 100 GW today. The projection is for that number to double to 200 GW by 2030. In English: the world needs to build as much data center infrastructure in the next five years as it built across the entire history of the industry.
Interconnection and permitting processes in critical U.S. regions now take anywhere from five to over 12 years. More than 60% of planned data center capacity targeted for 2027 is not yet under construction.
Bitcoin miners are sitting on something valuable #
Bernstein’s analysts have started calling certain miners the “power landlords of AI.” While traditional data center developers are stuck in permitting queues measured in years, miners already have operational power connections.
Bernstein highlights Iris Energy and Core Scientific by name, pointing to their pre-secured power pipelines as the core of their AI infrastructure thesis. The broader mining sector has accumulated pipelines exceeding 27 GW in total.
Announced deals linking Bitcoin miners to major AI infrastructure projects have crossed $90 billion in total value. Bernstein projects that AI-related revenue for mining operations could grow ninefold by 2030.
What investors should actually watch #
The risks are real too. Converting mining facilities to handle the specific power density and cooling requirements of AI workloads isn’t trivial. Not every site is suitable, and the capital costs of retrofitting can be substantial.
The interconnection backlog also cuts both ways. Yes, existing miners have power. But expanding that power, or securing new sites in response to AI contract wins, still runs into the same permitting queues.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our