IREN secures 4.5GW of GPU‑ready capacity to host large cluster deployments
- IREN said its new 1.6GW Oklahoma campus pushes secured grid-connected power past 4.5GW, giving the company a much bigger runway for hyperscale AI deployments. (nasdaq.com) - The concrete detail is timing: grid studies are complete, Oklahoma power is scheduled to start ramping in 2028, and Childress already anchors a 750MW site. (nasdaq.com) - Why it matters: hyperscalers increasingly need power first, and IREN is selling itself as a shortcut around the grid queue. (iren.com)
AI infrastructure is turning into a power business. GPUs still matter, obviously, but the thing buyers keep running into is not just chip supply — it’s whether anyone can(nasdaq.com)ers. That is why IREN’s latest expansion matters. By adding a 1.6GW campus in Oklahoma, the company says it now has more than 4.5GW of secured grid-connected p(nasdaq.com). (nasdaq.com) ### What did IREN actually secu(iren.com) number is total secured grid-connected power across multiple campuses, with the big new piece being a 1.6GW Oklahoma site spread over roughly 2,000 acres. IREN describes that campus as high-bandwidth and low-latency, which is shorthand for “this is meant for serious AI deployments, not just cheap power on empty land.” (iren.com) ### Why is power the story? Because a modern AI cluster is basically a data center wrapped around an electrical substation. Training and inference at f(nasdaq.com)ntinuous power. You can buy GPUs, but if the site is stuck in interconnection studies or waiting years for transmission upgrades, the cluster does not ship on time. That is the pitch behind IREN’s whole build-to-suit and colocation business — ready-to-build sites with grid access already lined up. (iren.com) ### Why is Oklahoma the important piece? Because it e(iren.com) Childress, Texas. Childress is a 750MW liquid-cooled site and already hosts the company’s biggest AI ambitions, including the Microsoft contract announced in November 2025. Oklahoma adds another very large option for future phases and for customers that need multiple campuses, not just one big building. (iren.com) ### How big are the clusters we’re talking about? Big enough that old cloud language starts to break. IREN’s Microsoft deal was framed around 200MW of critical IT load delivered in fo(iren.com)company said in March it was expanding AI Cloud capacity to 150,000 GPUs. So when IREN talks about hyperscaler deployments, it is pointing at cluster sizes that look more like utility projects than normal enterprise hosting. (iren.com) ### Is the 4.5GW all for AI? Not in the sense that every mega(iren.com)irection is clear. IREN started as a bitcoin miner and has been repositioning itself as an AI cloud and data center operator. Its own materials now emphasize AI-optimized campuses, colocation, build-to-suit deployments, and liquid-cooled infrastructure rather than just mining economics. (iren.com) ### What is the catch? Time. The Oklahoma site’s grid studies are complete, but power there is scheduled to begin ramping in 2028. So the headline numbe(iren.com)U fleet; the 4.5GW figure is really a statement that IREN wants to stay in the room when hyperscalers plan the next several years, not just the next few quarters. (nasdaq.com) ### Why are investors focused on this now? Because hyperscalers are spending like the bottleneck is physical infrastructure, not software. If AI de(iren.com) the ability to stand up large clusters fast. A specialist operator with land, substations, cooling design, and queue position can become more useful than a generic cloud reseller. That is what IREN is trying to be. (youtube.com) ### Bottom line The news is not just that IREN got bigger. It is that the company now has a power map big enough to matter in hyperscale AI conversations. In this m(nasdaq.com)t buyers cannot easily build for themselves on their own timeline. (nasdaq.com)
