Via ts2.tech
The 50-megawatt liquid-cooled facility in Texas marks the first of four deployments as IREN races toward gigawatt-scale AI infrastructure
IREN just handed Microsoft the keys to a 50-megawatt AI cloud facility in Childress, Texas, completing the first phase of what might be one of the largest cloud infrastructure deals quietly reshaping the AI landscape. The deployment, called Horizon 1, is the opening act of a $9.7 billion, five-year contract that aims to deliver 200 MW of total AI cloud capacity across four identical phases.
Microsoft has formally accepted the deployment, which means IREN is now earning revenue on what it spent the better part of a year building.
What IREN actually built #
The facility uses direct-to-chip liquid cooling, a technology that pumps coolant directly onto processors rather than relying on air conditioning to keep an entire room cold. The deployment runs on NVIDIA’s GB300 GPUs, and IREN earned NVIDIA Exemplar Cloud status for its GB300 NVL72 platform at the Horizon 1 site, meaning the infrastructure passed NVIDIA’s own performance benchmarks.
The $9.7 billion Microsoft contract, signed in November 2025, includes a 20% prepayment from Microsoft covering dedicated NVIDIA GPU capacity.
The scale ambition is staggering #
Horizon 1 is 50 MW. IREN wants to hit 480 MW of gross AI cloud capacity by the end of 2026. The roadmap doesn’t stop there: the company is targeting 1.2 gigawatts by 2027.
IREN’s approach is vertically integrated, meaning it handles site development, power procurement, and data center construction internally rather than outsourcing pieces to different contractors. Co-Founder and Co-CEO Daniel Roberts pointed to the company’s more than 3,000 team members as the engine behind the rapid execution.
Beyond Microsoft #
IREN has disclosed a pipeline of additional multi-year AI agreements potentially valued at $2.8 billion. The company is eyeing expansion across North America, Europe, and the Asia-Pacific region.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our