IREN delivers first 50MW Microsoft AI cloud build at Childress under $9.7 billion deployment plan

Microsoft is taking delivery of the first 50-megawatt phase of an AI-focused cloud deployment being built by IREN at its Childress, Texas site, according to a report carried by Yahoo Finance. The delivery marks an early milestone in a larger rollout that the reporting ties to IREN’s $9.7 billion contract with Microsoft.

The deployment delivered is described as part of a sequence of four liquid-cooled data center builds. Liquid cooling, in this context, is the use of water or other cooling liquid to remove heat from computing equipment more efficiently than traditional air-only cooling, a factor that matters as AI workloads increase power density and heat output.

Under the reported arrangement, Microsoft has accepted the first of those four deployments. The acceptance suggests the initial phase has met whatever commissioning or readiness criteria were set for the opening of that portion of the site, although the report does not spell out test results, performance targets, or formal acceptance documentation.

The reporting also links NVIDIA’s involvement to the same broader push. NVIDIA has awarded “Exemplar Cloud” status to the GB300 system after testing, indicating that the environment where those systems are deployed has been evaluated against NVIDIA’s readiness standards. “GB300” refers to a specific NVIDIA system configuration, and “Exemplar Cloud” is a designation used to highlight qualified cloud deployments for NVIDIA’s platform.

For Microsoft, these kinds of infrastructure milestones are typically less about the headline size of a site and more about the start of reliable compute capacity for training and inference, especially as AI demand rises. The early delivery at Childress aligns with that pattern, where hyperscale operators and their infrastructure partners move in phased increments to keep production timelines on track.

For IREN, a first delivery is also a cash flow and execution announcement. Large multi-phase data center programs often depend on meeting engineering and operational checkpoints before subsequent work can proceed at full speed. While the report emphasizes delivery and acceptance, it does not disclose payment schedules, any change orders, or whether future phases are on the same timeline.

On the technology side, liquid cooling and system-level validation through NVIDIA’s programs both reflect how AI infrastructure has become tightly coupled to hardware and platform qualification. As a practical matter, customers often need assurance that the full stack, including cooling, power, networking, and system management, will perform as expected under real workloads, not just in laboratory conditions.

Still, several items are not disclosed in the reporting. The post does not provide the reported deployment’s exact commissioning dates, the capacity breakdown by cluster or rack, the specific nature of the “testing” tied to NVIDIA’s Exemplar Cloud designation, or any performance metrics for the Childress environment. It also does not say whether the remaining three liquid-cooled deployments have been accepted yet or what pacing targets Microsoft and IREN are using for the rest of the program.