IREN delivers Horizon 1 AI cloud to Microsoft, earns NVIDIA Exemplar status on GB300 NVL72

IREN Limited announced on August 13 that its first AI‑cloud deployment, Horizon 1, has been delivered to Microsoft and officially accepted. The 50 MW, direct‑to‑chip liquid‑cooled facility marks the opening of a four‑phase rollout under a five‑year, $9.7 billion cloud services contract. In parallel, NVIDIA certified the installation as an Exemplar Cloud on the GB300 NVL72 platform, confirming performance and reliability benchmarks. The milestone showcases IREN’s vertically integrated approach and signals a new scale of AI infrastructure for enterprise customers. Industry observers see the move as a bellwether for how hyperscale providers will source AI compute in the coming years.

What happened

IREN reported that Horizon 1, a 50 MW direct‑to‑chip liquid‑cooled AI cloud module, has been handed over to Microsoft at its Childress, Texas campus and is now operational. The deployment is the first of four planned modules slated for 2026 under a five‑year, $9.7 billion agreement announced in November 2025.

NVIDIA’s testing of the GB300 NVL72 system at Horizon 1 earned the Exemplar Cloud designation, indicating the installation meets stringent performance, reliability, and scalability criteria for demanding AI workloads.

Looking ahead, IREN disclosed plans to expand its AI‑cloud footprint to 480 MW (gross) in 2026 and 1.2 GW (gross) by 2027, leveraging its vertically integrated design, engineering, and construction capabilities.

Why it matters

The delivery validates a vertically integrated model that can compress timelines for massive AI infrastructure, a critical advantage as demand for GPU‑heavy training and inference accelerates. Microsoft gains immediate access to a high‑density, NVIDIA‑certified compute pool, reducing reliance on third‑party colocation and potentially lowering total cost of ownership. For the broader market, IREN’s Exemplar status signals that newer liquid‑cooled architectures can meet or exceed traditional air‑cooled data center performance, nudging competitors toward similar designs.

ProsCons

How to think about it

Enterprises evaluating AI‑cloud partners should weigh three factors: (1) the provider’s ability to deliver integrated, certified hardware on a tight schedule; (2) the scalability path and whether the roadmap aligns with projected AI workloads; and (3) the risk profile of the underlying GPU ecosystem, including cooling technology and supply‑chain resilience. Matching these criteria to business objectives helps determine if a vertically integrated, NVIDIA‑validated solution like IREN’s Horizon 1 fits the organization’s compute strategy.

FAQ