On 13 August 2026 IREN Limited announced, in a release furnished to the SEC on Form 8-K, that Horizon 1 has been delivered to and accepted by Microsoft, and that IREN has achieved NVIDIA Exemplar Cloud status on the NVIDIA GB300 NVL72 platform. Horizon 1 is the first of four 50 MW IT-load, direct-to-chip liquid-cooled AI cloud deployments scheduled for delivery to Microsoft during 2026 at IREN’s Childress, Texas campus, under a five-year cloud services contract worth 9.7 billion dollars that the two companies announced in November 2025.
The Exemplar Cloud designation followed NVIDIA testing of the GB300 NVL72 deployment at Horizon 1. IREN describes it as demonstrating an ability to support demanding AI workloads at performance, reliability and scale; it is a vendor certification, so it says what NVIDIA is willing to certify rather than providing an independent benchmark. Co-CEO Daniel Roberts credited more than 3,000 people across the site team and said IREN expects to deliver Horizons 2 through 4 later this year.
IREN reaffirmed its capacity targets in the same release: a broader expansion to 480 MW of gross AI cloud capacity in 2026 and 1.2 GW in 2027. The company positions itself as vertically integrated, controlling design, engineering and construction of the data centers that house its GPU deployments, and points to a portfolio of land and grid-connected power in renewable-rich regions across North America, Europe and Asia-Pacific. Those forward capacity figures are targets carried in a release explicitly flagged as containing forward-looking statements.
The reason this matters more than a routine build update is that it converts a headline contract into delivered, accepted capacity. Multi-billion-dollar neocloud contracts have been announced far faster than they have been fulfilled, and the gap between signing and revenue recognition is where several of these businesses will be judged. One accepted 50 MW block against a 9.7 billion dollar five-year commitment is roughly a quarter of the Microsoft programme’s 2026 scope, on schedule, at a hyperscaler’s acceptance standard. It also shows how far Microsoft’s own AI capacity now depends on third-party operators rather than only on its own data centers, which is a supply-chain concentration worth tracking in both directions.