2 hours ago
20
Microsoft has formally accepted Horizon 1, the first phase of IREN Limited’s AI cloud deployment under a five-year contract worth approximately $9.7 billion. The milestone means IREN’s purpose-built liquid-cooled data center in Childress, Texas, is on track to deliver 50 MW of IT load by Q4 2025.
Inside the deal
The contract, first announced on November 3, 2025, commits Microsoft to purchasing dedicated GPU cloud services powered by NVIDIA’s GB300 GPUs across IREN’s Texas campus. The deployment is structured in four phases, labeled Horizon 1 through 4, with full buildout expected to reach 200 MW of total capacity by the end of 2026.
At full scale, IREN projects the arrangement will generate roughly $1.94 billion in annual recurring revenue.
Microsoft’s financial commitment goes beyond a handshake. The tech giant is providing a 20% upfront payment on the contract, a prepayment structure that gives IREN significant capital to fund the buildout.
On the hardware side, IREN has locked in a separate purchase agreement with Dell Technologies valued at approximately $5.8 billion. Dell will supply the infrastructure needed to house NVIDIA’s GB300 GPUs in liquid-cooled, Tier III-equivalent data centers designed to handle high-density racks running between 130 and 200 kW per rack.
The entire Childress campus will run on 100% renewable energy.
From Bitcoin mining to AI powerhouse
IREN’s trajectory is one of the more dramatic pivots in recent tech history. The NASDAQ-listed company built its early business around Bitcoin mining, an industry that shares more DNA with AI data centers than most people realize. Both require massive amounts of power, sophisticated cooling infrastructure, and the operational know-how to keep thousands of specialized processors humming around the clock.
The market clearly agrees. IREN shares surged as much as 30% following the original contract announcement, hitting record highs as investors priced in the revenue potential of the Microsoft partnership.
Why GPU capacity is the new gold rush
NVIDIA’s GB300 GPUs, the chips at the center of this deal, represent the company’s next-generation architecture designed specifically for the kind of large-scale AI training and inference workloads that Microsoft needs. Liquid cooling is essentially a requirement at the power densities these chips demand, which is why IREN’s facilities are built around that approach from the ground up rather than retrofitting air-cooled buildings.
The Tier III-equivalent designation means the data centers are designed for 99.982% availability, with redundant components that allow maintenance without taking systems offline.
With three more phases still to come and a full 200 MW buildout targeted by end of 2026, each subsequent Horizon phase will need to clear the same Microsoft delivery acceptance process, and any delays in power delivery, equipment installation, or commissioning could push back the revenue ramp.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) ·