Redmond, Washington, November 3, 2025- Microsoft has struck an approximately USD 9.7 billion cloud-services agreement with AI-cloud operator IREN that will give the software giant access to Nvidia’s latest GB300 GPUs over five years, part of a push to shore up urgently needed AI compute capacity without building as many new data centres as possible itself.
Under the arrangement, Microsoft will prepay roughly 20% of the contract value to IREN, money that IREN says will help fund a separate ~USD 5.8 billion equipment purchase agreement with Dell Technologies for Nvidia GB300 systems and ancillary hardware. The GPU systems are scheduled for phased deployment through 2026 at IREN’s Childress, Texas campus, where the company is building liquid-cooled facilities intended to deliver roughly 200 MW of critical IT capacity in the near term.
Microsoft framed the deal as a practical fix for a shortfall of AI infrastructure that has constrained the industry’s ability to scale advanced applications.
“IREN’s expertise in building and operating a fully integrated AI cloud, from data centers to GPU stack, combined with their secured power capacity, makes them a strategic partner,” Jonathan Tinter, president of business development and ventures at Microsoft, said in a company statement. “This collaboration unlocks new growth opportunities for both companies and the customers we serve.”
For IREN, the contract cements its rapid transformation from crypto-mining roots into an AI-cloud operator. The company says it operates nearly 2,910 MW of renewable-powered data-centre capacity across North America and has large campuses (including a 750 MW Childress site) designed to host high-density GPU deployments. IREN’s shares jumped more than 20% in premarket trading on the news.
Analysts say the deal lets Microsoft expand GPU capacity quickly while avoiding the long lead times and heavy capital outlays of building and connecting wholly new campuses, and it helps mitigate the risk that buying chips outright would leave the buyer exposed to rapid obsolescence as newer processors arrive.
The IREN-Dell equipment agreement will see hardware installed in phases at Childress and across IREN’s build-out plans through 2026; the initial liquid-cooled deployments are expected to target the high-power density typical of large-scale generative-AI training and inference clusters. Market reaction was swift: investors re-rated IREN on the prospects of large, long-dated revenue, while other vendors in the GPU supply chain also registered moves on the news.
The timing underscores how major cloud providers are increasingly turning to outside suppliers and creative commercial structures to secure scarce GPU capacity. Microsoft warned last week that its AI capacity crunch may persist into mid-2026, highlighting the strategic urgency behind such large-scale capacity agreements.
.jpg)
