Microsoft Corp. has entered into a $9.7 billion, five-year agreement with Australian infrastructure provider IREN Ltd. to acquire AI cloud capacity, marking a major expansion of the tech giant’s artificial intelligence infrastructure. The deal makes Microsoft IREN’s largest customer to date.
Under the agreement, Microsoft will gain access to Nvidia accelerator systems based on the GB300 architecture at IREN’s facility in Texas, designed to support advanced AI workloads. The contract includes a 20% prepayment, according to a company statement released Monday.
As part of the deal, IREN will purchase GPUs and related equipment worth $5.8 billion from Dell Technologies Inc. Once fully operational, the project is expected to generate around $1.94 billion in annualized revenue, IREN Chief Executive Officer Daniel Roberts said via email. The deal will utilize roughly 10% of IREN’s total capacity, leaving room for further client expansion and additional contracts.
“We’ve always seen major hyperscalers as natural partners,” Roberts said. “Our discussions with several of them have accelerated as their compute requirements — and our AI cloud capabilities — continue to grow.”
Following the announcement, IREN’s shares surged over 28% in U.S. pre-market trading, while Dell’s stock rose about 4%. Microsoft’s shares were little changed.
IREN belongs to a group of so-called “neocloud” providers — including CoreWeave Inc., Nebius Group NV, Crusoe Inc., and Nscale — that specialize in AI data center operations and compete to supply computing power to major hyperscalers such as Meta Platforms Inc. and AI companies like OpenAI. Many of these firms, including IREN, originally started as Bitcoin mining companies before pivoting to AI infrastructure.
IREN’s NASDAQ-listed stock has already surged more than 500% in 2025, fueled by investor optimism over the global AI boom that has also propelled Nvidia to a $5 trillion valuation.
Microsoft has been increasingly relying on leasing agreements with such AI infrastructure providers to meet growing cloud and AI demand. The company recently acknowledged capacity constraints in its Azure cloud business amid rising demand from clients like OpenAI.
The new GB300 systems in Texas will be installed in stages through next year at IREN’s Childress facility, which will support up to 750 megawatts of capacity. The company also operates a 2-gigawatt hub in Sweetwater, near Abilene, Texas, and reports strong demand for large-scale AI infrastructure deployments there, Roberts added.
