IREN (IREN), a data center operator focused on AI cloud infrastructure, said it agreed to buy more than 50,000 specialized processing chips from Nvidia (NVDA), expanding its capacity by about 50%.
The B300 GPUs, or graphic processing units, will take the Sydney-based company’s total AI compute fleet to about 150,000 GPUs. A GPU is a specialized chip for performing large numbers of parallel computations, enabling the training and operation of artificial intelligence models at high speed.
The company also filed for a potential at-the-market share sale of up to $6 billion as part of its broader capital management strategy. The shares dropped 5% in pre-market trading on Thursday due to potential dilution.
The additional hardware is expected to be deployed in phases through the second half of 2026 across the company’s air-cooled data centers in Mackenzie, British Columbia, and Childress, Texas. Once fully deployed, the expanded fleet is projected to support more than $3.7 billion in annualized AI cloud revenue, positioning IREN among the larger AI cloud infrastructure providers globally.
IREN said it has secured about $9.3 billion in funding over the past eight months through customer prepayments, convertible notes, GPU leasing and financing arrangements, with roughly $3.5 billion in additional capital expenditures expected for the new GPU deployments in the second half of 2026.
