IREN, a data center operator headquartered in Sydney, Australia, specializing in AI cloud infrastructure, announced two important plans: first, to reach an agreement with NVIDIA to purchase more than 50,000 B300 GPUs, expanding its AI computing power by about 50%; second, to submit a potential at-the-market (ATM) stock issuance plan, aiming to raise up to 600 million USD.

The purchased NVIDIA B300 GPUs are AI computing-specific graphics processing chips with large-scale parallel computing capabilities, capable of supporting the training and operation of artificial intelligence models at high speed. After the purchase is completed, IREN’s total AI computing GPU scale will reach about 150,000 units. These devices are expected to be deployed in phases in the company’s air-cooled data centers located in Mackenzie, British Columbia, Canada, and Childress, Texas, USA, in the second half of 2026. After full deployment, IREN’s AI cloud business is expected to achieve annualized revenue exceeding 3.7 billion USD, positioning it among the world’s larger AI cloud infrastructure providers.

In addition, after the company disclosed plans to carry out an ATM stock issuance plan of up to 6 billion USD, its stock price fell about 5% in pre-market trading.