IREN expands liquid-cooled AI capacity, with billions riding on the build-out

IREN is expanding liquid-cooled data-centre capacity for AI computing, including three further 50-megawatt deployments targeted for delivery in the fourth quarter of 2026. The expansion could help the company use existing power infrastructure more intensively, but its projected capital spending makes delivery and customer demand the rather large asterisks.

IREN Watch analysis

What happened

IREN says it delivered a 50MW IT deployment for Microsoft at its Childress, Texas site in August. It is targeting three more 50MW deployments for the fourth quarter, and plans liquid-cooled GPU installations at Mackenzie, Canal Flats and Prince George in British Columbia, with further expansion planned for 2027.

A Zacks Investment Research article republished by Yahoo Finance says IREN reported AI Cloud Services revenue of $128.8 million in fiscal 2026, up from $16.4 million in fiscal 2025. It also says contracts secured in August represented more than $20 million in revenue per MW of IT load over their terms, while active pricing discussions indicated about $25 million per MW. These are figures reported in the article, not a guarantee of future returns. Read the Yahoo Finance article.

Why it matters

Liquid cooling can support higher-density computing, giving operators a way to put more powerful GPUs into facilities where power is available. For IREN, the question is whether that capacity can be delivered on schedule and converted into sustained revenue.

The investment is substantial: the article puts IREN’s projected fiscal 2027 capital expenditure at $25 billion to $30 billion. Construction delays, financing needs, GPU obsolescence and customer demand are among the risks it identifies. More capacity is useful only if the economics survive the build-out.

Our read

This is a concrete AI infrastructure expansion, not just another forecast about how much computing the industry might need. The delivery targets and revenue figures give readers something to measure against. But the projected spending is enormous, and the article’s pricing figures do not settle whether the returns will justify it. Keep an eye on completed deployments and customer revenue, not just megawatts announced.

What to watch

• Whether the three targeted 50MW deployments arrive in the fourth quarter.
• Progress on the planned British Columbia GPU installations.
• Whether AI Cloud Services revenue grows enough to support the projected capital spending.

Discussion spark: Should IREN prioritise expanding AI capacity while demand is strong, or slow its build-out until the returns are clearer?

not affiliated with or endorsed by IREN