Northwise

Iren Prince George Site

Research Note · FreePublished · Updated

Prince George: IREN’s 50 MW campus, the move from mining to GPUs and the next planned liquid-cooled deployment.

In this article

Prince George is one of IREN's clearest examples of an existing electrical campus moving from Bitcoin mining toward AI computing. The property is smaller than the company's large Texas developments, but its role is different: it provides an established operating location in British Columbia while new campuses work through much longer construction programmes.

Its 50 MW power capacity is the starting point for understanding the site. It is not a count of commissioned GPUs, and it does not establish how much of the connection is carrying a particular customer's workload today. The distinction matters when equipment is installed in stages and a new cooling configuration is planned alongside the existing one.

The IREN infrastructure overview places Prince George in the global network. Use the IREN Data Center Tracker to compare locations and their disclosed development stages.

The physical foundation

IREN's facility page identifies a 50 MW, 12-acre campus. Its equipment headline combines installed and on-order GPUs, so that headline should not be reproduced as a live installed fleet. A machine ordered for future delivery cannot yet provide the same service as one installed, tested and available to a customer. Prince George facility specifications.

The site's established electrical infrastructure makes it different from an early development project. That advantage is real, but conversion still requires a complete system. GPUs depend on rack power, cooling, storage, high-speed networking and software. An operating mining facility demonstrates an industrial use of electricity; it does not by itself establish that every part of a new AI deployment is ready.

From mining to AI

IREN's December 2025 filing described the displacement of Prince George mining activity for GPUs. That is a useful dated marker of the transition. It also means that old site descriptions centred on mining capacity should no longer be used as the current operating picture. IREN's December 2025 filing.

The economics of the replacement workload depend on more than the power connection. An AI operator must obtain equipment, find customers, deliver the contracted service and support it over the relevant term. Hardware availability and customer demand can advance at different speeds. An old mining hashrate and a new GPU order therefore cannot be treated as interchangeable measures of productive capacity.

The additional 2027 deployment

IREN's August results included Prince George in its planned 2027 Canadian liquid-cooled expansion. That places a further development stage beyond the earlier conversion. It remains a company schedule rather than a completed operating milestone. IREN FY26 results.

The practical question is how the new deployment will use the existing site's power headroom and what equipment, cooling and construction work is required. A campus can support different generations of hardware over time, but the transition has to be engineered. The presence of an electrical connection does not remove the need to test the entire installation under the conditions required by its customer.

Compare deployments rather than headlines

Prince George is part of a Canadian group with Mackenzie and Canal Flats. Portfolio GPU counts and contracted annualized revenue are not automatically attributable to one of those properties. A useful comparison keeps site capacity, installed equipment, on-order equipment and accepted customer service separate.

The same care applies to revenue illustrations. Multiplying GPU count by an hourly price and the hours in a year describes a scenario. It does not establish realized revenue, because the equipment may not operate for every hour, all capacity may not be sold, and the actual contract may use different terms. The original example retained below is an illustration of those mechanics, not a newly issued forecast for Prince George.

What would sharpen the operating picture

Site-specific commissioning and customer-acceptance disclosures would be more informative than another company-wide equipment total. They would show which part of the available power has become a delivered service and how the next development stage relates to it. Until then, the site should be described as an established campus progressing through AI deployment and further planned conversion, with the limits of the available site-level evidence kept visible.

The IREN company page contains the wider research and separately dated models. This factual site review is current through September 17, 2026.

Historical calculations retained from the original study

The following limited examples are preserved from the original publication. Their assumptions and outputs have not been revised. They are not current operating measurements, a new Northwise forecast or a statement of current contracted prices. The factual discussion above uses the latest reviewed evidence.

Infrastructure Variable

Example Value

GPU Hourly Pricing

$2–$5 per GPU hour

GPUs per Cluster

72 GPUs

Hours per Year

8,760 hours

Annual Revenue Potential per Cluster

Hundreds of thousands to several million dollars depending on utilization

Northwise Research

Get Northwise Research in Your Inbox

New research, model releases, portfolio updates, and major Northwise developments delivered directly to your inbox.

Reader discussion

Discuss the research

0 published

Create a Free account to join the discussion with a display name.

Create a Free Account

No comments yet. Start a thoughtful discussion.