The Weekly Brief — Issue 09

Hello,

Canada’s AI story is getting more physical.

In British Columbia, IREN is refitting data-centre capacity built around Bitcoin mining for AI workloads, backed by new GPU financing. In Saskatchewan, the province has drawn up rules for the next wave of data centres before dozens of proposals move forward.

Plus: more Canadian businesses say AI is in their plans, and Quebec’s election campaign now has a voluntary code for how political parties use it.

here is the week.

What happened

IREN is accelerating its shift toward AI computing, and three of its existing British Columbia data-centre sites are part of the plan.

In results released August 27, the company said it plans new liquid-cooled deployments at Mackenzie, Canal Flats and Prince George in 2027. At Mackenzie, a US$2.4-billion financing package led by Blue Owl and PIMCO is intended to fund about 90% of the associated GPU spending for an air-cooled expansion.

IREN’s broader business is changing quickly, too. Company-wide AI Cloud Services revenue rose from US$16.4 million in its 2025 fiscal year to US$128.8 million in 2026. The company is also decommissioning Bitcoin-mining equipment and transitioning more of its data-centre capacity toward AI workloads.

Not all of IREN’s AI growth is Canadian, and the company has not disclosed how much of its global customer business will ultimately run from B.C. But the physical investment happening at its Canadian sites is real.

Why it matters

AI may feel like software, but the systems behind it need very physical things: chips, buildings, cooling, electricity and enormous amounts of capital.

That makes data centres part of Canada’s AI story whether most Canadians ever see the servers inside them or not.

The important shift is not simply that AI needs data centres. It is that existing Canadian digital infrastructure is now being redesigned around AI demand.

And that creates questions much bigger than one company: where new computing capacity should be built, how much electricity it should consume, who pays for supporting infrastructure, and what communities get in return.

Those questions are already showing up in provincial policy debates.

Your move

When you see a large Canadian AI or data-centre announcement, look past the investment number and find four things:

How much electricity does the project need?

Is the company building new capacity or converting infrastructure that already exists?

Who is responsible for new power and grid infrastructure?

And what does the host community actually receive — jobs, tax revenue, investment, ownership, or something else?

Those details will usually tell you more than the words “AI investment.”

Saskatchewan wrote rules for the next wave of data centres

Saskatchewan released a new Data Centre Framework on August 27 built around six principles, including Canadian ownership, Canadian data and AI sovereignty, Saskatchewan jobs and partnerships, industry experience, self-supplied electricity and a centralized provincial intake process.

The province says more than 30 data-centre applications are already in its queue.

The framework requires new proponents to supply their own power rather than adding pressure to Saskatchewan’s existing electricity system or displacing export capacity. Indigenous participation is encouraged, and environmental-review requirements still apply.

Why it matters: this moves the Canadian data-centre debate beyond attracting investment. Saskatchewan is trying to decide in advance who should own these projects, how they should be powered and what benefits should stay in Canada.

Thirty applications do not mean 30 data centres will be built. But a pipeline that large explains why provinces are starting to establish rules before projects arrive.

One-quarter of Canadian businesses now say AI is in their plans

Statistics Canada released fresh national business data Monday.

Some 25.2% of Canadian employer businesses say they plan to use AI to produce goods or deliver services over the next 12 months, up from 14.5% a year ago and 10.6% two years ago.

But the other half of the story matters just as much: 52.7% still have no plans to use AI.

Among that group, 79.1% said AI was not relevant to the goods or services they provide. Privacy and security concerns and a lack of knowledge about AI capabilities were also cited.

Among businesses planning to use AI, data analytics and large language models were the two most common planned applications.

Why it matters: more businesses are considering AI, but this survey measures plans, not actual use. The national picture is clearer: interest is widening, while more than half still say they have no plans to use AI.

Quebec’s election campaign now has an AI code

Four parties represented in Quebec’s National Assembly — the CAQ, Quebec Liberals, Parti Québécois and Québec solidaire — endorsed a voluntary code for responsible AI use in politics on August 24.

Developed by IVADO and CEIMIA, the code covers transparency, responsible use, security and governance, and awareness and prevention.

Among other things, it addresses AI-generated political messages, deceptive deepfakes, false social-media accounts, political targeting and the use of foreign AI providers or data brokers.

Quebec’s provincial election was formally called August 27. Voting day is October 5.

Why it matters: this is not a law, and it is not an Elections Québec regulation. It is a voluntary commitment by the participating parties.

But the campaign is now underway, and generative AI can make convincing political images, audio and text cheaply and quickly. The code gives voters something concrete against which to judge how participating parties actually behave over the next five weeks.

When people talk about Canada needing more AI compute, they are talking about the physical capacity required to run and train AI systems.

That starts with specialized computer chips — usually GPUs for today’s large AI workloads — but the chips are only part of it.

They need servers.

Servers need buildings.

Buildings need high-capacity electricity connections.

The equipment produces heat, so data centres need cooling systems. IREN’s 2027 B.C. plans include new liquid-cooled deployments.

Then all of it needs networking equipment and high-speed connections to move enormous amounts of data.

So when a government talks about building Canada’s AI capacity, it is not only talking about algorithms.

It is also talking about electricity, land, construction, cooling, chips and capital.

That is why AI policy is increasingly becoming infrastructure policy.

The next time you see an AI or data-centre announcement, take the biggest number in the headline and label it.

Is it current capacity?

Planned capacity?

Committed financing?

Projected jobs?

Or estimated economic impact?

Then look for the words around it: plans, targets, estimates, expects, could.

Those qualifiers tell you whether the number describes something that exists now or something that may happen later.

The habit takes less than a minute, and it makes big AI-infrastructure announcements much easier to read accurately.

A calm, practical briefing on the AI developments that matter in Canada—delivered every Tuesday.

That’s the week. See you next Tuesday.

— Padge T.

AI Brief Canada is an editorially independent publication published by Groundshift Advisory Inc. and written by Padge T.

Groundshift Advisory helps Canadian organizations adopt AI responsibly. Sponsored, affiliated, or commercially connected content will always be clearly disclosed.

Issue disclosure: Nothing in this issue is sponsored, affiliated, paid, or commercially connected.

AI use: AI tools supported research, source discovery, outlining and drafting for this issue. All material was reviewed, edited and fact-checked by Padge T. before publication. Read our AI Use Disclosure.

Not affiliated with the Government of Canada.

Corrections or feedback: We welcome corrections and reader feedback. Email [email protected] or review our Editorial Policy.