Canada wants more of the compute behind artificial intelligence built here. This year the federal government moved from strategy to specific projects. For a company that needs GPU capacity now, it's worth understanding what's coming and when.
What's happened
- January 15 to February 15, 2026: Innovation, Science and Economic Development Canada (ISED) ran a call for proposals to enable large-scale sovereign AI data centres. According to Torys, the call targeted facilities above 100 megawatts, with promising projects considered for a memorandum of understanding with the government.
- May 11, 2026: the government announced it is advancing work with TELUS on a proposed large-scale AI data centre in British Columbia.
- No money yet. ISED states plainly that no funding has yet been committed or distributed. The memoranda are non-binding.
TELUS's chief executive said demand had "completely sold out" its first AI facility in Rimouski, Quebec. That tells you as much about the market as any policy document.
What it means for buyers
These are major, multi-year builds, and they're aimed first at large-scale national capacity. They'll help, but they won't solve a GPU shortage for a Canadian company in the next few quarters.
In the meantime, the realistic options for a business that needs GPU capacity are:
- Hyperscaler GPU instances. Fast to start, but the newest GPUs are often scarce, frequently tied to reservations or commitments, and subject to the same foreign-jurisdiction questions as any U.S.-owned cloud.
- Specialist GPU clouds. Often better availability and pricing for training jobs, though location and ownership vary widely.
- Dedicated private GPU capacity. Hardware reserved for you, billed monthly, in a location and under a jurisdiction you choose. Better suited to steady workloads such as inference, fine-tuning and data that can't leave Canada.
Questions to ask before you commit
- Where will the hardware physically sit, and who owns the provider? The sovereignty question applies to GPUs just as much as to storage.
- How long until it's actually available? Ask for a date, not a waitlist position.
- What's the minimum commitment? Some GPU offers only make sense on a multi-year term; know what you're signing up for.
- What will moving data cost? Training data in and model outputs out can be a large part of the bill.
- Is it shared or dedicated? Performance on shared hardware varies with the neighbours.
How we help
We help Canadian organizations find GPU capacity that fits the workload, the budget and the jurisdiction, often in days rather than quarters. See our AI and GPU capacity use case, or book a discovery call and tell us what you're trying to run.