Canada’s biggest telecom is making a very loud statement about where the economy is heading: AI isn’t just software anymore — it’s infrastructure.
BCE says it will invest an additional $1.7 billion to build a 300-megawatt, purpose-built AI data center in Saskatchewan, with Cerebras and CoreWeave signed on as tenants. The project is being developed through BCE’s Bell Canada unit in collaboration with the Saskatchewan provincial government, and BCE is calling it the largest purpose-built AI data center development in Canada.
This isn’t a small “innovation hub.” It’s an industrial-scale compute plant.
What BCE is building (in plain terms)
A 300-MW AI data center is not designed for everyday web hosting. It’s designed for the kind of workloads that power the AI economy:
- large model training
- high-volume inference (serving AI at scale)
- dense GPU/accelerator clusters
- massive data throughput requirements
BCE expects to spend roughly $1.3 billion in capital expenditures in 2026 related to construction. The project will be funded through a mix of debt and cash on hand.
Construction is set to begin this spring, and the facility will come online in phases, with the first stage expected to be operational in the first half of 2027.
Why Cerebras and CoreWeave matter
The tenant mix explains what this site is intended to be: a serious AI compute hub, not a generic server farm.
- Cerebras will supply its specialized AI chips aimed at large-scale training and computing.
- CoreWeave will provide AI compute capacity using Nvidia processors.
That pairing signals a “best tools for the job” approach — combining different architectures and capacity models that are in high demand as enterprises and governments race to secure compute.
Why Saskatchewan?
At first glance, Saskatchewan might surprise people used to thinking “AI = Toronto, Montreal, Vancouver.” But the logic is straightforward:
- Large data centers need power availability, long-term planning, and room to scale.
- They benefit from regions that can support major new industrial loads.
- They increasingly depend on partnerships with local infrastructure players and governments.
The project is also being positioned as a regional economic engine — a hub that can attract additional investment and expand Canada’s AI ecosystem beyond the usual urban cores.
The SaskTel connection: fibre + go-to-market partnership
Connectivity is the second half of an AI data center story. BCE says the facility will connect into Bell’s fibre network through a partnership with SaskTel, and the two companies will act as go-to-market partners to offer AI-powered products and services to SaskTel customers.
That’s an important detail: BCE isn’t only building compute; it’s building a channel to sell AI-enabled services into real organizations — government, business, and public-sector clients across the region.
BCE is raising its AI revenue ambitions
Alongside the project, BCE raised its forecast for AI-powered solutions revenue to about $2 billion by 2028, up from about $1.5 billion previously.
That’s the strategic punchline: the data center isn’t just a construction project. It’s a growth engine aimed at turning Bell into a bigger AI services provider — not only a connectivity company.
What this signals for Canada
This move fits a larger pattern: AI leadership is increasingly determined by who controls compute, power, networks, and deployment capacity.
Canada has plenty of AI talent and research strength — but compute capacity has become the bottleneck. A project of this scale says: Canada wants to host more of the “AI factory” layer domestically, rather than exporting its ideas and importing the infrastructure.
Bottom line
BCE’s Saskatchewan build is a big bet on the next economic reality: AI is becoming a utility-grade industry, and the countries (and companies) that win won’t just have smarter models — they’ll have the infrastructure to run them at scale.
