The movement to develop sovereign AI, which is the development of foundational AI models entirely based in a country’s own data centers and based on a country’s own large language models, continues to gather steam. TELUS this past week announced its launch of Canada’s first fully sovereign AI Factory in Rimouski, Quebec
This effort, developed in joint with NVIDIA and HPE, is a functioning AI center tied together by TELUS’ fiber backbone, and is intended to give Canadian enterprises the ability to develop, train, and operate AI models without crossing a border.
Darren Entwistle, TELUS CEO, framed the launch by saying, “with the launch of our nation’s first fully Sovereign AI Factory in Rimouski, we are maximizing Canadian autonomy over sensitive data.”
That message echoes what regulators and policymakers are saying around the world about AI sovereignty. Everyone can see that AI is the new battlefield for data and platform dominance, and this time around, countries are not waiting passively to see who emerges as the victor.
Inside the TELUS Facility
The Sovereign AI Factory operates entirely under Canadian operation. Every component, from chips to cooling, is managed locally. The site runs on renewable energy and has been designed for efficiency, with cooling methods that cut water use sharply and reduce reliance on mechanical systems for most of the year.
Using the AI factory, enterprises can carry out the full AI lifecycle from using LLM trained and hosted entirely on Canadian infrastructure to new model training, fine‑tuning, and deployment in Canadian data centers. To accelerate adoption, TELUS has lined up partners such as NVIDIA, Accenture, League, and OpenText. The company has also committed to building a second center in Kamloops, British Columbia.
For years, Canadian companies in sectors like healthcare and finance have faced barriers tied to data residency and compliance. When workloads leave the country for global cloud providers, sovereignty concerns follow. TELUS’ approach removes that obstacle. Enterprises and Canadian governmental organizations can get the benefits of AI along with national control without trading offer performance or compute power.
TELUS is betting that Canadian firms will pay for trust, compliance, and local oversight, even if alternate options in the market might be cheaper, more rapidly developed, and have broader global adoption. From the company’s perspective, Rimouski is a first test. If enterprises adopt this first factory approach, more sites will follow.
Where This Fits Globally
“Sovereign AI” is emerging as a phrase with growing weight. At its core the idea is that critical AI compute, data, and governance must live under the control of a nation’s own jurisdictions and under their local legal, regulatory, and operational guardrails.
Recent news coverage shows how countries from Sweden to Korea, United Kingdom to UAE, and even local and tribal governments like the Cherokee Nation are seeing the critical need to maintain control over a technology that is increasingly powering a large number of digital systems.
And momentum continues. France is backing a homegrown stack that combines national cloud capacity with foundation models from Mistral AI to serve banks, health systems, and the public sector without relying on U.S. dominated infrastructures. In the EU, EuroHPC sites such as LUMI, Leonardo, and MareNostrum 5 give the group shared data center capabilities under European rules.
In the UAE, Abu Dhabi’s Technology Innovation Institute keeps enhancing and innovating Falcon family models and is building out local inference and training routes. The region is directing large capital pools toward data centers, chips, and model builders. In Saudi Arabia, national funds and research hubs are positioning AI as core economic infrastructure, with domestic residency and control as selling points.
In India, The IndiaAI mission concentrates GPU capacity and public‑private funding to give startups and agencies domestic data center capabilities.
Of course, in China, Beijing is pressing domestic options across the stack, from models like ERNIE, DeepSeek, and Qwen to accelerators in the Ascend line. The aim is to train and run state‑scale AI without foreign chips or cloud computing.
The TELUS move fits into this accelerating trend. It signals Canada’s intention not merely to consume AI but to anchor it, controlling the “plumbing” of AI rather than leasing it.
Read the full article here
