The announcement was not a policy intention or a budget line item. It was a concrete project: three world-class data centre facilities in British Columbia, co-developed by TELUS and real estate partner Westbank, backed by an MOU with the federal government, powered by 85 megawatts of clean energy from BC Hydro, and designed to scale to over 60,000 high-performance NVIDIA GPUs and 150 megawatts of total capacity by 2032. Total projected economic value: $9 billion injected into the Canadian economy.
It is the largest sovereign AI infrastructure commitment in Canadian history. And it arrives in a global context where the question of who controls AI compute — and therefore who controls AI capability, data, and intellectual property — has become one of the defining geopolitical questions of the decade.
What Is Being Built: The Three-Site B.C. Cluster
The announcement centres on a three-facility cluster in British Columbia, with staggered timelines designed to match infrastructure delivery to demonstrated market demand.
Kamloops AI Factory (expanding existing facility): The first site to come online — expected later in 2026. TELUS is expanding its existing Kamloops data centre into a full Sovereign AI Factory format. Kamloops's geography provides advantages relevant to AI infrastructure: cooler ambient temperatures reducing cooling costs, proximity to BC Hydro's renewable power grid, and physical distance from the seismic risk zones concentrated along the Pacific Coast.
M3 Facility, Mount Pleasant, Vancouver: Planned to open at the end of 2026 and scale through 2028. Located in Vancouver's Mount Pleasant neighbourhood, the M3 facility is a new-build co-developed with Westbank — the real estate developer whose partnership brings deep expertise in urban mixed-use development and district energy systems. Mount Pleasant's position within Vancouver's emerging tech corridor makes it accessible to the city's startup ecosystem, universities, and provincial government institutions.
150 West Georgia, Vancouver: Designed to come online in 2029, this is the flagship facility of the cluster — a central downtown Vancouver location at 150 West Georgia Street that places sovereign AI compute at the geographic and economic heart of Canada's most internationally connected city.
At full scale, the three-site cluster will house infrastructure featuring high-performance NVIDIA GPUs capable of supporting large-scale AI model training, complex simulations, and production-scale deployment — the complete AI development lifecycle in a single sovereign platform. The cluster's total capacity grows to over 150 MW by 2032, with the initial phase drawing on 85 MW of clean, renewable power secured from BC Hydro.
The project is developing under the federal Enabling Large-Scale Sovereign AI Data Centres initiative — a programme created by Budget 2025 to identify a limited number of large-scale sovereign commercial AI data centre projects and negotiate memoranda of understanding with the most promising proposals. The call for proposals ran from January 15 to February 15, 2026. TELUS's B.C. cluster is the project that emerged from that process as the federal government's priority partnership.
The Origin Story: How Rimouski Proved the Demand Was Real
The B.C. cluster was not built from a policy brief. It was built from a sold-out product.
Canada's first Sovereign AI Factory was launched by TELUS in Rimouski, Quebec — a smaller market in the Lower St. Lawrence region that most technology investors would not have considered a natural location for a flagship AI infrastructure investment. TELUS chose Rimouski partly for its access to Quebec's abundant, low-cost hydroelectric power, and partly as a deliberate test of whether demand for sovereign AI compute would materialise from Canadian institutions and enterprises.
The result was unambiguous. The Rimouski facility sold out completely. The unprecedented demand that completely sold out Canada's first AI Factory in Rimouski proves that Canadian innovators want cutting-edge AI infrastructure built right here on Canadian soil.
Darren Entwistle, President and CEO of TELUS, described the B.C. expansion as being developed as a direct response to that market demand — not as a speculative infrastructure build but as a capacity expansion driven by a waiting list of Canadian customers who wanted sovereign compute access and could not get it from the existing facility.
This demand-driven origin story matters for evaluating the project's prospects. AI infrastructure investments fail when they are built ahead of demand and struggle to fill capacity. The Rimouski precedent suggests the opposite dynamic is operating in Canada: the sovereign AI compute market is currently supply-constrained, not demand-constrained. Canadian businesses, universities, government agencies, and startups want access to AI infrastructure that does not require their data to leave the country — and they are willing to commit to it when it is made available.
Why "Sovereignty" Is Not Just a Political Slogan
The word "sovereignty" appears throughout the project's official communications — in the programme name, the facility branding, the ministerial statements, and the CEO remarks. It is worth being precise about what AI sovereignty actually means in 2026, because it is a concept that has genuine technical and legal substance beneath the political framing.
Data Sovereignty: The Legal Dimension
When a Canadian university, hospital, or government agency uses AI infrastructure operated by a U.S. hyperscaler — AWS, Azure, Google Cloud — the data processed on that infrastructure is subject to the legal jurisdiction in which the infrastructure operates. Under the U.S. Cloud Act, American law enforcement agencies can compel U.S. cloud providers to produce data stored in their systems, regardless of where that data physically resides. A Canadian hospital running patient data through a U.S.-operated AI system is exposing that data to a legal access regime that Canadian privacy law does not fully control.
Sovereign AI infrastructure — owned, operated, and governed within Canada — keeps that data under Canadian legal jurisdiction from ingestion to output. For healthcare organisations bound by provincial health privacy legislation, for financial institutions subject to OSFI's data residency requirements, and for government agencies handling sensitive citizen information, this is not a marginal compliance consideration. It is a foundational operational requirement.
TELUS's Sovereign AI Factory supports the full AI development lifecycle — from model training and fine-tuning to real-world deployment to inference at large scale within a single platform — all on infrastructure owned, operated, and governed entirely within Canada. The emphasis on ownership and governance, not merely physical location, is the legally meaningful distinction.
Intellectual Property Sovereignty
When an organisation trains a proprietary AI model on hyperscaler infrastructure, the training process exposes model architecture, training data, and proprietary outputs to the infrastructure operator's terms of service, logging practices, and data use policies. The competitive intelligence embedded in a model trained on an organisation's proprietary data — its customer patterns, its product demand forecasts, its scientific research — is an asset that Canadian organisations have an interest in protecting from exposure to foreign commercial actors.
Sovereign infrastructure provides a different assurance framework: the infrastructure is Canadian, the governance is Canadian, and the terms under which data and model outputs are handled are subject to Canadian legal oversight rather than a U.S. technology company's terms of service.
TELUS CEO Entwistle directly framed the competitive dimension: Canadian businesses, entrepreneurs, startups, researchers, public institutions and government organizations require world-class AI compute without sending their data, intellectual property and competitive advantage outside Canadian borders.
Compute Sovereignty: The Strategic Dimension
The NDP leader Avi Lewis raised a pointed counterargument: data sovereignty doesn't really exist in an AI ecosystem dominated by U.S. technology companies. It's a marketing gimmick.
This objection is partially correct and partially misconceived. It is correct that using NVIDIA GPUs and operating a facility with partial ties to U.S. technology ecosystems does not produce absolute technological independence. The AI hardware supply chain remains heavily concentrated in U.S.-aligned companies, and the dominant AI frameworks and models originate from U.S. research institutions.
But the objection conflates complete independence with meaningful sovereignty, which is a different standard. Canada does not need to design its own AI chips to benefit meaningfully from AI infrastructure that keeps data processing within Canadian legal jurisdiction, employs Canadian workers, contributes Canadian tax revenue, and reduces the country's economic dependence on foreign AI infrastructure operators who charge premium prices for access to compute that Canadian institutions cannot influence or audit.
The question is not whether Canada can achieve total technological autarky. No country can. The question is whether Canada can establish enough domestic AI infrastructure to ensure its most sensitive data, its most important research, and its most strategically valuable AI development does not depend entirely on foreign infrastructure operators over whom it has no legal, economic, or regulatory leverage.
The $9 billion B.C. cluster is a meaningful step toward that standard. It is not a complete solution. It is not nothing.
The Sustainability Architecture: Using Every Electron Twice
One of the most technically interesting aspects of the TELUS B.C. cluster is its sustainability architecture — specifically the district energy integration that transforms waste heat from AI compute into residential heating for Metro Vancouver.
Standard AI data centres generate enormous quantities of heat as a byproduct of compute operations. Most manage this heat as a disposal problem: cooling systems pump heat outside the building and into the atmosphere. It is a thermodynamic inefficiency that the AI industry has largely accepted as unavoidable.
The TELUS B.C. cluster treats it differently. The waste heat generated by AI compute at the M3 and 150 West Georgia Vancouver facilities will be captured by Creative Energy — Vancouver's district heating network — and distributed into the residential heating grid for Metro Vancouver. At full scale, this waste heat recovery system is projected to heat more than 150,000 homes in Metro Vancouver, lowering energy costs for British Columbians and eliminating the facilities' net carbon footprint.
Westbank founder and CEO Ian Gillespie described the design philosophy: after years of research and development working with TELUS, BC Hydro, Creative Energy, the City of Vancouver, the Province of B.C., and the federal government, the project has arrived at what he called the most elegant low-carbon solution — effectively using every electron twice, to produce environmentally responsible sovereign AI infrastructure.
The specific performance claims: the cluster will run on 98% clean energy, using liquid cooling throughout, with waste heat recovery integrated into district energy distribution. The combination of BC Hydro's renewable hydroelectric power and the heat recovery system positions TELUS's claim to be building the world's most sustainable sovereign data centres as technically substantive rather than purely aspirational.
For comparison: a standard AI data centre running on grid-average energy consumes enormous quantities of electricity and water while producing waste heat that contributes to urban heat island effects. The TELUS design inverts that relationship — the data centre becomes a net energy contributor to the surrounding residential community rather than purely an energy drain.
The Political Dimension: How This Fits Canada's Broader AI Strategy
The B.C. cluster announcement did not emerge from a vacuum. It is the most visible output of a policy framework that the Carney government has been assembling since Budget 2025.
The specific policy mechanism — the Enabling Large-Scale Sovereign AI Data Centres initiative — was created in Budget 2025 to deploy government support for a limited number of commercially viable, large-scale AI data centre projects on Canadian soil. The call for proposals that ran January 15 to February 15, 2026 produced multiple applications, and the government is engaging with the most promising projects. The TELUS B.C. cluster is the first to reach the MOU stage — the formal commitment to detailed collaboration on project structure, financing, and governance — but it is explicitly described as one of a limited number of projects the government is advancing, not the sole recipient.
Minister of Artificial Intelligence and Digital Innovation Evan Solomon framed the national policy position clearly: securing Canada's technological independence is a national priority, and it requires building the infrastructure to back it up. The B.C. cluster represents concrete action to strengthen Canada's sovereign AI capacity and ensure that Canadian innovation, data, and economic advantages are anchored in Canada.
The political context in which this statement lands is significant. Canada in 2026 is navigating a complex relationship with the United States under trade tensions that have made Canadian political and business leadership more attentive than at any point in recent decades to the risks of deep economic dependence on a single foreign partner. The sovereign AI infrastructure initiative fits within a broader pattern of Canadian policy moves toward building economic resilience against external disruption — including the AI compute infrastructure on which the country's innovation economy increasingly depends.
The NDP's critique — that public investment in AI infrastructure displaces more immediate social needs — reflects a genuine values tension that democratic societies face when allocating capital between long-term strategic infrastructure and immediate social services. That tension does not invalidate the strategic rationale for sovereign AI infrastructure, but it represents a real political constraint on how far and how fast the programme can move.
The Global Sovereign AI Race: Where Canada Sits
Canada's $9 billion B.C. cluster is a significant national commitment. In the global context of sovereign AI infrastructure investment, it is one of many moves on a board that is filling up rapidly.
The European Union's AI Continent Action Plan explicitly targets sovereign AI compute as a strategic priority, with investments across France, Germany, Spain, and Scandinavia aimed at reducing European dependence on U.S. hyperscalers. Saudi Arabia's HUMAIN initiative — backed by a $600 billion investment framework announced during the Trump administration's Gulf visits — is building sovereign AI infrastructure at a scale that dwarfs any single national programme outside the United States. India's IndiaAI Mission is investing $1.3 billion in domestic AI compute as part of a broader strategy to develop indigenous AI capability.
The pattern is global and accelerating: nations that can afford to build sovereign AI infrastructure are doing so, driven by a shared recognition that dependence on foreign AI compute creates strategic vulnerabilities in data security, economic competitiveness, and technological autonomy that they are not willing to accept indefinitely.
Canada's position in this race is shaped by specific advantages that not all nations share. Its abundant renewable hydroelectric power — particularly in British Columbia and Quebec — provides a clean energy foundation for AI infrastructure that coal-dependent economies cannot match. Its strong research universities — Toronto, Waterloo, UBC, McGill — provide a talent pipeline that supports both AI infrastructure operations and the AI research that generates demand for sovereign compute. And its proximity and economic integration with the United States creates both a market relationship that facilitates access to U.S. AI hardware and a political motivation to ensure that relationship does not become one of unconstrained technological dependency.
Who Will Use This Infrastructure?
TELUS's characterisation of the target market is specific and worth examining. The Sovereign AI Factory is designed to serve Canadian businesses, entrepreneurs, startups, researchers, public institutions, and government organisations that require world-class AI compute without sending their data, intellectual property, and competitive advantage outside Canadian borders.
The segments that are most likely to generate early, sustained demand:
Healthcare and life sciences. Canada's provincial health systems handle some of the most privacy-sensitive data in the economy. AI applications in medical imaging, diagnostic support, drug discovery, and population health management require AI compute that can be demonstrated to comply with provincial health privacy legislation — a requirement that sovereign infrastructure satisfies more cleanly than U.S.-operated alternatives.
Financial services. Canada's major banks and insurance companies are subject to OSFI data residency requirements that create compliance considerations for AI workloads processing sensitive customer financial data. Sovereign compute provides a cleaner compliance pathway than attempting to structure U.S. cloud agreements to satisfy Canadian regulatory requirements.
Government and public sector. Federal agencies, provincial governments, and municipalities running AI applications on citizen data have both legal obligations and political incentives to process that data within Canadian jurisdiction. The federal government's own role as a driver of the Enabling Large-Scale Sovereign AI initiative reflects its own institutional demand for compute that aligns with its data governance obligations.
Research institutions. Canadian universities conducting sensitive research — in defence, energy, health, and other strategic domains — have interests in ensuring their research data and proprietary methodologies are not exposed to foreign legal access regimes through the compute infrastructure they use.
Technology startups. The Rimouski sellout demonstrated demand from the startup ecosystem specifically. Canadian AI startups building proprietary models on Canadian data have commercial reasons to prefer sovereign infrastructure — it simplifies investor due diligence, reduces compliance complexity, and positions the company's data governance practices favourably with enterprise customers.
The Honest Caveats
The B.C. cluster announcement comes with a set of honest caveats that balanced coverage requires acknowledging.
The MOU is not a completed project. The government and TELUS are advancing work and exploring mechanisms — language that describes a commitment to collaborate on project details, not a finalised agreement with committed funding and binding delivery dates. The gap between an MOU and a functioning data centre involves planning approvals, construction procurement, equipment orders, and financing arrangements that can each introduce delays.
60,000 GPUs by 2032 is a long horizon. The comparison point matters: Microsoft's Fairwater superfactories are being built to hundreds of thousands of NVIDIA Vera Rubin chips. Saudi Arabia's HUMAIN is targeting investments that dwarf the B.C. cluster. Canada's sovereign AI cluster is a meaningful national commitment, but it is a fraction of the compute being assembled by the world's largest AI infrastructure investors.
The NDP's sovereignty critique has substance. Avi Lewis's observation that data sovereignty is limited in an ecosystem dominated by U.S. technology companies is not entirely wrong. The Sovereign AI Factory cluster uses NVIDIA GPUs — American hardware — and operates software stacks with significant U.S.-origin components. The sovereignty is real and meaningful at the data governance and legal jurisdiction level. It is partial rather than complete at the hardware and software dependency level. Honest communication about what sovereign infrastructure does and does not provide matters for setting appropriate expectations.
Environmental claims require ongoing verification. The claim of 98% clean energy and net-zero carbon footprint through heat recovery is technically achievable given BC Hydro's renewable power base and the district energy integration. Whether those claims are verified by independent third parties on an ongoing basis — rather than as design specifications — is a transparency question that the project's governance framework should address.
Conclusion: Infrastructure Is Policy
The $9 billion Canada-TELUS Sovereign AI Factory cluster in British Columbia is not primarily a technology story. It is a policy story about what a country with Canada's specific assets — abundant clean energy, strong research universities, proximity to the U.S. AI ecosystem, and a tradition of resource-based infrastructure investment — chooses to do with those assets in an era when AI compute has become as strategically significant as energy infrastructure or transportation networks.
The federal government's framing is deliberate and correct: securing Canada's technological independence is a national priority, and it requires building the infrastructure to back it up. Infrastructure that keeps Canadian data under Canadian legal jurisdiction, employs Canadian workers, generates Canadian tax revenue, and provides Canadian researchers, startups, and institutions with world-class AI compute without foreign dependency is infrastructure worth building.
The project's success will be measured not by whether the facilities are built on schedule — though that matters — but by whether they become the operational foundation for a Canadian AI innovation ecosystem that would not have been possible without them. Whether a genomics startup in Vancouver, a healthcare system in Alberta, or a federal agency in Ottawa uses the B.C. cluster to build something that sovereign compute made possible — something that would have been constrained or prevented by the data governance requirements that foreign cloud compute cannot satisfy.
That is the bet the $9 billion represents. It is a bet that Canadian innovation, anchored in Canadian infrastructure, is worth building.
Quick Reference: Canada's Sovereign AI Factory Cluster
| Detail | Information |
|---|---|
| Lead proponent | TELUS Communications |
| Government partner | Government of Canada (federal) |
| Real estate partner | Westbank |
| Energy partner | BC Hydro |
| Programme | Enabling Large-Scale Sovereign AI Data Centres (Budget 2025) |
| Minister | Evan Solomon — Minister of AI and Digital Innovation |
| Total economic impact | $9 billion projected |
| Site 1 | Kamloops AI Factory — online later 2026 |
| Site 2 | M3 Facility, Mount Pleasant, Vancouver — end of 2026, scales through 2028 |
| Site 3 | 150 West Georgia, Vancouver — 2029 |
| Full-scale capacity | 60,000+ NVIDIA GPUs / 150 MW by 2032 |
| Initial power | 85 MW clean renewable (BC Hydro) |
| Energy source | 98% clean energy |
| Sustainability | Liquid cooling + heat recovery — heats 150,000+ Metro Vancouver homes |
| Predecessor facility | Rimouski, Quebec AI Factory — completely sold out |
| Jobs created | 1,000+ construction + hundreds of high-skilled operations roles |
| GPU hardware | High-performance NVIDIA GPUs |
| Status | MOU stage — details being finalised between TELUS and federal government |
| Key opposition | NDP leader Avi Lewis — questioned sovereignty claims and investment priorities |
0 Comments