Canada’s artificial intelligence industry is facing an uncomfortable contradiction. Billions of dollars are flowing into new data centres across the country, yet some homegrown technology companies are struggling to secure the computing power they need to compete. Toronto-based Transformer Lab has become a striking example, with CEO Ali Asaria revealing that discussions with more than 50 computing suppliers in recent weeks failed to produce the additional capacity his company needed.
The shortage comes as American technology giants accelerate their Canadian expansion, including Meta’s $13-billion Alberta data centre project and Anthropic’s planned computing operations in Nova Scotia. Meanwhile, Ottawa is proposing new financial measures to help Canadian businesses secure access to powerful AI infrastructure. The growing tension raises a fundamental question about Canada’s technological future: Will the country’s expanding computing industry primarily serve Canadian innovators, or the American corporations with the deepest pockets?
More Than 50 Suppliers, but No Reliable Extra Computing Capacity
For Transformer Lab, the scramble for computing power has become a frustrating exercise in persistence. Asaria described spending weeks pursuing suppliers without securing the additional capacity his company needed. In a November 4 social media post, he recounted assigning an employee to contact roughly 50 vendors, attending supplier dinners where new equipment remained unavailable, and receiving a time-sensitive offer involving just two Nvidia B300 GPUs located in India. At one conference, he said, challenging a supplier publicly finally led to a conversation about its actual inventory.
The experience is especially significant because Transformer Lab develops technology designed to automate scientific and artificial intelligence research. Its Primus platform, introduced earlier in 2026, can conduct experiments, analyze results and generate research reports, according to the company. That makes reliable computing access essential to its business. Asaria’s account does not prove that every Canadian supplier lacks available equipment, but it illustrates how difficult securing suitable computing resources can become. For smaller technology companies, the problem is not necessarily a shortage of ideas or talent. It can be the inability to obtain the machines required to put those ideas into practice.
Why Canada’s AI Computing Shortage Is More Complicated Than It Looks
Modern artificial intelligence systems require more than conventional computer servers. Advanced workloads depend on specialized graphics processing units, or GPUs, supported by high-speed networking, substantial electricity supplies and sophisticated cooling systems. These resources must work together efficiently, particularly when companies train large AI models or conduct complex experiments. A new data centre announcement therefore does not automatically mean that advanced processing capacity is installed, operational or available for smaller businesses to rent.
Canada’s federal government has already acknowledged the seriousness of these limitations. Innovation, Science and Economic Development Canada has described the country as ranking at the bottom of the G7 for publicly available computing capacity. Earlier federal consultations found that 45% of surveyed Canadian AI firms and developers identified cloud computing costs as a barrier, while 43% of research respondents identified availability as a concern. These findings illustrate why the shortage extends beyond individual businesses. When computing resources become expensive or difficult to obtain, promising research projects can be delayed, reduced in scope or redirected toward what existing equipment can handle. Building additional facilities helps, but ensuring those facilities contain suitable, accessible technology is equally important.
American Technology Giants Are Moving Into Canada’s Data-Centre Market
The competition for Canadian computing infrastructure is intensifying as some of the world’s largest technology companies make substantial investment commitments. Meta broke ground on a one-gigawatt data centre in Sturgeon County, Alberta, in July 2026. The project is expected to represent more than $13 billion in investment when completed and is intended to support Meta’s own AI workloads and digital services. In Nova Scotia, Anthropic, the American company behind Claude, has signed a long-term lease for capacity at EverWind’s planned Point Tupper data-centre campus. The development’s initial 95-megawatt phase is targeted for late 2027.
These investments could bring substantial economic benefits to their host communities, including construction employment, infrastructure upgrades and additional business activity. Meta anticipates approximately 3,000 construction workers at the peak of its Alberta project and more than 300 operational jobs once completed. However, the larger concern for Canadian AI developers is who ultimately controls the computing capacity being created. A data centre located in Canada does not necessarily provide Canadian businesses with affordable access to its processors. If powerful American corporations secure most of the capacity through long-term agreements, domestic startups could find themselves competing for whatever computing resources remain available.
Why American Tech Giants Have an Advantage Before Construction Even Begins
The competition is not determined solely by which company can afford the most advanced computer chips. Financing plays an equally important role. Simon Ahdoot, chief executive of Montreal-based Hypertec Group, has explained that borrowing costs for data-centre developers depend heavily on the financial strength of their customers. A supplier with a major American technology company committed to purchasing computing capacity for years can present lenders with a relatively predictable revenue stream. That makes expensive infrastructure projects easier to finance than facilities depending on smaller, less-established customers.
Sean Mullin, a former federal policy adviser involved in designing Canada’s $2-billion AI computing strategy, has highlighted the difficulty faced by developers serving domestic startups. These businesses may need to combine numerous smaller customer contracts to demonstrate sufficient demand to lenders. Even then, financing can be more expensive or difficult to obtain. The result is a reinforcing disadvantage: established technology giants secure access through large commitments, while smaller Canadian companies struggle to attract the infrastructure investment they require. For a startup hoping to expand its research operations, the obstacle may therefore begin long before negotiations over individual GPUs. It can originate with the financial decisions that determine which data centres are built in the first place.
Ottawa Wants to Unlock 500 Megawatts of Computing Power for Canadian Companies
The federal government is attempting to address the imbalance through a new initiative outlined in Budget 2026, tabled on November 5. Ottawa says it intends to enable Canadian AI companies to access 500 megawatts of sovereign computing capacity by using financial mechanisms such as backstops and offtake agreements. Rather than relying exclusively on traditional grants, these arrangements are intended to encourage private investment. An offtake agreement can provide a developer with a commitment to purchase future services, while a financial backstop can reduce certain risks that discourage lenders from funding expensive projects.
The approach could help Canadian computing suppliers attract the financing required to purchase equipment and expand their operations. However, the 500-megawatt figure represents a policy objective, not computing capacity that has already become available. Details about individual agreements, implementation timelines and customer access will determine how effective the initiative becomes. Asaria has expressed support for the federal approach while warning that developers could potentially exploit financing guarantees. That concern highlights the importance of ensuring public financial support produces measurable benefits for domestic businesses rather than simply accelerating speculative construction. Ottawa must find a way to encourage new investment without allowing the companies most capable of securing government-backed financing to monopolize the resulting capacity.
Canada’s Existing $2-Billion AI Strategy Has Not Eliminated the Access Problem
The latest proposal builds on an earlier federal commitment to strengthen Canada’s computing infrastructure. Budget 2024 introduced $2 billion over five years for sovereign AI computing initiatives. The strategy included up to $700 million to support commercial data-centre development, up to $1 billion for public supercomputing infrastructure and $300 million for an AI Compute Access Fund. In May 2026, the government announced $66 million in support for 44 projects through that fund. The recipients work in industries ranging from health care and agriculture to advanced manufacturing, transportation and financial services.
These measures demonstrate that Canada has begun addressing the shortage, but financial announcements and functioning computing systems are not interchangeable. Federal support can help a business afford processing power, yet that assistance cannot immediately produce GPUs that are unavailable. Some infrastructure projects require lengthy construction, equipment procurement and electrical upgrades before customers can begin using them. Ottawa’s June 2026 national AI strategy also proposed an additional $700 million for affordable sovereign computing access. The challenge now is translating these commitments into reliable services. For a company developing a medical diagnostic tool or an AI research platform, the meaningful measure of progress is not the total value of government announcements. It is whether enough appropriate computing capacity can actually be obtained when the work needs to happen.
Canadian Telecom Companies Are Expanding Their Own Computing Infrastructure
Canada is not relying entirely on American technology companies to develop its AI infrastructure. TELUS has been expanding its sovereign computing operations after reporting that its AI factory in Rimouski, Quebec, which opened in September 2025, had sold out its available capacity. In May 2026, the company outlined plans for three British Columbia facilities that could collectively scale to more than 60,000 GPUs and 150 megawatts by 2032. Bell Canada is pursuing an even larger project in Saskatchewan. A September agreement with the provincial government outlined a potential 1.2-gigawatt AI infrastructure hub with capital investment that could reach $52.5 billion.
Canadian technology partnerships are also forming around these developments. Bell, Cohere, Hypertec and BUZZ High Performance Computing announced a collaboration in June 2026 intended to strengthen Canadian-controlled infrastructure for advanced AI workloads. Such partnerships could help domestic companies obtain services that would otherwise depend heavily on foreign operators. However, the scale of announced investments should not obscure their development status. Bell’s Saskatchewan expansion involves a non-binding agreement and phased construction subject to further conditions, while TELUS’s longer-term expansion targets extend several years into the future. These projects could materially improve Canada’s position, but startups facing immediate shortages cannot depend on capacity that has yet to enter commercial operation.
Canadian AI Sovereignty Means More Than Keeping Servers Inside the Country
Canada’s growing emphasis on sovereign artificial intelligence involves more than the physical location of computing equipment. It also raises questions about ownership, operational control and who ultimately determines how infrastructure is used. During an April 2026 parliamentary committee hearing, Cohere chief AI officer Joëlle Pineau explained why the Canadian company had worked with American-owned CoreWeave to establish advanced computing infrastructure. According to Pineau, no Canadian-owned provider had the necessary capabilities when Cohere evaluated potential partners for its specialized research cluster.
That experience illustrates the difficult choices facing Canadian AI companies. Using foreign infrastructure partners may allow domestic businesses to develop advanced technology sooner, but it can complicate efforts to establish independent computing capabilities. Canada’s sovereign computing program guidelines emphasize that infrastructure should remain under Canadian governance, with domestic control over essential systems and decisions about access. These safeguards become particularly important when computing resources support sensitive research, government services or commercially valuable intellectual property. A Canadian company may develop sophisticated AI models, but its long-term independence can still depend on outside organizations controlling the infrastructure beneath them. Strengthening sovereignty therefore requires both expanding domestic computing capacity and ensuring Canadian institutions retain meaningful decision-making authority over how that capacity is used.
Communities Are Questioning the Cost of Canada’s Data-Centre Expansion
The rapid development of AI infrastructure is generating concerns beyond the technology industry. In Nova Scotia, more than 200 residents attended a November 3 community meeting about EverWind’s proposed data centre near Point Tupper. Residents questioned the project’s environmental protections, electricity requirements and potential effects on local water resources. Some also challenged whether adequate consultation and oversight had taken place. Those concerns are significant because the facility’s proposed expansion would eventually bring its electricity demand to 420 megawatts, although the initial development is substantially smaller.
Large data centres can create employment and investment opportunities, but their energy requirements introduce difficult decisions for governments and local communities. Meta’s Alberta project, for example, is associated with plans for a dedicated natural-gas-fired electricity facility to help meet its substantial power requirements. In September 2026, Ottawa introduced Responsible Data Centre Development Principles emphasizing environmental protection, transparency, community benefits and preventing additional electricity costs from being transferred to existing ratepayers. These expectations reflect the need to evaluate infrastructure growth from more than one perspective. Canadians may welcome technology investment while still expecting clear answers about electricity consumption, water protection and local economic benefits. Ultimately, the success of the industry’s expansion will depend partly on whether developers can demonstrate that their projects create lasting value without imposing unreasonable costs on nearby communities.
Canada’s Biggest Challenge May Be Building the Right Amount of Capacity
Although many Canadian AI businesses want additional computing power, not everyone agrees that rapidly constructing more data centres is the best solution. Craig McLellan, chief executive of Toronto-based cloud services company ThinkOn, has questioned whether domestic demand currently supports the 500 megawatts of capacity Ottawa wants to enable. He has warned that financial guarantees could encourage excessive construction and reduce incentives to prioritize efficient computing equipment. His concern is not that Canada should abandon its ambitions, but that infrastructure development needs to reflect realistic demand rather than speculative expectations.
The debate underscores a delicate balancing act. Build too little capacity, and Canadian researchers may continue losing opportunities to better-funded international competitors. Build too much without sufficient customers, and developers could end up operating expensive facilities that struggle to generate adequate returns. An effective approach would measure success through operational computing capacity, affordability, utilization and meaningful access for Canadian businesses rather than investment announcements alone. Foreign customers can still contribute to a successful Canadian industry, provided domestic innovators are not consistently pushed aside. For companies like Transformer Lab, the immediate priority remains straightforward: obtaining reliable access to the processing power needed to develop competitive technology. Whether Canada can deliver that access will help determine how much of the AI economy’s future value remains in Canadian hands.
