Canada’s plans to bridge the AI compute gap and how it can make industry policy inclusive and sustainable

The new frontier in national competitiveness

As countries race to solidify their positions as leaders in artificial intelligence (AI), the foundation of this innovation increasingly rests not just on AI talent and research capacity but on the robustness of computational infrastructure. Canada, a country celebrated for its pioneering spirit in AI research and development, now faces a pivotal challenge: the AI compute (AIC) gap. This gap threatens Canada’s standing on the global stage and raises questions about inclusive and sustainable economic growth.

This week, the government announced a $2.4 billion investment to cultivate domestic AI infrastructure and realign itself with peer nations. Last month, we released a report outlining Canada’s policy options to close the AI compute gap. Here, we go over our main findings and use the OECD’s framework of national computing infrastructure to understand Canada’s AIC capacity, its effectiveness and resilience challenges.

Canada’s compute dilemma

“We have the brain. Now we need the mainframe.” Canada’s Minister of Innovation, Science and Industry, Francois-Phillipe Champagne, highlights a critical shortfall in the country’s AI infrastructure—the computational horsepower required to develop and test state-of-the-art AI systems. Despite a commendable history of innovation and a strategic national AI agenda initiated in 2017, Canada’s research and enterprise-ready computational capabilities are lagging not just in comparison to its southern neighbour, the United States, but to all other G7 Nations.

Canada’s storied history in AI research and policy

Recent AI research accolades

The country is home to trailblazers in the AI field, notably Geoffrey Hinton and Yoshua Bengio, who worked extensively on neural network algorithms, and Richard Sutton in reinforcement learning. From 2022 to 2023, Canada showcased significant advancements in its AI sector, illustrating the country’s growing influence and innovation in the global AI landscape. With over 140,000 actively engaged AI professionals, representing a 29% increase from the previous year, Canada has cemented its status as a leading hub for AI talent. The country is home to 10% of the world’s top-tier AI researchers, ranking second globally in this metric. Additionally, Canada leads in gender diversity within the AI field, experiencing a 67% growth in the number of women in AI during 2022-23, the highest year-over-year increase worldwide. It also tops the G7 nations in both the growth of AI talent and the number of AI-related academic papers published per capita since 2019.

Canadian compute capacity in figures

Canada’s computational deficit is an urgent call to action. Even when adjusting for population and economic differences, Canada’s high-performance computing infrastructure lags behind most of its peer G7 nations. In comparative terms, the US leads Canada in adjusted AIC capabilities by a factor of 8 to 11 times. Japan leads by approximately eight times and France and Germany by two to three times.

Country	Unadjusted Compute Performance Capacity[CC1]	Per Capita Performance	GDP Adjusted Performance	Core Performance (PFLOPS)	Population Ratio	GDP Ratio
USA	90.4	10.6	7.6	3725.85	8.56	11.90
Japan	16.3	7.6	8.5	669.83	2.15	1.91
Italy	8.5	5.6	8.9	351.76	1.51	0.96
Germany	6.2	1.9	3.1	256.27	3.21	1.98
France	4.2	2.4	3.2	173.23	1.75	1.30
UK	2.0	1.2	1.4	81.71	1.72	1.44
Canada	1.0	1.0	1.0	41.21	1.00	1.00

---

To account for economic market size and population differences, we calculate a per-capita adjustment ratio, (https://data.worldbank.org/indicator/SP.POP.TOTL?locations=CA-US-GB-FR-DE-JP-IT) and GDP (https://data.worldbank.org/indicator/NY.GDP.MKTP.CD?locations=CA-US-GB-FR-DE-JP-IT) adjustment ratio using TOP500’s supercomputer sub-list generator as of November 2023, https://www.top500.org/statistics/list/.

---

Government support for industry and forthcoming AI compute strategy

In 2017, Canada was the first to launch a national plan for AI development, the (PCAIS), led by the Canadian Institute for Advanced Research (CIFAR). The $443 plan is the backbone of Canada’s position as a leader in AI research, talent, and start-up growth. Yet only $40 million of that funding was directly earmarked for AI compute initiatives. Most AI compute infrastructure is the responsibility of the Digital Research Alliance of Canada and its research partners. However, the PCAIS does not include programmes directly targeting private-sector actors’ AI compute needs, including companies seeking to scale or grow public AI computing infrastructure.

In April 2024, the Canadian Government announced a $2.4 billion investment to cultivate domestic AI infrastructure, a crucial step in addressing its AIC compute gap.

This investment package includes $2 billion to enhance computing capabilities and technological infrastructure for AI researchers, startups, and scale-ups. It provides consultations for a new AI Compute Access Fund and a Canadian AI Sovereign Compute Strategy, ensuring access to state-of-the-art infrastructure, attracting global investments, and fostering Canadian talent. An additional $200 million will support AI startups and the adoption of AI in critical sectors such as agriculture, healthcare, and manufacturing.

Furthermore, $100 million will be allocated to the Natural Research Council’s Innovation Research Assistance AI Assist Program to aid small and medium-sized businesses integrate AI solutions to boost productivity. The plan also includes $50 million for the Sectoral Workforce Solutions Program to offer new skills training for workers in sectors potentially disrupted by AI advancements and $50 million to establish a Canadian AI Safety Institute for the safe development and deployment of AI. Lastly, $5.1 million will be dedicated to enforcing the Artificial Intelligence and Data Act, guiding AI innovation responsibly and protecting Canadians from potential risks.

We posit that the efficacy of this newest investment hinges on its ability to democratise AI’s benefits across all Canadian sectors. This underscores the urgency of a comprehensive, inclusive strategy that addresses the computing gap and ensures equitable access to AI’s transformative potential.

The path forward: recommendations for a unified strategy with AI infrastructure for all

Addressing the compute gap requires more than a financial investment; it demands a commitment to inclusive and equitable access to AIC infrastructure and services. The process must start with extensive public consultation, incorporating voices from a wide array of sectors, particularly those traditionally marginalised in tech innovation. Inspired by international examples such as Taiwan’s digital democracy, Canada’s approach should focus on creating infrastructure and policies that foster inclusive prosperity, ensuring that the benefits of AI extend across the entire spectrum of society. This approach ensures that the voices of scale-ups, small-to-medium enterprises, and marginalised communities are heard in the wake of AI’s growing impact on everyday life.

Once available, the AIC procurement and adoption strategy should be integrated with existing business and digital adoption support policies, such as the Canadian federal government’s innovation subsidies through the Industrial Research Assistance Program and tax incentives via the Scientific Research & Experimental Development Program, to provide rapid and ongoing support for those who need the infrastructure and services to reap AI’s potential for the broader Canadian economy.

AIC policy considerations: Applying the OECD’s national computing framework

Public policy must be designed so that investments benefit the broadest array of AI ecosystem firms and stakeholders. Public investment must not simply replace private investment. Instead, AI computing support strategies should be well-governed and adaptive to evolving policy and industry circumstances. Expediency is key when addressing the current challenges of Canada’s AI sector, and any executed policy has short-to-long-term considerations for expanding high-performance computing infrastructure.

Here is a strategic set of policy options to address current and longer-term computing gaps for Canada developed using the OECD’s framework for national AI computing infrastructure:

• Build domestic AI computing capacity: This approach aims to provide Canadian researchers and firms with direct access to top-tier AIC, fostering innovation and reducing reliance on foreign computing resources. However, this approach faces high upfront costs and the risk of not procuring effective or sufficient AIC capacity in the long run. Programmes to offset domestic infrastructure expansion, like Finland’s LUMI supercomputer business access program, can alleviate downstream operational and maintenance costs.
• Effectively subsidise existing AIC cloud services: To boost the availability and use of AI computing resources, the federal government would need to immediately provide direct subsidies to start-up and early-growth firms. This could be through incumbent cloud service computing discounts and by incentivising domestic incorporation and operations. While this offers an expedient solution, it is temporary and risks increasing dependency on foreign cloud services and could lock in dependence on these digital ecosystems.
• Ensure sustainable, resilient AIC supply chains by collaborating with key trade partners: By partnering with countries like the UK, France, and others within frameworks like the European High-Performance Computing Joint Undertaking, Canada can secure more favourable purchasing power and access to AIC infrastructure procurement and cloud services for its AI ecosystem. This would benefit domestic firms and research institutions that face AIC shortages for future research and development needs and infrastructure procurement.

At the most macro level, any strategy to bridge the AI compute gap must align with Canada’s overarching goals to promote economic growth, ensure digital sovereignty, and maintain leadership in responsible AI innovation. Finally, Canada must distribute AI’s benefits across every part of society, including expanding AI adoption among small-to-medium enterprises and ensuring it supports the middle class and populations currently furthest from tech.

A call to action for a sustainable and inclusive AI sector in Canada

We recommend that Canada’s forthcoming AI policy be human-centred, carefully balancing equitable access to AI-driven economic growth while addressing the urgent computing needs of industry and innovation ecosystems. By prioritising public consultation, leveraging international partnerships, and focusing on sustainable and inclusive infrastructure development, Canada can navigate its current challenges and emerge as a leader in the practical, ethical, and equitable use of AI.

The journey to close Canada’s AI compute gap is not just an innovation challenge; it’s a societal imperative. This gap affects economic productivity and innovation sovereignty. As we venture further into the age of AI, today’s decisions will define Canada’s role in the global digital economy and its ability to foster a prosperous future. The time to act is now, with foresight, inclusivity, and a steadfast commitment to keeping up with computing incumbents and galvanising the prosperity and well-being of all Canadians in the new digital age.