How Canada is poised to become a world leader in AI supercomputing

"With a top-10 supercomputer, Canada is ready to compete in the global AI race based on the talent we have here at home" 
Dr. Ryan Grant, Associate Professor 
Queen’s University 

Canada is a global leader in Artificial Intelligence (AI). To sustain this advantage, world-class domestic infrastructure is essential for our researchers and industries. The Government of Canada is committed to building state-of-the-art, Canadian-owned and -operated high-performance supercomputing. This investment is expected to dramatically enhance Canada's supercomputing capabilities, empowering universities and AI businesses to succeed on the world stage. By incorporating environmental priorities from the start, these systems will also deliver significant benefits to local communities. 

Queen’s and Bell have proposed building just such a supercomputer via the Sovereign Compute Infrastructure Program (SCIP) funding competition. Dr. Ryan Grant is an associate professor of electrical and computer engineering and the head of Queen’s Computing at Extreme Scale Advanced Research (CAESAR) lab. He believes this represents a massive opportunity for Canada. 

“The benefits to productivity and competitiveness could be huge,” says Grant. “It will allow us to keep up and even surpass the current world leaders in the space.” 

Why made-in-Canada, world-leading supercomputing matters 

This project seeks to exponentially increase the compute power available in Canada. “Compute” refers to the computational resources needed for AI systems to perform tasks like training, which involves processing data and even running simulations to produce data for the models. More compute power generally means faster, more accurate AI development. Grant says that’s a big deal for several reasons.  

New research opportunities 

A world-leading supercomputer will enable research and development to advance innovations in areas like health sciences and pharmaceuticals. Researchers will be better able to tackle incredibly complex problems: from designing new drug treatments to unlocking new frontiers in fundamental physics. Businesses can also tap into that research to embed all-new capabilities into their products, helping the economy and creating jobs. 

Stronger AI talent pipeline 

The new supercomputer will position Canada as an attractive destination for the world’s best AI researchers and experts. That includes people who may not have considered coming here in the past, keeping the world-leading experts we already have, and also helping ensure the new leaders we develop have their best opportunities right here in Canada.   

“Top talent tends to migrate to the top supercomputing centres,” says Grant. “They’re always motivated to work with the best tools and supercomputing teams on the biggest, most interesting things.”  

Improved data sovereignty 

Made-in-Canada supercomputing helps ensure researchers and industry won’t have to rely on AI models and infrastructure from other countries. That will help prevent sensitive data from leaving the country.  

It also means more AI-powered solutions can be developed domestically to address Canadian needs and priorities. This is especially important in times of crisis, like a pandemic. In such situations, Grant says it’s vital that Canada is not stuck “waiting in line” to get access to compute power and capacity in another country to formulate vaccines or run public health risk scenarios. 

“Top talent tends to migrate to the top supercomputing centres. They are always motivated to work with the best tools and supercomputing teams on the biggest, most interesting things.” 

The current state of AI supercomputing in Canada  

This new supercomputer won’t be the first in Canada. Our country is already home to several, like the machines at the Cedar Supercomputing Centre at Simon Fraser University. It’s enabling world-class research in climate modelling, health care and more. But we have an opportunity to improve: Canada’s best supercomputer is the 78th most powerful in the world. While a top 100 supercomputer can do amazing things, Grant says it can’t compare to the “leadership class” systems in the top 10.  

The actual ranking isn’t what’s important – but the number also represents an echelon of capability, which is a helpful way of understanding the comparative power of these systems. “A system in the top 30 is twice as powerful as one in the top 70. The 15th-ranked is then twice as powerful as that,” says Grant. “In short, the top 10 supercomputers are together, more powerful than the next 490 combined.” 

What does that look like in practice? A top 100 system can run simulations on a city like Vancouver to determine how vulnerable the city is to an earthquake. A system equivalent to those in the top 10 can go much deeper, modelling the soil and bedrock underneath individual buildings to determine why they’re vulnerable.  

Getting the most from Canadian expertise 

Canada’s aim is to have a system that belongs in that “top 10” class. But that kind of computing is very expensive to build – and it also requires a lot of scientific and technical expertise to operate and maintain.  

The good news is Canada has a lot of that expertise already. It’s in universities across the country, and in institutes like Mila, Vector and AMII – we even have a Nobel Prize laureate in AI physics! To keep these experts in Canada, and to attract more from abroad, we need more places where they can put their skills to use – rather than watching them leave for the AI giants in the United States and elsewhere.  

“It’s like having the talent to field the best hockey team in the world, but not having a world-class arena. Let’s build the facility and have the right people running it – so it’s the best it can be for our players.” 

Building an AI supercomputer, the green way  

Bell and Queen’s University plan to build that arena. If successful with SCIP, Bell will lead the design and construction phases of the building and housing the supercomputer, while also providing the high-speed network connection. Grant’s team at Queen's will develop and operate the supercomputer itself, leading research, chip procurement, system architecture, and advanced technology programs. They’ve also reached a memorandum of understanding with the team at Simon Fraser University to share expertise and deliver scalable, high-performance computing to academia, government and industry from coast to coast. 

But Grant’s goal is to do more than just build a leadership-class system. He wants to create a green supercomputing centre of excellence that will be both environmentally and community friendly. 

“To me, green computing is about more than just being energy efficient,” says Grant. “It’s about how a supercomputer integrates with and benefits the communities around it.” 

Going beyond the state of the art 

Water is a common cooling mechanism, but what if the resulting hot water was then directed to heat nearby homes as part of a district energy heating system in a closed loop, saving the water usage as well? That’s just the start of what’s possible in Grant’s vision of a green supercomputing centre of excellence. 

“We’re going to apply the technologies that exist today to make our centre as green as possible. Then we’re going to push things even further, using the capabilities of the supercomputer to design a more efficient process for heat reclamation,” he says. “We’ll use the state of the art to create an entirely new state of the art.”  

What’s more, Grant sees green supercomputing as an area where Canada can be a real leader. Not every supercomputer is built this way; going green requires a lot of upfront planning. And while it’s possible to recoup much of the cost, there isn’t the same short-term return on investment compared to building less efficient facilities as fast as possible.  

“It’s going to change the landscape. But being green is also just the right thing to do, especially when there’s public investment involved,” says Grant. 

“To me, green computing is about more than just being energy efficient. It’s about how a supercomputer integrates with and benefits the communities around it.” 

Staying ahead of the competition 

Canada’s leadership-class supercomputer will enable new research and business opportunities at both the local and national levels. But Canada isn't the only country looking to make its way into the top 10. Many others are also building out their supercomputing infrastructure. So, while the facilities and equipment are clearly important, it is the brains and talent behind the infrastructure that will make the biggest difference. 

“With a top 10 supercomputer in place, I’d be very happy to compete with anybody worldwide based on the AI talent we have here in Canada,” says Grant. “Our people are amazing. We just have to make sure they have the right tools so they can do the most with their talent.”