Kingston-based Queen’s University wants to build a Canadian supercomputer, and it has snagged an American leader in the high-performance computing space to help.
Queen’s has recruited Ian Karlin as assistant professor to co-lead its supercomputing research efforts alongside associate professor and existing technical leader Ryan Grant. Together, the pair hope to help Canada catch up in the global compute race.
Canada lags behind the US but many other nations in its compute capacity. The country’s most powerful supercomputer is Telus’ Sovereign AI Factory, which was established earlier this year and placed 78th globally in Top500.org’s latest rankings of the most powerful non-distributed computer systems in the world.
“It makes sense to work with the folks that have already done that.”
Ryan Grant,
Queen’s University
But the Government of Canada has made big commitments to change that. The feds have allocated up to $705 million out of the $2-billion compute spending package announced last year to build a large national AI supercomputing facility to support researchers and companies.
Karlin joins a growing list of talented US researchers who have moved to Canada amid US government research funding cuts and attacks on post-secondary institutions.
He most recently served as a principal engineer at chip giant Nvidia, and has worked on El Capitan, the world’s most powerful supercomputer. He has also served as technical lead on two other upcoming US supercomputers, Doudna and Mission, and collaborated with research labs across Europe and Japan.
In an interview with BetaKit, Grant and Karlin explained their decision to join forces, how shifting US attitudes towards academia could lead to a Canadian “brain gain,” and pitched why Queen’s should be tasked with building a top 10 Canadian supercomputer.
“I was thinking about moving back into research,” Karlin said. Between this role, which Karlin described as “almost written for what I do,” and the situation in the US research community, “it seemed like the right time,” he added.
RELATED: Federal government outlines $2 billion in AI compute spending commitment
Grant said that a Canadian supercomputer would give the country the in-house horsepower it needs to conduct massive simulations, train large AI models, and fuel “revolutionary” new research. He said a supercomputer could help answer big questions from academia and industry about a wide variety of topics, from jet engines to forest fires and pandemics.
“We’re toolmakers and tool improvers, and so we very much see our role as service to others, where we build the big tool that others use to make those big Nobel Prize discoveries,” Grant said. “We just make it possible for them to get there faster.”
A local supercomputer would also keep this system and the data that is fed into it in Canadian hands and free from the coercion of foreign governments—an increasing concern given the ongoing US trade war.
Queen’s wants to build and house that facility, and is in the process of preparing a bid for when the Government of Canada begins seeking proposals. Grant is hopeful that this will happen in 2026, but said no official timeline has been released yet. Queen’s intends to collaborate with Bell Canada on its proposal.
Grant argued that the university’s geographic proximity to Ottawa, Toronto, Vector, and Mila put it in the right location, and thinks that its existing supercomputing expertise, coupled with his and Karlin’s experience—he claims they are the only two people in Canada who have helped design and procure next-generation supercomputers—make it the right fit for the job.
“It makes sense to work with the folks that have already done that,” Grant said.
Efforts to slash funding for research and put pressure on US universities under President Donald Trump have led postsecondary institutions in other countries to try luring talented US scientists away. The University of Toronto has been campaigning to attract some of them to Canada, most recently hiring three top US scholars.
RELATED: U of T establishes AI research chair to recruit the next Geoffrey Hinton
Grant and Karlin see room for Canada to poach more talented US scientists amid these conditions, and said the pair are keeping a list of folks who have expressed “serious interest” in playing a part in Queen’s efforts to build a Canadian supercomputer.
“There’s some very senior, very big names on that list that would make their bosses scratch their heads,” Karlin said.
This could be important: Grant argued that Canada does not have enough talent domestically to fully staff a top 10 supercomputing centre. To run such a facility, Grant said the country will need to bring in some foreign experts to train its existing workforce.
Grant added that this global supercomputing supply and demand issue creates “a really massive opportunity for us for brain gain and brain retain.”
Feature image courtesy Queen’s University.