Feds to plan boost AI development, subsidize compute costs and build data centres
The federal government plans to spend $1-billion to build supercomputing infrastructure to power artificial intelligence models and another $1-billion to subsidize companies that want to build data centres and access resources for AI.
Innovation, Science and Economic Development Canada provided new details Thursday of the government’s AI infrastructure strategy, which was first announced in April. At the time, Prime Minister Justin Trudeau said Ottawa would spend $2-billion to help Canadian companies and academia access the expensive graphics processing units (GPUs) and other infrastructure, generally called “compute,” that is crucial to building and running advanced AI models. Compute is necessary to increase AI adoption and capture the economic benefits of the technology, proponents say.
ISED said Thursday that the government will spend as much as $700-million to spur the development of new AI data centres in Canada and expand existing facilities. The government said it has opened an application process to companies and public-private partnerships to access the funding. The criteria notes that proposals should include significant provincial or territorial involvement and have “meaningful participation” by Canadian industry.
Foreign companies such as Amazon.com Inc. and Google dominate the market for data centres and AI infrastructure and are already spending billions of dollars to build facilities around the world. The federal criteria does not appear to prohibit foreign companies from applying along with Canadian partners to receive funds.
“It is essential to ensure that the allocated funds are deployed effectively and efficiently to maximize their impact on Canadian companies and not simply flow to foreign AI companies or large multinational infrastructure providers,” said Dana O’Born, the vice-president of strategy and advocacy at the Council of Canadian Innovators, an organization that represents tech companies.
The government will also prioritize projects that set aside affordable capacity for Canadian customers and use sustainable energy sources.
In the near term, another $300-million will subsidize the AI compute costs for small and medium-sized businesses in key sectors such as health care, energy and advanced manufacturing. But the details will not be made public until the project launches next spring.
“The barrier to entry for many small and medium-sized enterprises comes strictly down to cost,” said Graham Dobbs, a senior economist at the Dais think tank at Toronto Metropolitan University. “These costs can run up fairly high before you have a substantial customer base or revenue generation.”
About $800-million of the funding will be used to build a new publicly owned supercomputing facility, along with a smaller secure site, that can be accessed by academics and industry players. Another $200-million will go to institutes that already manage publicly owned computing infrastructure, such as the Digital Research Alliance of Canada, the National Research Council of Canada and the country’s three national AI hubs: the Mila-Quebec Artificial Intelligence Institute, Amii (Alberta Machine Intelligence Institute) and the Vector Institute.
Canada has fallen behind other countries when it comes to AI, according to some measures. The country dropped to eighth place from the No. 5 slot in the Tortoise Global AI Index this year, which evaluates countries on a variety of factors related to artificial intelligence. Canada also ranks 18th in the world for AI infrastructure, behind Germany, France and Britain.
When Ottawa announced $125-million in funding to build three AI institutes in 2017, the compute resources to support researchers were not much of a consideration. Since then, AI models have grown bigger and more resource-intensive, and Canada’s compute capacity for academic research has lagged.
In the private sector, demand for GPUs has also surged in recent years as generative AI has taken off, with some companies having trouble accessing these resources at an affordable price.
Separately, Alberta is making a pitch to companies to build AI data centres in the province. Nate Glubish, the Technology Minister, said on Wednesday he’s hoping to attract as much as $100-billion worth of AI data centres to Alberta in the next five years. The province can use its abundance of natural gas to power these data centres, which require immense amounts of electricity, he has said.
But concern about the emissions associated with data centres and the stress on power grids is mounting. A report from Royal Bank of Canada released this week said that if an additional six gigawatts of data centre capacity is added to electrical grids and powered by natural gas, Canada’s annual emissions could rise 3 per cent, though that could be mitigated through carbon capture and storage.
This article was first reported by The Globe and Mail
