Sovereign artificial intelligence | The strategy to provide Canada with its own AI is launched

The path to a sovereign Canadian AI is starting to take shape. The Minister of Innovation, Science and Industry, François-Philippe Champagne, formalized Thursday the 2 billion federal strategy first announced by his government last spring.

The first $700 million from this envelope will serve as financial assistance for Canadian companies wishing to become “Canadian AI champions”, by building or expanding computer data centers that will increase the computing capacity installed in Canada. A sum of 300 million will also be intended to ensure better access to cutting-edge artificial intelligence tools for Canadian SMEs.

Patience…

The other billion dollars planned for this strategy will be used to establish a public computing infrastructure for AI. We will see next spring how this infrastructure takes shape, since that is when Ottawa will launch a call for proposals to find suppliers able to develop this technology.

We will have to be even more patient before seeing this infrastructure be functional. Consultations with industry last summer allowed the government to note that this could take a long time to achieve. A sum of 200 million was therefore set aside to improve current public infrastructure. Another portion of this total investment will be used to build a smaller, more secure AI infrastructure, the use of which will be managed by Shared Services Canada and the National Research Center of Canada (NRC).

“We are a driving force in the global ecosystem and, by increasing access to secure national computing capacity, we will help businesses, innovators and researchers boost the Canadian economy and stand out on the international stage », declared François-Philippe Champagne in a press release.

A strategy and its challenges

The federal minister thus poses himself in opposition to the increasingly present feeling in the AI ​​sector according to which the creation from scratch of a new infrastructure capable of competing with the existing systems of OpenAI, Google and of Meta, in particular, will cost a lot more than we think.

Dario Amodei, the CEO of AI company Anthropic, recently estimated that developing such technology could cost more than US$5 billion.

The availability of electricity necessary to power this infrastructure could also represent another challenge for the realization of the federal strategy.

Data centers consume energy, and electricity producers like Hydro-Québec, who see demand for electricity increasing and their production capacity stagnating, are now reluctant to accommodate such centers on their network.

The Quebec company QScale recently had to put on hold a project to expand its data centers, precisely because Quebec refuses to grant it the necessary energy block.

Read “QScale: customers, financing… but no electricity”