# What is the Canadian Sovereign AI Compute Strategy, and where does the $2 billion go?

Date: 2024-12-06 · Author: Robert Scales · https://guidetoartificialintelligence.com/en/learn/canada-sovereign-compute-strategy/

**Answer:** It is a federal plan launched on December 5, 2024 that spends up to $2 billion on the computing power behind AI, split among commercial data centres, public supercomputers and help for smaller companies. Its aim is to give Canadian researchers and businesses access to that capacity at home instead of relying only on systems elsewhere.

On December 5, 2024, Innovation Minister François-Philippe Champagne unveiled the Canadian Sovereign AI Compute Strategy. On the next day, Deputy Prime Minister Chrystia Freeland announced its first investment: up to $240 million for Cohere, the Toronto company that builds language models for businesses.

The money itself is not new. Budget 2024 promised the $2 billion. What arrived this week is the plan for dividing it, and the first application process. Consultations over the summer drew input from more than 1,000 people and organizations in research, industry and civil society, according to the government.

## What is compute, and why does a government fund it?

**Compute** is the computing capacity needed to train and run AI systems: specialized chips, the servers around them and the data centres that house and power them. Training a language model means repeating enormous numbers of calculations on those chips, so a lab without enough of them can test fewer ideas and train smaller models.

Think of compute as the road network of AI. A country can have excellent drivers, and Canada has many: the government says it hosts 10% of the top AI researchers in the world. But if the roads belong to someone else, others decide the tolls and the access. That worry is what the word "sovereign" points at here. In this strategy it means capacity located in Canada, with data processed and stored inside the country.

## How the $2 billion is divided

| Part | Up to | What it pays for | When it helps |
|---|---|---|---|
| AI Compute Challenge | $700 million | New or expanded AI data centres, from companies, consortiums and academic-industry partnerships | Medium term |
| Public computing infrastructure | $1 billion | A large sovereign supercomputing facility and a smaller secure facility; $200 million of this tops up existing public systems | Short to long term |
| AI Compute Access Fund | $300 million | Help for Canadian businesses to buy compute, with details expected in spring 2025 | Near term |

The middle row has the longest timeline. BetaKit reports that Ottawa plans to start seeking partners for the large facility in spring 2025, with a preference for proposals that bring in private capital. The smaller facility would be led by Shared Services Canada, which supplies technology to federal departments, and the National Research Council. Because new buildings take time, the $200 million goes to systems run by the National Research Council, the three national AI institutes and Canada's Digital Research Alliance.

The Compute Challenge is competitive. The government says priority goes to Canadian projects that show a high return on public money and sustainability. BetaKit adds that the use of Canadian-made hardware or software also counts in a project's favour.

## The first cheque: Cohere

The December 6 announcement says the $240 million will help Cohere secure enough private capital to persuade its strategic partners to build a Canadian AI data centre worth several billion dollars, online in 2025. Cohere and other Canadian firms would use its capacity. Freeland described the investment as "crowding in" private money, meaning a smaller public amount is meant to draw a much larger private one.

Two things are not in the announcement: the names of those partners and the location of the data centre. Anyone weighing how "sovereign" this project is will want to know who owns and operates it once those details are public.

## Why it matters, and the questions it leaves open

Seen from this week, the strategy puts Canada in a group of countries treating compute as national infrastructure. *The Register* notes similar plans in India, Japan and Europe. It also makes a point worth keeping in mind: in the short term, American chipmakers stand to gain, which it suggests may be one reason Nvidia CEO Jensen Huang has promoted the idea so actively. A data centre in Canada will still run chips designed and largely built abroad.

Scale is another question. Ottawa wants its public money to pull in far larger private sums, so the strategy's success depends on how much private money the public share attracts, and on whether the new capacity is priced so that smaller Canadian players can really use it.

Here is a simple way to test any future project announced under the strategy. The last column is yours to fill in.

| Question | Why it matters | Answer for the project |
|---|---|---|
| Who owns the building and the chips? | Ownership decides who earns the profit | |
| Who operates the systems day to day? | The operator holds administrative access | |
| Which country's laws bind the operator? | Those laws may reach the data | |
| What share of capacity is open to other Canadian firms, and at what price? | That is the stated public benefit | |

> **Try it:** When the next AI Compute Challenge project is announced, fill in this table from the press release. Write "not stated" wherever the release is silent, and notice how many lines that is.

## How could a Canadian business or researcher benefit from this strategy?

For a small or medium-sized company, the AI Compute Access Fund is the most direct route, because it is meant to help pay for compute. Its rules are not out yet, so the practical step today is to work out what you would use it for. Estimate how many hours of training or inference you need, and what that costs now. Having that figure ready makes an application easier to prepare.

For a researcher, the near-term gain is likely to come through the existing public systems that receive part of the $200 million, so your institution's research computing office is the place to ask. For a company or a university consortium planning a data centre, the AI Compute Challenge is open to applications now.

If you only use AI through an app, nothing changes for you this month. The strategy is about the layer underneath, and new capacity takes time to build.

## Sources

- Canada to drive billions in investments to build domestic AI compute capacity at home: https://www.canada.ca/en/innovation-science-economic-development/news/2024/12/canada-to-drive-billions-in-investments-to-build-domestic-ai-compute-capacity-at-home.html
- Deputy Prime Minister announces $240 million for Cohere to scale-up AI compute capacity: https://www.canada.ca/en/department-finance/news/2024/12/deputy-prime-minister-announces-240-million-for-cohere-to-scale-up-ai-compute-capacity.html
- Federal government outlines $2 billion in AI compute spending commitment: https://betakit.com/federal-government-outlines-2-billion-in-ai-compute-spending-commitment/

---

From the Guide to Artificial Intelligence, by Robert Scales, published by Holistic Hybrid Inc. Share and adapt it under CC BY-NC-SA 4.0 (https://creativecommons.org/licenses/by-nc-sa/4.0/): credit the guide, link to https://guidetoartificialintelligence.com/en/learn/canada-sovereign-compute-strategy/, no commercial use without permission, and keep the same licence.