AI
Toronto's $2.9 billion chip deal and Ottawa's compute plan: what it means for Canadian business
AMD is paying about $2.9 billion for Toronto chip startup Taalas, and Ottawa plans 500 MW of new compute capacity. What Canadian businesses should take from both.

Two pieces of Canadian AI news landed within days of each other. AMD disclosed it will pay about 2.9 billion US dollars for Taalas, a Toronto startup that builds AI chips. And the federal budget proposed new public backing to create 500 megawatts of Canadian compute capacity. Together they say something simple: Canada is trying to own more of the AI stack, from chips to data centres, instead of only renting it.
What Taalas builds
Taalas was founded in August 2023 in Toronto by Ljubisa Bajic, Lejla Bajic and Drago Ignjatovic. The company builds model-specific inference silicon: chips with a single trained AI model built into the silicon, so that model runs fast and cheaply. The two publicly named products are the HC1 and the HC2.
AMD announced the acquisition agreement on 6 August 2026, without naming a price. The quarterly filing lodged on 4 November put the value at about 2.9 billion US dollars in cash and stock, including up to about 1.4 billion dollars in contingent consideration. Closing is expected by the end of 2026, subject to regulatory approval. AMD's chief software officer reportedly said the Taalas team would join AMD's Canadian workforce.
Why AMD is paying that much
Inference, running models, is where the volume is. A specialised chip can run a single model far more cheaply than a general-purpose processor. AMD launched the Instinct MI400 series in July 2026. Taalas adds a different kind of chip to that picture.
For Canada, this is a talent and research story, not just a deal story. A three-year-old company from Toronto built something a global chipmaker wanted badly enough to pay billions for.
Ottawa's 500 megawatt plan
The federal budget, presented in the first week of November, proposed using Ottawa's financial tools and guarantees to increase domestically controlled compute capacity. Federal agencies could issue backstops and offtake agreements to help create 500 megawatts of new compute capacity.
The background: Meta is building a 13 billion dollar data centre campus in Alberta. Anthropic will lease space at a facility in Nova Scotia. Developers have announced nearly 200 new data centres across Canada. In the United States, voters in dozens of municipalities approved bans on new compute facilities, which pushes developers to look north.
Ottawa also announced on 29 October an equity stake of up to 1.6 billion Canadian dollars in Toronto-based Cohere. Cohere has committed to creating 1,000 new tech jobs and investing 4 billion Canadian dollars in research and infrastructure by the end of 2031. Those are company commitments, as reported.
What it means for a Canadian business
First, talent. Chip and AI engineering talent is more likely to stay in Canada when global employers build teams here. For small businesses that means more competition for skilled people, but also more partners nearby who understand AI systems.
Second, work around the build-out. Data centres need construction, power, cooling and IT services. Local firms that offer those services should follow the project lists in their province.
Third, costs and choice. More domestic compute capacity could mean more choice among providers. That is a possibility, not a promise. Only count on lower prices when quotes are actually on the table.
Fourth, data at home. For businesses that must keep data in Canada, it matters where the compute actually runs. Ask your providers, instead of assuming.
Fifth, Nova Scotia is at the centre of the debate. Alongside the Anthropic lease there is the proposed 10 billion dollar data centre and protests against it on 7 November. Businesses in Atlantic Canada should also note the ACOA call for AI projects, which closes on 20 November. The details are in our article on AI funding in Atlantic Canada.
What to watch next
Regulatory approval and the closing of the Taalas deal, expected by the end of 2026. Whether the 500 megawatt plan gets concrete projects. And how the questions about power, sites and communities are answered. That is the part of the story that decides whether announcements become jobs and contracts.
The takeaway
A 2.9 billion dollar deal and a 500 megawatt plan will not change your AI bill overnight. But they change the direction: Canada is building its own AI infrastructure, and businesses that know early where compute, talent and funding are forming will have a head start. For what Budget 2026 means for AI jobs, see Canada Budget 2026: AI jobs.
Frequently asked questions
What does Taalas make?
Taalas builds model-specific inference chips. A single trained AI model is built into the silicon, so it runs fast and cheaply. The products are named HC1 and HC2.
How much is AMD paying for Taalas?
About 2.9 billion US dollars in cash and stock, including up to about 1.4 billion in contingent consideration. Closing is expected by the end of 2026, subject to approval.
What is Ottawa's 500 megawatt plan?
The federal budget proposes using backstops and offtake agreements to create 500 megawatts of new, domestically controlled compute capacity. Concrete projects are still to come.
Does this change what my business pays for AI?
Not directly, and not yet. What matters over time is more choice among providers, more talent in the country, and more clarity about where data is processed.
Working on something like this?
Tell us what you are building. We reply to every enquiry.
Related service: AI Search Visibility (GEO)


