Canada's AI Plan Forgot the Chips, Council Says — and It Doesn't Want New Money

Canada's chip plan is not a call to rival Nvidia. It is, in the Canadian Semiconductor Council's telling, a plumbing exercise: take the money that already exists in the AI Compute Access Fund, the BDC LIFT Fund, and the Canadian Tech Growth Fund, and direct AI-strategy procurement toward the photonics, advanced-packaging, and interconnect firms the country already has.

CSC's position paper, "Sovereign AI Needs Sovereign Silicon", lands in the formal comment window for Ottawa's freshly released "AI for All" strategy. The strategy correctly flags dependence on foreign AI chips as a strategic vulnerability, the council says. It stops short of treating semiconductors as a policy priority, and it does not say who should build the hardware behind the compute capacity it wants to grow.

"What we are saying is that the funding vehicles already exist. It's about letting Canadian industry plug into them," Paul Slaby, CSC's managing director, told EE Times.

The three vehicles the council names are already capitalized. The AI Compute Access Fund is a federal program that subsidizes Canadian firms' access to AI computing capacity. The BDC LIFT Fund is the Business Development Bank of Canada's late-stage growth-capital arm. The Canadian Tech Growth Fund is the country's flagship domestic growth-equity vehicle. None of them is new; none of them needs a fresh appropriation. The ask is procurement language that names Canadian lanes, not new spending.

The paper rejects the idea that Canada should try to build a domestic fab rival, and rejects the idea that Canada should compete in the frontier-model race, the global race to build the largest AI systems currently led by a handful of well-funded U.S. and Chinese labs. The procurement ask is not a Canadian Nvidia. It is a Canadian tier in the global supply chain.

The position paper identifies four Canadian strengths the strategy should steer procurement toward: photonics, the light-based components that move data on and off chips; compound semiconductors, the specialty materials such as gallium nitride used in power and radio-frequency devices; advanced packaging, the assembly steps that turn bare silicon into usable systems; and chip design. The named assets are concrete. The Canadian Photonics Fabrication Centre (CPFC) was recently spun off from the National Research Council in Ottawa and now operates as an independent fabrication facility for photonic chips. Bromont, Quebec, hosts Canada's cluster of advanced-packaging facilities, a node the global semiconductor industry has spent the last decade treating as strategically scarce. The council also notes that Canadian optical interconnect technologies are already inside products shipped by Nvidia and AMD, two of the three companies that effectively set the floor for AI compute.

The mechanism only works if Ottawa writes it in. The AI strategy, as currently drafted, calls for substantial growth in sovereign compute capacity, demand the position paper pegs at hardware worth billions of dollars. Without a procurement-instruction clause favoring domestic suppliers in the named lanes, that demand flows to incumbents. With one, the existing funds can be steered toward Canadian firms without new appropriations, which is the only way this gets through a federal budget process that is already crowded.

The falsifier is in the final strategy text, not in this position paper. If "AI for All" ships without a semiconductor pillar and without procurement language that names domestic lanes, the existing vehicles keep flowing where they currently flow, and the proposed mechanism does not engage.

The comment window on the AI strategy is the lever. Whether Ottawa writes silicon in as a pillar, and whether the procurement language follows, is the watch item.