Carney Names AI Critical Infrastructure as Canada Courts Middle Powers on Compute

Carney Names AI Critical Infrastructure as Canada Courts Middle Powers on Compute Canada's plan to lead non-superpower nations in sovereign AI puts national energy systems at the center of the compute race. Canada's Prime Minister Mark Carney released a national artificial intelligence strategy on Thursday (2026-06-04) that treats AI as critical infrastructure on par with energy, a bid to lead the world's middle powers as they race to build sovereign AI capacity.⁷ The energy read is in the framing. When a G7 government places compute alongside the grid as protected national infrastructure, it is committing to the power demand that large-scale AI deployment requires, in a country whose generation mix is one of its few competitive levers.⁷ The contest Carney is entering has so far been defined by the two largest economies. The United States, third by population with just under 325 million residents, anchors one AI stack; China anchors the other.¹ Sinocism's reading of recent US-China summit outcomes described the strategic contest entering a phase of stalemate, with senior Chinese officials Li and Ding addressing AI as a domain of competition.^3,2 That leaves a tier of capable but smaller states looking for room to maneuver. Canada, a major economic player with around 36.5 million residents for its landmass, is positioning itself as the convener of that group rather than a follower of either stack.^1,7 The talent question sits underneath all of it. The Atlantic Council frames readiness as a ladder from AI literacy, understanding core capabilities and concepts, to fluency, knowing how to use the technology to solve real problems.⁶ The distinction matters for any government promising to stand up sovereign capacity, because the build-out needs operators, not just chips and power.⁶ There is a labor-market sweetener in the data. LinkedIn figures cited by the Atlantic Council show AI added 1.3 million jobs in 2024-25, a counter to the assumption that the technology only displaces work.⁶ But the allied-stack approach has a tell. The Atlantic Council notes that frameworks for exporting technical talent to deploy AI in developing-world public services still emphasize the US AI stack, even where import partners are invited in.⁵ A middle-power bid for sovereignty runs into the gravity of the larger ecosystems it is trying to sit between.⁵ The wider setting is an investment environment turning more defensive. UNCTAD counted a record number of new measures restricting foreign investment, and calculates that 63% of global investment flows were subject to a screening regime last year, up from 52% in 2020.⁴ Industries accounting for 60% of the value of America's stock markets fall under the potential remit of CFIUS, judging by deals submitted in 2021.⁴ Compute and the energy assets that feed it are exactly the kind of strategic infrastructure screening now captures.⁴ For Germany, the read-across is direct. With a population near 82 million, it sits inside the same middle-power calculus Carney is appealing to, and faces the constraint sovereign AI ambition collides with: power prices, with German baseload front-month near €101/MWh and ICE Endex TTF front-month near €47/MWh making data-center electricity a live cost rather than an afterthought.¹ For now none of this moves a price. The strategy is a statement of intent, not a procurement order, and the gap between designating AI as critical infrastructure and actually wiring gigawatts of new load is where most of these national plans will be tested.⁷ Watch whether Canada's plan attaches firm power-supply commitments or stays at the level of designation, whether other middle powers, Germany among them, follow with their own sovereign-compute strategies, and whether tightening investment screening starts to capture the cross-border data-center and grid deals the AI race depends on.^7,4