Engineering optionality
What Canada’s Defence Industrial Strategy gets right on AI Sovereignty
Canada’s new Defence Industrial Strategy is not a document about tanks and ships with a performative paragraph about “innovation.” It’s a document that quietly articulates the reality of AI in 2026:
AI is not a feature. It’s infrastructure. And infrastructure is geopolitics.
The Strategy even spells it out in bureaucratic plain English: Artificial Intelligence sits inside “Digital Systems” as a Key Sovereign Capability, alongside secure cloud and quantum.
That’s the tell. Once a government starts using the word sovereign around a technology, it’s no longer talking about “cool models.” It’s talking about freedom of action under pressure. And for a middle power, freedom of action is not built by declaring sovereignty. It’s built by engineering optionality; across the stack, across supply chains, across alliances. Variable geometry, but with contracts.
The Strategy’s AI claim: “this matters for power”
The document frames “frontier technologies, such as AI and quantum computing”, as game-changing on the battlefield and economically transformative. That matters because it puts AI into the category my innovation diplomacy framework calls high-salience: domains where marginal advantage translates into real leverage.
Once you accept AI is high-salience, a lot of things follow:
You can’t treat research as purely “open by default.”
You can’t treat procurement as mere purchasing.
You can’t treat alliances as vibes.
You need strategy that can survive contact with rivalry. Which brings us to the Strategy’s real architecture.
BUILD–PARTNER–BUY is variable geometry written in procurement verbs
The Strategy’s core triad, BUILD, PARTNER, BUY, isn’t just industrial policy. It’s a middle-power doctrine.
BUILD: develop domestic capacity where sovereignty is non‑negotiable, including “frontier areas, such as AI”.
PARTNER: work with “trusted allies” when Canada can’t (and shouldn’t) do it all, explicitly naming deeper relationships beyond the U.S. with the EU and UK, and Indo‑Pacific partners like Japan and Korea
BUY: acquire from others when necessary, but with domestic reinvestment and conditions that protect Canadian control.
That is variable geometry as engineering: different coalition shapes for different capability layers.
The point isn’t purity. It’s stable flight.
The real “AI stack” move: universities become defence infrastructure
If you want the most consequential AI paragraph in the whole Strategy, it’s not a line about algorithms. It’s the part where the document rewires the relationship between universities and defence capability.
It describes modern conflict expanding into the digital domain “driven by technologies such as AI…”, then immediately grounds Canada’s response in the research base: colleges and universities, backed by the granting councils and CFI.
From there, it gets operational:
A Science and Research Defence Advisory Council to connect post‑secondary research to defence priorities.
BOREALIS, a new mechanism to accelerate defence research “particularly in frontier technologies such as AI…”
A national network of Defence Innovation Secure Hubs, where security‑cleared academic researchers collaborate with government and industry in secure environments.
That last point is the giveaway that we’ve crossed into a new regime. Canada is not just funding AI. It’s building secure institutional plumbing to move ideas from lab to deployment without pretending dual‑use risk is imaginary. In innovation diplomacy terms: that’s investment in verification capacity, the hinge variable that lets trusted coalitions expand without naïve leakage.
Procurement is diplomacy, and the Strategy admits it
The Strategy puts the Defence Investment Agency (DIA) at the centre and explicitly ties it to multilateral capability development, joint procurement programs aligned with SAFE and Readiness2030, and “national diplomatic efforts”.
That’s not window dressing. It’s a recognition that in high‑tech domains, diplomacy is often conducted through:
joint procurement,
interoperability requirements,
shared standards,
export promotion,
and coalition supply chains.
The Strategy even builds out the diplomatic machinery at a time when government is set on reducing the size of the civil service: more trade commissioners, stronger roles for defence attachés, “deal teams” for major contracts.
This is the state saying: we are going to compete as a coordinated system: research, industry, procurement, and diplomacy pulling in the same direction.
“Walls and bridges” is not hypocrisy. It’s the pattern.
My innovation diplomacy note argues that governments are building walls and bridges simultaneously, restricting flows in high‑risk areas while accelerating collaboration inside trusted clubs. The Strategy follows that logic:
Bridges:
secure hubs,
accelerated defence R&D,
joint procurement programs,
expanded partnerships across allied networks.
Walls:
explicit emphasis on protecting sensitive technologies and research from hostile actors,
research security frameworks and tools of screening and control
Add the supply‑chain layer (e.g., critical minerals, coordinated with allies through initiatives like the G7 Critical Minerals Production Alliance and NATO-led efforts) and you can see the stack logic peeking through the defence language.
AI isn’t floating in the cloud. It’s sitting on minerals, energy, chips, compute, standards, and institutions.
The real test: does this create optionality, or just better dependency?
A strategy like this can succeed… or it can become theatre. The difference is in the measurement. My innovation diplomacy framework insists on diagnostics: policies should specify observable indicators that trigger review, and we should defend openness where closure yields no security dividend.
So the question isn’t “does Canada say ‘sovereignty’ a lot?”
The questions are:
Do Canadian AI-capable firms actually win procurement in meaningful ways, or do they remain subcontractors in someone else’s value chain?
Do secure hubs and clearances reduce friction for serious collaboration, or do they create a new layer of process that mainly rewards incumbents?
Does partnering with allies increase optionality: multiple pathways, multiple suppliers, shared standards, or does it quietly harden dependency on a narrower club of primes and platforms?
That’s the middle-power dilemma in the AI era: you can’t do everything yourself, but you also can’t afford to be merely “a safe place to park other people’s compute and IP.”
The Strategy’s real story is a new kind of statecraft
Canada is practicing realist AI diplomacy by treating procurement as a coalition instrument and universities as capability infrastructure. That’s variable geometry applied to the hard layers of the AI stack: research pipelines, secure collaboration, allied clubs, supply chains, and the demand signal of procurement.
It won’t be perfect (no strategy survives first contact with vendors, timelines, or politics) but it’s a strategy that feels like it understands the era we’re in: not a transition, a rupture; where the fight isn’t just over ideas, but over the machinery that turns ideas into power.