Sovereignty, security and the AI chip question

Concealed in a forward position on the edge of a contested city, a reconnaissance team studies a tablet. Drone feeds, signals intercepts and surveillance data converge. A pattern emerges: phone activity, short journeys, irregular hours. Within seconds, the system assigns probability: an observation post.

A target package follows: location, behaviour, associated individuals. Aircraft receive coordinates and strike. Identification to engagement takes minutes. The cycle repeats, limited less by human capacity than by available compute. Military professionals describe this as a “targeting factory”.

This is not a vision of future warfare. It is already how modern warfare operates. Forces that lack this advantage will struggle against those that possess it. The implications are now shaping defence policy in Europe and increasingly for America’s allies more broadly.

AI is much more than an enabler. It compresses time and expands scale, altering how force is applied. Western militaries recognise this shift but face a structural constraint: limited sovereign AI capability. Building it will take time. The operational environment will not wait.

Policymakers therefore face a growing tension. Immediate capability lies in the American AI stack. Sovereign control implies a slower, more uncertain path.

The compute problem

VIEW ALL

At its core, AI rests on compute. Training models, running inference and processing sensor data all depend on it, and therefore on chips.

High-end AI compute is concentrated among a small number of suppliers, largely American. Firms such as Nvidia have built not just hardware, but the architectures and software ecosystems that underpin modern AI. Defence systems are designed around these foundations. Without access to compute, capability degrades sharply.

And that’s a problem for all Western military powers.

The dependency extends beyond Silicon. Systems such as NATO’s Maven are best understood as stacks: hardware, software, data infrastructure and integration. Europe participates across much of this stack but remains weaker at scale in compute.

Sovereignty and trusted foundations

Compute is not neutral. In defence AI, the source matters.

Across the West, policymakers increasingly worry about dependence on American technology, particularly as Washington becomes more willing to use export controls and industrial policy as strategic tools. For countries seeking greater technological sovereignty, reducing that dependence can appear attractive.

Yet the alternatives are not straightforward. China is investing heavily in AI infrastructure, semiconductors and digital ecosystems, and would welcome greater influence over the technological foundations of Western defence. The danger is not that Western militaries suddenly buy Chinese chips. It is that efforts to escape one dependency gradually create another.

The relevant question is therefore not foreign versus domestic. It is whether Europe and its allies can build sovereign capability on trusted foundations while retaining control over how that capability is applied.

Why Western militaries are concerned

The West’s defence institutions seek sovereignty because it confers control over supply, systems and decision making. Accepting reliance on external compute complicates that aim. In crisis, access might be shaped, delayed or conditioned. This is not a prospect that sits easily with policymakers already uncertain about the durability of trans-Atlantic guarantees.

Fair enough.

The concern is not uniquely European. Australia, Japan and other US-aligned states face the same dilemma: how to build greater technological resilience without weakening the security architectures on which they depend.

But at the same time, much of the enabling technology sits within a US regulatory framework. A full-stack alternative – spanning chips, infrastructure, software and integration – is a decades-long industrial effort.

Recent defence debates, at least the public ones, have focused on limited mass and readiness. Yet in defence departments and operational headquarters, the ability to compress time and scale effects through AI is becoming equally central. Defence planners know that adversaries are not pausing.

Managing the constraint

The practical response is alignment and managed interdependence. Access to US-designed chips remains the fastest route to fielding advanced systems. Without it, non-US forces risk operating at a disadvantage.

This is an extension of an established model. Trans-Atlantic and Indo-Pacific defence cooperation has long relied on shared systems, standards and industrial integration. AI moves that logic into a more foundational layer.

A twin-track approach follows. First, secure access: deepen relationships with US suppliers, ensure supply and maintain interoperability. Treat compute as a critical defence input. Second, build selectively: focus investment on integration, applications and control of data and systems. These are the layers where sovereignty has most practical effect.

Overly rigid precaution may slow deployment. Defence has historically operated with greater latitude than civilian sectors and is likely to continue to do so.

A hybrid model

The most plausible outcome is not separation, but overlap. US compute is likely to remain at the core of allied defence AI for years to come, while sovereign capability develops around integration, applications and data control. These determine how capability is used rather than simply where it originates.

American incentives

Dependence is not one-sided. US export controls introduce friction, but they are balanced by commercial and strategic incentives.

Western militaries represent a growing market for AI compute and defence systems. The United States also relies on capable allies; restricting access in ways that weaken them carries costs of its own.

These factors point in the same direction. Continued supply of high-end compute serves American interests. Managed integration serves the whole of the Western alliance.

If Western defence increasingly depends on AI and AI increasingly depends on compute, the policy challenge is not simply to achieve sovereignty. It is to avoid exchanging one dependency for another.

The development of hybrid systems allows Western militaries to retain operational advantage while laying the groundwork for greater autonomy over time. For now, that means building capability on trusted foundations while investing in the sovereign layers that will matter most in the future.

AI capability cannot wait for full sovereignty. Nor can sovereignty be indefinitely deferred. The challenge for non-US Western powers is to strengthen both without drifting into a competing technological ecosystem.

James Maclaren is a freelance journalist and writer specialising in defence and security. He is a former UK military officer with command, staff, joint experience in a variety of roles and operational environments. He is based in London and the Far East.