Skip to Content

Who Owns the Infrastructure Owns the Future. France Knew It. Australia Just Said It.

8 April 2026 by
Arnaud Couvreur


There is a particular kind of morning in Perth (the Indian Ocean still too bright to look at directly) when the distance between France and Australia collapses into something almost manageable. I had one of those mornings recently, reading through a government media release with my second espresso, thinking: this is the kind of thing that would barely register in a Paris boardroom, and that almost nobody in Canberra will have connected to what France has been quietly building for the past five years.

The document was Australia's new national framework for data centres and AI infrastructure, published on March 23. Two weeks later, I was still thinking about it. Not because of the technology. Because of the question underneath the technology.


A question hiding in plain sight


I have been working recently on AI integration projects: how organisations actually adopt these tools, what assumptions get buried in the process, what the risks look like when you get close enough to see them clearly. (I wrote about transparency and mandatory labelling here, if you want the longer version.) What struck me, working through those projects, was that the most interesting questions were never really about the software. They were about where the data lives. Who has jurisdiction over it. What happens to it when the vendor's country of incorporation decides it has a right to access it.

These are not abstract questions. They are the operating conditions that boards and executive teams are navigating right now, whether they have noticed it or not. The Australian government's framework made them explicit. And in doing so, it echoed something I have been watching France wrestle with for years.


What Australia just decided


The framework, titled Expectations of data centres and AI infrastructure developers, creates no new legal obligations. That is worth saying clearly, because the temptation is to read it as regulation and dismiss it on those grounds. It is not regulation. It is something subtler, and in some ways more significant: a statement of conditions. Five of them: national interest, energy transition, water sustainability, local skills investment, and innovation. The government will prioritise projects that meet them. Those that do not will find their path through regulatory processes considerably less smooth.

The message to the hyperscalers is not complicated. Between 85 and 135 billion Australian dollars in data centre investment is expected over the next decade. The government is saying: the money is welcome, but it comes with expectations. The social licence is not a formality. It is the price of entry.

I read that and thought: France has been saying this for years. It just uses different words.


The French habit of naming things


Living between two countries for a decade trains you to notice when the same argument is being made in different languages. The French call it l'intérêt général. The Australians call it national interest. The logic is identical: there are some domains where the state has an obligation to set the terms of engagement, because the consequences of getting it wrong are not recoverable by the market alone.

France's Doctrine Cloud, adopted in 2021, conditions access to sensitive state data on sovereign cloud certification (specifically, immunity from the American CLOUD Act and other extraterritorial instruments). The EU AI Act is the world's first binding legal framework on artificial intelligence. Mistral AI was built in part on the conviction that a serious economy cannot outsource its cognitive infrastructure to actors who answer to a foreign jurisdiction. Each of these decisions reflects the same underlying argument: digital infrastructure is strategic infrastructure. Not a commodity.

I spent fifteen years in France working inside national employer organisations, negotiating collective agreements, watching the French state intervene in markets with a confidence that never entirely made sense to me until I left. From Australia, looking back, it makes considerably more sense. The French are not wrong about the principle. They are sometimes wrong about the implementation. But the principle (that some things are too important to be left entirely to market logic) is one that democracies are slowly rediscovering everywhere.

Australia just added its name to the list.


What is already being built


This is not, it turns out, a hypothetical conversation. In February 2025, the French Treasury's regional economic service hosted a gathering of Australian AI companies and experts at the French Consul General's residence in Sydney, ahead of the Paris AI Summit. The agenda was explicitly about AI governance, digital sovereignty, and the role of public policy. That is not a casual seminar. That is a government signalling where it wants the relationship to go.

The FACET (the Franco-Australian Centre for Energy Transition) is already running joint AI projects with CSIRO, CEA, and Swinburne University, co-funded by both governments at roughly one million euros and one million Australian dollars per year until 2027. The work is applied and expanding. And since December 2025, France has a dedicated Ambassador for Digital and AI at the Quai d'Orsay: Clara Chappaz, whose brief covers international digital diplomacy and the global AI summit circuit. Her appointment signals genuine institutional intent, not diplomatic courtesy. The interlocutors exist. The bilateral mechanisms exist. What does not yet exist is a specific conversation on data sovereignty and AI infrastructure, the terrain where both countries are now asking the same question.

Is that conversation coming? I think so. The conditions for it have never been better.


The part that matters for business


If you are a board director, a chief executive, or a senior leader with operations touching both France and Australia, this convergence is not background noise. Data localisation requirements are tightening in both jurisdictions: for different reasons, through different mechanisms, but in the same direction. AI governance frameworks are becoming conditions of market access. The definition of "sensitive" data is expanding. And the question of where your infrastructure sits (whose jurisdiction it falls under, who can access it) is moving from IT policy to board-level risk.

The organisations that will navigate this well are not those with the most sophisticated compliance function. They are those led by people who understand both sides of the argument. Not just the regulatory layer, but the political logic underneath it. Why France insists on sovereign cloud. Why Australia now frames compute as a question of national interest. These are not bureaucratic eccentricities. They reflect a genuine and growing anxiety about what it means to let critical infrastructure be owned by actors who are, ultimately, accountable to someone else.

I have spent enough time in both countries to know this gap is real and frequently underestimated. I have also learnt (sometimes at some cost) that it is an opportunity for the organisations, and the individuals, who take the trouble to understand it before it becomes a problem.

Policy convergence does not automatically produce partnerships. It creates the conditions for them. Someone still has to be in the room, and someone has to have prepared for the meeting.

In fine, that has always been true. Wait and See!

The ocean is flat this morning. Good thinking weather.


Arnaud Couvreur 8 April 2026
Share this post
Australia's Moment: Why the Iran Strike Should Change How We Think About AI Governance