In February 2026, the Israeli AI investor Eli David posted that Macron was a “clown” for trying to make France an AI leader with €30 million, a reference to a French grant for attracting about forty international researchers. The line went viral. Macron clapped back on X, pointing to the €109 billion in private AI investment pledged at the Paris summit he had hosted a year earlier, and for a brief, glorious moment a researcher grant was the funniest thing to happen in European industrial policy.

About three months later, a French-led consortium called AION is bidding for roughly €10 billion from the European Commission’s AI gigafactory fund. The consortium includes Scaleway (Xavier Niel’s Iliad Group), Capgemini, Orange, EDF, Ardian, Bull, and Artefact. Iliad alone is committing €4 billion through Scaleway. The project would deploy 288,000 H100-equivalent GPUs at roughly 200 megawatts, with a first phase near 100 megawatts and a target of one gigawatt, doubling France’s total compute capacity.

The gap between €30 million and €10 billion is the gap between a meme and a procurement bid. One is a talking point; the other requires banks, industrial partners, site plans, and a power source, and AION has all four.

The power source is the interesting part. EDF, the French state-controlled utility that operates the largest nuclear fleet in Europe, is in the consortium. France generates roughly 70 percent of its electricity from nuclear, which means low-carbon, dispatchable baseload power at rates below the European average. Every other EU member bidding for gigafactory funds has to solve the energy question. France solved it decades ago for unrelated reasons and is now collecting the dividend.

This is not a minor detail. AI datacenters are power-constrained before they are GPU-constrained. Mistral chose the Paris region for its first owned datacenter. AION is making the same bet at twenty times the scale. Nuclear baseload is France’s genuine structural advantage in a competition where electricity cost and carbon intensity are becoming selection criteria for where to build.

The EU framework behind AION is called InvestAI, launched by the Commission with a target of €200 billion in combined public and private capital: €50 billion public, €150 billion private. Of the public portion, €20 billion is earmarked for up to five AI gigafactories across member states. Seventy-six consortia submitted expressions of interest totaling €230 billion, which is either a sign of genuine demand or a sign that everyone in Europe filed an expression of interest to see what would happen.

The obligatory context: five US hyperscalers have committed between $660 billion and $690 billion in AI capital expenditure for 2026 alone. The EU’s entire InvestAI target, public and private combined over multiple years, is €200 billion. The United States has produced roughly 40 foundation models at frontier scale. China has produced 15. Europe has produced three. The independent European frontier lab is already consolidating into transatlantic partnerships rather than scaling alone.

So the correction is real but proportional. Europe went from a memed press conference to a ten-figure infrastructure bid backed by an actual industrial consortium with an actual power source. That is genuine progress, even if it amounts, in absolute terms, to roughly 1.5 percent of what the US hyperscaler class is spending this year. The shift is from performative to structural, not from small to large. AION, if funded, would make France the largest compute market in continental Europe. It would not make Europe competitive with Virginia.

The nuclear angle is the part worth watching. Cheap, low-carbon, dispatchable power is the one input to AI infrastructure that cannot be bought with venture capital or replicated by executive order. France has it, almost nobody else in Europe does, and if the AI buildout follows the energy, the buildout follows France.

Sources: Reuters, Euronews, European Commission

–
By the Control Plane Editorial Team