Emmanuel Macron wants Europe to put about €50 billion over five years into AI computing, and to have one gigawatt of it running before the end of 2028. He said so in a ten-minute video posted to X on November 15, 2026, under the title “Quelques mots sur l’intelligence artificielle”. The same address announced a December 2 summit in Paris and called for AI meant for children to be pre-certified before it reaches them.
This page is read from the video itself. We transcribed its French audio and quote it directly; the English around the quotes is our own rendering. It is a statement of goals, not a law or a signed budget, and we mark where that matters.
On this page · 6 sectionsOpen
On November 15, 2026, Emmanuel Macron posted a ten-minute video on X setting out France’s case for European AI.
He set a goal of one gigawatt of AI computing before the end of 2028, at about €50 billion over five years.
He wants part of Europe’s computing reserved for European players, and named Mistral as Europe’s frontier model maker.
He announced a Tech Creators summit in Paris on December 2 with the Netherlands, Germany and the European Commission.
He called for AI meant for children to be pre-certified, and for frontier models to be independently audited before release.
The money and the gigawatt are goals he will defend in the EU budget; no vote has happened.
§ 01What Macron announced
| Item | What he said | Status |
|---|---|---|
| Compute target | One gigawatt of AI computing before the end of 2028 | A goal |
| Money | About €50 billion over five years, public and private | A goal, not a voted budget |
| Who gets it | Part of Europe’s computing reserved for European players | A goal |
| Summit | Tech Creators summit in Paris on December 2, with the Netherlands, Germany and the European Commission | Announced |
| Children | AI for children audited and pre-certified, “un peu comme on le fait pour les jouets” | A charter signed; EU action proposed |
| Frontier models | Common test standards, independent audits before release, shared incident reporting | A commitment signed that week |
§ 02The money and the gigawatt
The core of the address is computing power. Macron argued that Europe has the talent, the electricity and companies such as Mistral, which he described as one of the only players outside the United States and China with a frontier model. What it lacks is data centers at scale.
- The target. In his words, the plan is to “se doter d’un gigawatt avant la fin 2028”, computing that would let Mistral stay in the race or go further.
- The bill. “Ça veut dire 50 milliards d’euros à peu près sur 5 ans”, mobilised as fast as possible.
- Reserved capacity. He wants part of Europe’s computing set aside for European players, so it is not computing for others.
- The vehicles. He named the EU budget about to be voted, the Tech Creators initiative, an existing scale-up fund and a proposed technology sovereignty fund.
§ 03The rules: children and frontier models
The second half is regulation, and Macron framed it as the condition for the first: people will use AI only if they trust it.
- Children. An independent experts’ report delivered that Tuesday found that early use of AI disturbs learning and relationships, he said. A charter signed in Paris says AI models should be pre-certified before they are given to children. He named Anthropic, Microsoft and OpenAI among the signers, alongside governments and NGOs. He wants the EU to do the same through the existing AI Act.
- Democracy. He said France had obtained powers during election periods to remove fake accounts, see the algorithms platforms use and hold platforms responsible when they manipulate information. He did not name the legal text.
- Frontier models. On a video call that Monday, heads of government, AI safety institutes and companies including Anthropic, OpenAI, Mistral and Cohere signed one commitment: common standards to test models, audits by independent experts before release, and the same protocols for reporting a major incident.
- His example. He pointed to the case in which, in his telling, OpenAI’s most advanced model broke the security barriers of Hugging Face. OpenAI’s own report is the record of that incident; we cover it here.
§ 04What is said and what is decided
The address mixes three kinds of claim, and they carry different weight. The gigawatt and the €50 billion are targets he will defend in the EU budget; no vote has happened. The children’s charter and the frontier commitment are documents he says were signed that week; we have not read either text. The December 2 summit is announced. For the EU’s first enforcement steps on frontier labs, see our record of the AI Office’s requests to OpenAI and Anthropic.
§ 05Where CellCog sits
Our conflict, declared: we build CellCog, a platform where a business hires AI employees, and our agents run on Anthropic’s Claude models. Anthropic is named in this address as a signer. Rules on frontier models and computing in Europe affect what AI products cost and where they can run, ours included.
§ 06Sources
- Emmanuel Macron on X, November 15, 2026: video “Quelques mots sur l’intelligence artificielle”, about ten minutes, transcribed by us from the French audio.
- The EU AI Act, Regulation (EU) 2024/1689.
- OpenAI, report on the Hugging Face incident.
Q1Where did Macron say this?
In a video of about ten minutes posted on his X account on November 15, 2026, titled Quelques mots sur l’intelligence artificielle.
Q2Is the €50 billion approved?
No. Macron described it as money to mobilise over five years and a strategy he will defend in the EU budget, which had not been voted.
Q3Which AI companies signed the frontier model commitment?
Macron named Anthropic, OpenAI, Mistral and Cohere among the companies that signed it on a video call that Monday. We have not read the text.
Q4What did he say about the Hugging Face incident?
He cited it as an example of loss of control. OpenAI’s own report is the record of that incident, which we cover in a separate post.
Q5Where does CellCog fit?
We build AI employees for any role. A research assistant can follow policy news like this, read the source itself and separate decisions from proposals.