On June 29, 2026, French outlet La Voix de France detailed a government plan, announced June 16, to reallocate €655 million from the France 2030 program into AI infrastructure, research and public-sector deployment. The package includes rolling out a sovereign AI assistant to about one million civil servants and replacing the DGSI’s Palantir system with French vendor ChapsVision.
This article aggregates reporting from 1 news source. The TL;DR is AI-generated from original reporting. Race to AGI's analysis provides editorial context on implications for AGI development.
France is sharpening its AI sovereignty play by putting real money and procurement decisions behind the rhetoric. The €655 million reallocation from France 2030 into AI will not rival private hyperscaler capex, but it is targeted: national compute, open public data platforms, and a sovereign assistant deployed to a million public employees. Combined with a high‑profile switch in the intelligence services from Palantir to French group ChapsVision, Paris is clearly trying to build a domestic AI stack that can coexist with, and sometimes displace, US vendors. ([lavoixdefrance.fr](https://www.lavoixdefrance.fr/actualites/655-millions-deuros-pour-lia-francaise-lecornu-mise-sur-la-souverainete-un-million-dagents-publics-equipes-8103/))
For the AGI race, France’s move matters less as raw capacity and more as a template for mid‑sized powers. By seeding demand for local models (e.g., Mistral AI) in the state and building shared data and compute infrastructure, the government is trying to ensure French and European actors remain relevant in an ecosystem dominated by US and increasingly Chinese giants. The public‑sector assistant also creates a huge in‑country testbed to harden French models on real workflows.
This approach may not accelerate the absolute frontier of AGI, but it shapes who gets to commercialise and govern it. A European bloc with sovereign options in security, translation and admin tools will have more leverage in future standards talks on safety, watermarking and cross‑border data access.


