ANALYSIS: Franco-German digital sovereignty — the urgent tech bet defying Big Tech and Beijing

ANALYSIS: Franco-German digital sovereignty — the urgent tech bet defying Big Tech and Beijing

Introduction: November 2025 in Berlin, June 2026 in Brussels — a late but real awakening

From the Berlin summit to the European Commission

On November 18, 2025, in Berlin, the French and German governments jointly convened a Summit on European Digital Sovereignty. Emmanuel Macron and Friedrich Merz spoke side by side. Macron declared that Europe did not want to be a "customer" of large American corporations, nor a "vassal" of Washington or Beijing. "We want to create our own solutions," he said. Merz added: "Europe cannot abandon this field to others." The tone was set. The two largest economies in the eurozone agreed on the urgency of technological independence — no longer merely an aspiration but a strategic necessity.

The Berlin summit officially launched a joint Franco-German task force on digital sovereignty, mandated to produce a report in 2026. It identified three priority areas: artificial intelligence, data governance, and digital public infrastructure. Six months later, on June 3, 2026, the European Commission published its "Tech Sovereignty Package" — a legislative bundle including the Cloud and AI Development Act (CADA), a revision of the Chips Act, and a formal definition of "digital sovereignty" that the EU had, surprisingly, never codified before. The late but real awakening of a Europe that has decided to fight for its data.

The context: an alarming structural dependency

To understand why this technological bet is called "urgent," you need to grasp the scale of European dependency. Google, Microsoft, and Amazon control roughly 80 percent of the European cloud market, according to data cited by French Digital Minister Clara Chappaz in April 2025. Most enterprise software used by European governments — word processors, video conferencing, messaging systems — comes from American companies subject to the US Cloud Act, which allows American authorities to access data stored by these companies, including outside US territory. The generative AI models widely deployed in Europe — OpenAI's GPT-4, Google's Gemini, Anthropic's Claude — are all American.

What Paris and Berlin formally decided

The task force and its deliverables

The joint Franco-German task force launched in November 2025 produced in 2026 a set of recommendations that fed both countries' common positions in Brussels negotiations. The two governments agreed on "a shared definition of a European digital service" — a formula that may sound technocratic but is actually revolutionary: for the first time, Paris and Berlin set aside their historic disagreements on the definition of digital sovereignty to build a common position.

France, traditionally more interventionist, had tended to define digital sovereignty as the state's capacity to control data within its territory. Germany, more economically liberal, preferred a definition centered on the competitiveness of European companies and freedom of choice. The joint document produced what the Atlantic Council calls a "digital sovereignty triad": data control (who can process which data, and where), legal control (which legal frameworks govern the infrastructure), and provider nationality (where the headquarters and decision-making centers of tech companies are located). These three dimensions define a demanding standard that few non-European companies can satisfy.

Eurostack and its supporters

On April 9, 2025, the two governments had already formally backed the "Eurostack" initiative — a project aimed at creating a fully sovereign European technology infrastructure from end to end: semiconductors, cloud, platforms, AI, social networks. The German coalition document explicitly supports Eurostack. French Digital Minister Clara Chappaz confirmed that the French and German administrations were "in contact to build a joint roadmap." This Franco-German convergence on Eurostack gives the project a political legitimacy it lacked when it was carried only by industrial interest groups.

The American threat: Trump, digital tariffs, and the Cloud Act

Trump's pressure on European digital regulation

The geopolitical context for this Franco-German initiative is one of growing tension with the Trump administration. In a Truth Social post, Trump explicitly threatened additional tariffs on countries with "Digital Taxes, Digital Services Legislation, and Digital Markets regulations" — directly targeting European digital regulation. He described the Digital Markets Act (DMA), the Digital Services Act (DSA), and digital services legislation as barriers designed to "discriminate against American tech companies." The European Commission, France, and Germany unanimously rejected that interference.

This American pressure is not merely rhetorical. The Trump 2.0 administration pushed several major American companies to lobby for a weakening of European digital regulation within transatlantic trade negotiations. The fact that Paris and Berlin chose this moment to strengthen — rather than dilute under pressure — their common position on digital sovereignty is a clear political signal: European democracies intend to regulate their digital space independently of Washington's preferences.

The Cloud Act: the permanent legal threat

The US Cloud Act of 2018 is the most concrete legal manifestation of the European digital sovereignty problem. It allows American authorities to demand that American tech companies provide access to their clients' data, regardless of the country where that data is stored. For European governments, companies that store their data in Amazon, Microsoft, or Google data centers in the Netherlands or Ireland are theoretically exposed to unilateral American surveillance. This is why France developed its SecNumCloud certification, which requires, among other things, that no non-European entity hold more than 24 percent of a certified cloud company.

The Chinese threat: state AI and Beijing's tech champions

China's technology strategy and its internal coherence

If the American threat to European digital sovereignty is one of economic and legal dependency on commercial partners, the Chinese threat is of a different nature. China is deploying a state AI strategy at a scale that has no Western equivalent. Its tech companies — Huawei, ByteDance, Alibaba, Baidu — operate within a framework where the state can demand data access at any time, and where commercial interests are aligned with Beijing's strategic objectives. TikTok, owned by ByteDance, has accumulated data on hundreds of millions of Western users — a significant share of them young Europeans — under a legal regime that allows the Chinese government to access it.

More directly concerning for infrastructure: Huawei had supplied a significant proportion of European mobile infrastructure, notably in central and eastern European countries, before American restrictions and European decisions pushed for a progressive expulsion of that equipment. The European Commission is now preparing, according to Bloomberg, a legal requirement for the progressive elimination of Huawei and ZTE equipment from all European networks. Commission Vice President Henna Virkkunen is advocating turning the 2020 recommendation into a legal obligation. Thirteen member states have already taken restrictive measures.

Chinese AI and the risk of algorithmic dependency

China is investing heavily in AI models that could compete with American models in markets where Big Tech is perceived as a dependency risk. DeepSeek, developed by Chinese researchers in 2025, demonstrated that it was possible to build language models comparable to OpenAI's at a fraction of the cost — and without the most advanced Nvidia chips. If European governments or companies were to massively adopt Chinese AI models to avoid American dependency, they would be replacing one dependency with another, potentially more dangerous one. That is a risk the Franco-German digital sovereignty initiative explicitly seeks to prevent.

The Cloud and AI Development Act: the cornerstone of the puzzle

What CADA provides

The Cloud and AI Development Act (CADA), published on June 3, 2026 as part of the Commission's Tech Sovereignty Package, is the most ambitious piece of Europe's digital sovereignty strategy. It aims to address Europe's main deficit in the AI race: computing capacity. Europe has one of the most significant data center deficits of any advanced economy relative to its artificial intelligence needs. CADA seeks to remedy this by accelerating data center building permits, providing subsidized computing capacity to European AI startups, and establishing eligibility requirements for cloud providers in sensitive public procurement.

That last point is the most controversial: the eligibility requirements for public procurement could de facto exclude non-European providers from government contracts in sensitive areas — defense, domestic security, health, critical infrastructure. This is a major paradigm shift. Europe was moving from regulation to active European preference. The Commission was careful to phrase this in sufficiently technical language to avoid triggering an open trade war with Washington, but the intent is clear.

InvestAI and the 200 billion euros

In parallel, the European Commission launched InvestAI in 2025, an initiative to mobilize 200 billion euros in AI investment across Europe. This ambition must be set against the spending of American hyperscalers: Amazon, Microsoft, Google, and Meta announced for 2026 a combined total of 725 billion dollars in AI infrastructure spending — an increase of 77 percent over the 410 billion of 2025. This ratio reveals the scale of the challenge: Europe aims to mobilize 200 billion euros over several years, while four major American tech companies are spending 725 billion dollars in a single year. The gap is enormous. The European mobilization is necessary but structurally very far behind.

Europe's AI champions: what exists and what is missing

Mistral AI and the French hope

Europe is not without assets in the AI race. The French startup Mistral AI, founded in 2023 by former Google DeepMind and Meta researchers, developed high-performing language models — notably Mistral 7B and its successors — in open-source, giving them rapid international reach. Mistral's valuation reached several billion euros in 2024-2025, making it one of the rare European tech unicorns in generative AI. Experts from DFKI (the German Research Center for Artificial Intelligence) cited by DW note that providers like Mistral can develop "competitive advantages" on the dimensions of data sovereignty, regulatory compliance, transparency, and infrastructure control.

France has made Mistral AI a symbol of its digital industrial policy. Prime Minister Sébastien Lecornu announced in June 2026 that the DGSI, France's domestic intelligence service, was replacing the American firm Palantir's AI tools with those of the French company ChapsVision. Symbolically important, this decision is also an acknowledgment that the Palantir contracts created unacceptable "strategic dependencies." Lecornu also announced an investment of 655 million euros in AI for government services.

The glaring gaps

Despite these positive examples, Europe's structural weaknesses in AI remain significant. Europe has no hyperscaler — a company capable of running cloud infrastructure at the scale of Amazon, Microsoft, or Google. It has no cutting-edge chip manufacturer capable of producing GPUs equivalent to Nvidia's H100 or H200. The European semiconductor project — the Chips Act of 2023 with its target of 20 percent of global production by 2030 — is already behind on its objectives. Dutch company ASML, the world's only manufacturer of EUV lithography machines essential to advanced chip production, is a considerable European strength — but it is an equipment supplier, not a chip producer.

French policy in action: from Palantir to ChapsVision

A strong signal for European administrations

The decision by the French DGSI to replace Palantir's tools with those of ChapsVision — announced by Prime Minister Lecornu on June 16, 2026 — is both a political signal and a significant operational decision. Palantir, founded by Peter Thiel and backed by venture capital close to the Trump administration, is an American company whose data analytics tools are used by governments worldwide for surveillance and intelligence analysis. The DGSI contract had been extended through 2025, but the French government decided not to renew it.

Prime Minister Lecornu explained the decision in plain terms: "We cannot depend on tools designed by foreign entities. France must have its own resources." This language is sovereignty in its most concrete sense. It is all the more notable that ChapsVision, which takes over the contract, is also being chosen by the BfV — Germany's domestic counterintelligence service — creating a precedent for Franco-German cooperation in securing domestic intelligence tools. This is concrete digital sovereignty, not merely rhetorical.

SecNumCloud certification and "sovereign washing"

France developed the SecNumCloud certification of ANSSI (the National Agency for Information Systems Security) as a tool for qualifying "sovereign" cloud services usable by government agencies for sensitive data. This certification requires, among other things, that no single foreign entity hold more than 24 percent of the capital of the certified company, and no more than 39 percent in total by foreign entities. This capital structure rule makes it de facto impossible for Amazon, Microsoft, or Google to obtain certification — or to access it through joint ventures dressed up as European companies.

The Franco-German AI research center initiative

DFKI and Inria: the academic cooperation model

On the concrete research front, DFKI (German Research Center for Artificial Intelligence) and its French counterpart Inria (National Institute for Research in Digital Science and Technology) are about to sign an agreement to establish a Franco-German AI Center, according to DW in June 2026. Offices will open in Germany and France as early as July 2026, with operational launch in the fourth quarter of 2026. This center aims to coordinate fundamental and applied AI research, share computing resources, and train AI researchers based on a shared vision of European values — algorithmic transparency, privacy protection, non-discrimination.

This center will not produce GPT-5 or Gemini Ultra 2. It lacks the computing capacity for that. But it can contribute to training the next generation of European AI researchers, developing AI methods that use less computing power — an area where Europe has a potential advantage — and embedding in European research practices the principles of ethical governance that will become global standards. It is a long-term investment whose effects will be measured in decades, not quarters.

The European Parliament and the sovereign search engine

In parallel, the European Parliament made a decision as symbolic as it is practical: since June 4, 2026, the default search engine on Microsoft Edge and Mozilla Firefox browsers used by European institutions is Qwant, a French search engine. This choice, presented as a contribution to reducing dependency on foreign tools, is a statement of intent as much as a technical decision. Qwant is not Google — its result relevance remains lower by most comparative tests. But it stores its data in Europe, under European law, and without the advertising-profiling business model that defines Google search.

Structural obstacles: why it's so difficult

The problem of the fragmented single market

One of the paradoxes of European digital sovereignty is that its principal adversaries — American Big Tech — succeeded precisely by treating Europe as a single market of 450 million consumers, while European companies ran into 27 national markets with their local regulations, languages, and cultural preferences. Amazon builds a single AWS and sells it everywhere. A European cloud company must satisfy the regulatory requirements of 27 member states to reach the same scale. The absence of a genuine single market for digital services is structurally disadvantageous for European companies.

The Atlantic Council noted in its January 2026 report that Europe could move toward a form of "effective European control" as a qualifying criterion for sovereign cloud services — a formulation aligned with France's SecNumCloud certification but which, if adopted at European scale, would create significant entry barriers for non-European providers and could trigger a substantial American commercial response.

Talent and funding: the two bottlenecks

The DFKI identified four essential elements for Europe to become a real alternative to the United States in AI: data sovereignty, regulatory compliance, transparency, and infrastructure control. But these four elements are not enough without two prerequisites that remain problematic: talent and funding. Europe trains excellent AI researchers — but loses them massively to the United States through brain drain. The salaries offered by OpenAI, Google DeepMind, or Anthropic to a top-tier researcher — between 5 and 10 million dollars annually for the most sought-after profiles — are simply beyond the reach of any European company or institution.

The EU as a global regulator: the "Brussels Effect" in AI

The AI Act and its global reach

Europe holds a card that neither the United States nor China has: its capacity to impose its regulatory standards globally through what international lawyers call the "Brussels Effect." The GDPR (General Data Protection Regulation) of 2018 is the most striking example: a European law that transformed data protection practices worldwide, including in the United States, because global companies found it simpler to adopt a single high standard than to operate different systems by market. The European AI Act, which entered into force in 2024, aims to replicate that mechanism for AI.

If the AI Act succeeds in its regulatory bet — imposing on any company that wants to sell AI systems in Europe the obligation to comply with standards of transparency, non-discrimination, and auditability — it becomes de facto a global standard. OpenAI, Google, and Alibaba will all have to comply if they want access to the European market. This regulation does not directly create European champions. But it imposes constraints that favor companies whose culture is compatible with those values — and those companies are often European.

Trump versus the AI Act: the coming confrontation

The Trump administration views the European AI Act as a disguised regulatory barrier against American AI companies. White House representatives described European regulation as "anti-innovation" and included its relaxation in their demands during transatlantic trade negotiations. The European Union has so far refused to negotiate the substance of its digital regulation within trade agreements — a position defended notably by the European Parliament. But under pressure from European companies fearing American commercial retaliation, compromises are possible. Holding the line on digital regulation against Trump's pressure is one of the most important political tests facing the von der Leyen Commission.

European digital talent: brain drain to the US and retention strategies

The exodus of AI developers and engineers to Silicon Valley

One of the most documented weaknesses in the Franco-German digital sovereignty strategy is the brain drain toward the United States. Tens of thousands of engineers, AI researchers, and talented developers trained at French and German universities — the Grandes Écoles, the ENS, the CNRS, German technical universities (TU Munich, KIT) — choose every year to join American companies in San Francisco, Seattle, or New York. The reasons are well known: salaries two to three times higher, stock options, an ambition culture, access to the best tools and most ambitious projects.

This brain drain is not new, but it is accelerating with the AI race. American companies like Google DeepMind, Meta AI, and OpenAI actively recruit at European universities, often with offers that European startups or public laboratories cannot match. For the Franco-German digital sovereignty strategy, this reality creates a paradox: we invest in developing talent that foreign companies capture before it can contribute to building local capabilities. Resolving this paradox is a challenge as important as building infrastructure.

Retention measures: between tax incentives and corporate culture

France and Germany have implemented various mechanisms to try to retain their digital talent. In France, the jeune entreprise innovante status, the tax regime for inpatriates designed to attract foreign talent, and Inria's programs seek to create a competitive ecosystem. In Germany, initiatives like DFKI and industry-university partnerships attempt to maintain a critical mass of local researchers. The French "Make Our Planet Great Again" program — despite its borrowed Trumpian rhetoric — attracted dozens of foreign researchers in climate and AI.

These measures have produced tangible but insufficient results relative to the scale of the challenge. The European AI startup ecosystem is active — France ranks third globally for the number of AI startups by some measures. But the transition of promising startups to global-scale companies rarely happens in Europe. Mistral AI, considered the European champion of language models, has raised significant capital — and continues to face the temptation of an American acquisition. Retaining talent is a problem at the start of the chain. Keeping companies in Europe once they grow is another, equally difficult one.

The impact of the Iranian crisis: digital energy sovereignty

Data centers and energy sovereignty

The Strait of Hormuz crisis of 2026 added a new dimension to the urgency of digital sovereignty: energy sovereignty. Data centers are enormous energy consumers. The European Commission's plan to triple data center capacity by 2030 will carry substantial energy requirements. If that energy depends in part on Gulf hydrocarbons — whose price and availability can be disrupted at any time by the Hormuz crisis — digital sovereignty itself becomes vulnerable. Digital sovereignty policy and energy transition policy are not only linked by shared values. They are linked by a concrete infrastructural dependency.

The European Commission integrated into CADA a sustainability requirement for new data centers, with a four-tier energy rating system. The double objective is to reduce the climate impact of computing centers, and to reduce their dependence on fossil fuels whose supply can be disrupted by geopolitical crises. Building data centers powered by local renewable energy — solar, wind, nuclear depending on the country — is a component of digital sovereignty that too few commentators highlight adequately.

Nuclear power as a digital sovereignty lever

France, with its nuclear fleet, holds a competitive advantage for hosting high-energy-consumption data centers with a relatively low carbon footprint and independence from oil prices. This is a concrete dimension of digital sovereignty that French nuclear power offers to Europe. Several major American hyperscalers have in fact concluded nuclear energy supply agreements in France for their European computing centers — an energy dependency on France that partially reverses France's technological dependency on American cloud services. These complex mutual dependencies constitute the real fabric of technological geopolitics.

Franco-German cooperation: a history of digital ambivalence

Two visions, two industrial cultures: the structural frictions

Franco-German cooperation on digital sovereignty faces a reality that official announcements tend to downplay: the two countries do not share exactly the same vision of what it means. France has a tradition of technological dirigisme — the state as actor, investor, protector of national champions. The French approach to the "sovereign cloud", embodied in projects like Bleu (the joint venture between Orange, Capgemini, and Microsoft) or S3NS (with Thales and Google), is accepted even when it involves American partners — the essential thing being that data remains under French jurisdiction.

Germany has a different industrial tradition — rooted in the Mittelstand (industrial SMEs), complex public-private partnerships, and a historical wariness of national champions too close to the state (inherited from the history of the twentieth century). The German vision of digital sovereignty is more pragmatic about partnerships with large international companies, more focused on the security of industrial data, and less focused on building national technology champions. These differences are not insurmountable, but they slow the construction of a truly integrated common vision.

The flagship Franco-German projects: GAIA-X and its successors

GAIA-X — the Franco-German initiative for sovereign infrastructure cloud launched in 2020 — is both a symbol of common ambition and of its difficulties. The initiative, extended to other EU members, aimed to create an interoperable and sovereign European data space. It produced standards, certifications, a governance organization. It has not yet produced a credible cloud alternative to AWS, Azure, or Google Cloud. The temptation to define "sovereignty" criteria broadly enough to include the European subsidiaries of American Big Tech diluted the original ambition.

The successors and complements to GAIA-X — the sectoral "data spaces" in mobility, energy, health, and manufacturing — have better chances of success because they are more concrete and more tied to specific industrial needs. Digital sovereignty may be won less through broad generic architectures than through mastery of critical data in specific strategic sectors. A Europe that controls its health data, its automotive data, its energy data — even if it uses partially foreign cloud infrastructure — will be more sovereign than a Europe that theorized a perfectly sovereign cloud but never built it.

The regulatory calendar: what's coming in 2026-2027

The next legislative steps

The schedule for European digital sovereignty over the next 18 months is packed. The Cloud and AI Development Act must be negotiated between the Commission, Parliament, and Council before entering into force. The Digital Fairness Act, announced for mid-2026, will introduce protections for digital consumers — including protections against addictive design, targeted advertising for minors, and manipulative pricing practices. The ICT Supply Chain Toolbox, which will allow the coordinated identification and exclusion of suppliers deemed security risks (such as Huawei) across member states, is also expected.

These pieces of legislation, if they reach their ambitious form, will constitute a regulatory arsenal without global equivalent for digital governance. Their actual enforcement — requiring considerable human and technical resources to monitor global companies — will be the real test. The AI Act entered into force in 2024. In 2026, the first enforcement investigations are only just beginning. The challenge is not only to write the rule. It is to enforce it against companies whose revenues exceed the GDP of many member states.

France and Germany as engines — and their limits

Paris and Berlin agreeing on digital sovereignty is indispensable — but not sufficient. The other 25 member states must be brought on board, and some hold very different positions. Countries in central and eastern Europe, more economically dependent on the United States and culturally more wary of Franco-German regulatory ambitions, are often reluctant. The Netherlands, which hosts a disproportionate share of Big Tech's European data centers, has immediate economic interests that can conflict with a policy of European preference in public procurement. Building a coherent European digital sovereignty policy is also a domestic politics project within the Union itself.

Conclusion: a necessary bet, a long road ahead

What the Franco-German initiative has already accomplished

The balance sheet of the Franco-German digital sovereignty initiative in June 2026 is mixed. In the achievements column: a common definition of digital sovereignty that fed the European position; an ambitious Tech Sovereignty Package adopted by the Commission; concrete decisions such as replacing Palantir with ChapsVision in France; the launch of a Franco-German AI research center; and a firm stance maintained against American pressure on the AI Act. This is not nothing. It is, even by the standards of European institutional history, a remarkably rapid advance.

In the remaining challenges column: the absence of European industrial champions capable of competing with Big Tech at global scale; the investment gap (200 billion euros over several years against 725 billion dollars American in a single year); the talent drain toward American laboratories; intra-European divisions between member states; and constant pressure from Washington for compromises on digital regulation. This road will be long. The urgency is real. But honesty requires acknowledging that Europe is starting from far behind.

Why this bet is still worth making

The question is not whether Europe can catch up with the United States or China in the AI race in the coming years. It cannot. The question is whether it can build sufficient digital capacity to avoid being entirely dependent on foreign powers for critical functions of its economy and security. And whether it can impose regulatory standards that humanize AI globally — ensuring that the algorithms increasingly governing our lives respect principles of transparency and non-discrimination. Both objectives are within reach. They are worth the bet. They justify the Franco-German urgency. The world cannot afford to have only non-liberal democracies define the AI of the future.

By Maxime Marquette, columnist

Columnist's transparency note

My biases and my perspective

I am in favor of building European digital sovereignty, which orients my view on this subject. I believe that excessive technological dependency on non-European actors — whether American or Chinese — is a real strategic risk. This conviction may lead me to be more indulgent toward European ambitions and more critical of Big Tech and Chinese companies than a neutral observer would be. I own that.

What I don't know

I do not know the precise content of all joint Franco-German documents produced by the task force, some of which were not published in full. The AI investment figures come from reliable secondary sources — Bloomberg, Financial Times, Atlantic Council reports — but may vary depending on the definitions used. The actual effectiveness of the Cloud and AI Development Act will depend on its enforcement, which I cannot prejudge.

Sources

Primary sources

DW — Merz and Macron at the Berlin Digital Sovereignty Summit — November 18, 2025

Le Monde — EU publishes its technological sovereignty plan — June 3, 2026

The Guardian — France replaces Palantir with ChapsVision — June 16, 2026

Atlantic Council — Digital sovereignty: Europe's declaration of independence — January 2026

Secondary sources

Euractiv — Germany and France support Eurostack — April 2025

DW — Can Europe build its own AI heavyweights? — June 19, 2026

Atlantic Council — The digital sovereignty triad — May 21, 2026

Le Monde — Europe's painful awakening on digital sovereignty — June 5, 2026