On 2 November 2026, the German government published the AI Task Force's findings report, "Germany as an AI Nation by 2030," at the Digital Summit in Berlin. According to the Federal Government's summit page, it sets measurable 2030 targets and bundles 18 cross-departmental measures. At least 80 per cent of large enterprises and 55 per cent of small and medium-sized enterprises are expected to use AI. AI-generated output should exceed ten per cent of the economy. And 90 per cent of federal authorities should use at least one AI application for core tasks.
The report is, first, an adoption plan. Whether it becomes a trade problem depends on choices it has not yet made.
The strongest case for the sovereignty framing
The sovereignty argument deserves a fair hearing. Europe's AI stack, from chips to models to hyperscale cloud, leans heavily on a handful of US firms. A government that routes core administrative functions through a foreign supplier takes on real exposure to export controls, sanctions, outages and commercial leverage. Chancellor Merz said at the summit that "Europa darf die technologische Gestaltungsmacht nicht anderen überlassen" (Europe must not leave technological shaping power to others). He also said partnerships with like-minded states are essential to "reduce one-sided dependencies," according to the transcript of his press conference. Digital Minister Karsten Wildberger described the plan as "a reliable framework to promote greater digital sovereignty and competitiveness." Those are legitimate concerns, and a strategy that ignored them would be naive.
What the plan actually commits to
The most concrete commitment is compute. Merz described a national data-centre strategy "to quadruple our computing capacity by 2030," and Wildberger said at least one European AI gigafactory should operate in Germany. Trade press adds detail. ZfK reports that AI data-centre capacity is meant to grow 300 per cent against 2025. The time from application to commissioning of a data centre is to be capped at two and a half years. Access to frontier models is to be secured through "strategic partnerships and joint European procurement."
This is the pro-innovation core of the plan, and it is sound. The binding constraint on German AI adoption is not a lack of rules. It is slow permitting, expensive power and scarce compute. A permitting deadline does more for sovereignty than any procurement preference could.
The same ZfK report flags the gaps. The task force does not quantify the financing needed. It relies on private investment. It says it does not prejudge budget procedures. It also gives no figure for future electricity demand. ZfK sets this against private AI investment of about $286 billion in the US in 2025 versus $3.9 billion in Germany. The targets, in other words, outrun the funding.
Where the trade tension sits
Nothing in the sources I could verify shows the report excluding non-EU suppliers. Merz's remarks, as published, do not name US providers at all. The tension is structural rather than textual. A plan this reliant on private capital will need foreign hyperscalers, chip makers and model developers to build much of the capacity it targets. Meanwhile, the wider European sovereignty agenda is moving toward preference rules. A Steptoe analysis from May 2026 describes a proposed Cloud and AI Development Act that would seek to triple "sovereign" EU cloud capacity within seven years. It could bring "Buy European" rules for the public sector. The same analysis warns of higher compliance costs and lost business for American firms.
That matters because Washington is already treating European digital regulation as a trade issue. A CSIS analysis from March 2026 finds the US has sought anti-digital-regulation clauses in bilateral trade agreements with at least nine countries. It also notes that the Commission has insisted the Digital Markets Act and Digital Services Act are not part of trade negotiations.
The 90 per cent federal-authority target gives the issue a specific shape. If those government deployments are procured under sovereignty-first criteria that exclude or disadvantage foreign suppliers, public procurement becomes the first place a dispute over market access arises. My reading is that this risk is real but still avoidable. It is my inference, not a finding of the report.
A proportionate way to read sovereignty
There is a defensible version and a counterproductive one. The defensible version defines sovereignty by outcomes: portability of data and workloads, auditability, exit rights, multi-vendor architectures, and enforceable security requirements that any supplier can meet wherever it is headquartered. The counterproductive version defines it by nationality of ownership. That version raises costs for the very SMEs the 55 per cent target depends on. It narrows the supplier pool during a period when capacity is scarce. And it invites retaliation that would hit German exporters, whose economy depends on open markets.
The evidence-based test is simple. Government should state its security and resilience requirements in neutral, technical terms. Where a requirement can be met by a non-EU supplier under verifiable controls, the procurement rules should allow it. Where it cannot, officials should say why in public.
What to watch
Three things will show which version Berlin chooses. The first is whether the budget process attaches money to the 18 measures; Wildberger has said the government will review progress and report publicly. The second is whether the data-centre permitting cap becomes law rather than aspiration. The third is whether the procurement criteria for the federal AI deployments are published as neutral technical requirements or as origin-based preferences.
Germany's targets are ambitious, and ambition is not the problem. The risk is that the sovereignty label slides from building capacity to excluding competitors. That would slow adoption, raise prices and hand Washington a clean example of the discrimination it alleges.