AI Report: DACH Mid-Market Action Items Before Summer
On April 17, 2026, the working groups of the Franco-German AI Executives’ Dialogue submitted their final report to the French and German governments. The report, led by Inria, IMT, and Fraunhofer, identifies three industrial levers for a sovereign AI roadmap in Europe: joint compute sharing, data sovereignty via GAIA-X nodes, and use case prioritization along industrial value chains. For mid-sized companies in the DACH region, the real question is not what Brussels will do with it, but how to make the recommendations visible in their own numbers by summer.
- Report submitted on 17.04.2026 to the German Federal Ministry for Economic Affairs and Climate Action (BMWE) and the French Ministry of Economy; three mandatory levers: compute sharing, GAIA-X data space, and industrial use case prioritization.
- The report is being fed into the IPCEI AI, making it immediately relevant for funding and procurement for companies in Germany and France.
- For mid-sized companies in the DACH region, this means three operational tasks: establishing a “data owner” role, clarifying GAIA-X anchor points, and prioritizing AI use case portfolios.
- Companies that complete these three tasks by summer will be among the first wave of projects eligible for funding and will have a head start in procurement-critical tenders.
- The interfaces to the EU AI Act and NIS2 implementation are explicitly addressed in the report, facilitating legally secure integration.
What the April 17 report specifically demands
What is the Franco-German AI Executives’ Dialogue? The dialogue is an industry-driven initiative to coordinate German-French AI industrial policy, launched in January 2025 in Berlin. The initiative is led by Inria (France), IMT (France), Fraunhofer (Germany), and executive delegates from the largest industrial companies in both countries. The result is a consensus report updated three to five times a year, serving as input for national funding policy and the EU’s IPCEI AI strategy. The version submitted on 17.04.2026 is the first to include concrete operational recommendations rather than just declarations of intent.
The report outlines three work packages that directly affect mid-sized companies. Firstly, compute sharing through a joint funding framework for SMEs with access to sovereign high-performance infrastructure (AI Gigafactories, Jupiter in Germany, Jules Verne in France). Secondly, data space connection via GAIA-X nodes as a binding interface for industrial data ecosystems. Thirdly, use case prioritization along defined industrial AI domains (manufacturing, mobility, energy, health, environment). While the details are technical, they translate into three concrete questions for mid-sized companies.
Question one: Who within the company will take on the role of data owner for the next GAIA-X connection? Question two: In which industrial AI domain does the company generate its main revenue; which use cases there are eligible for funding? Question three: How will compute sharing be operationalized when workloads move between on-premises, hyperscalers, and sovereign data centers? Companies that answer these three questions in writing will have an operational anchor for the upcoming funding rounds.
Why the document has a stronger impact than expected
Many federal policy documents rarely have an immediate impact on mid-sized companies. This report is different because it feeds directly into the IPCEI AI program. IPCEI AI is the vehicle through which large funding programs are implemented in Germany and France; the aid rules allow for significantly more generous support quotas on an IPCEI basis than classic individual funding. In concrete terms, this means that companies aiming to secure funding for projects in the next twelve months should be familiar with the report’s recommendations and reference them cleanly in their own application architecture.
Three numbers from the report that stick with the board
Any good web architecture can be sketched on a Post-It. The same applies to a mid-sized business AI roadmap: if it doesn’t fit on half a page, prioritization is lacking.
The three operational homework tasks in detail
Homework 1: Establish a data owner role for GAIA-X connectivity
GAIA-X nodes require a clear responsibility structure per data category. SMEs must designate a role that controls data ownership and technical connectivity: a data owner with a mandate over production, supply chain, and engineering data, not just IT technical data assets. This role is ideally located at the executive level or as a dedicated mandate directly below it. We observe in the first integration projects that companies that formally designate this role are GAIA-X-ready within three months. Companies that “just leave it to IT” are still stuck in data catalog discussions after nine months.
The operational connection to a GAIA-X node can now be handled via commercial service providers (Teralab, International Data Spaces, Catena-X for mobility). The technical integration is solved, but the organizational one is not. The pragmatism check: if you can’t name three data categories in your company that can be audited and shared externally, you haven’t even started your GAIA-X hour.
Homework 2: Prioritize use case portfolio along the five industrial AI domains
The report prioritizes five domains: manufacturing, mobility, energy, health, environment. SMEs should categorize their existing AI use cases into these domains and prioritize the top five that have the greatest leverage in their business model. This sounds trivial but breaks with the usual use case discussion, which prioritizes AI projects primarily based on technical feasibility. The industrial policy line is: value contribution plus strategic fit with the funding landscape. A use case that cleanly falls into one of the five domains and delivers a genuine efficiency or quality advantage has a higher chance of co-financing in 2026 than a technically sophisticated standard optimization case.
Task 3: Compute-Sharing Strategy for the Next 24 Months
Compute-sharing is the most operational of the three levers but is also the most misunderstood. It does not mean that medium-sized companies gain access to a national supercomputing center. Instead, it means that funded AI workloads should run via defined hubs (Jupiter, Jules Verne, the announced AI gigafactories). For your own company, this means: assess which workloads are latency-critical (and thus remain on-premises), which are suitable for hyperscalers (and remain there), and which can be outsourced to sovereign data centers in the future. Those who decide this by summer will be part of the first wave of funded pilot projects.
The compute-sharing strategy is the one that involves the board of directors the most. It affects investment decisions, existing hyperscaler contracts, and possible exit strategies. A pragmatic interim step: over the next three months, create a compute inventory list that records the category (latency-critical, sensitive, standardizable) for each workload. This list will be more valuable than any executive summary in the funding applications of the coming 18 months.
Practical Check: What a Medium-Sized Company Project Looks Like in Reality
From our consulting mandates of the last few months, three project archetypes can be derived that fit neatly into the report logic. Archetype one, “Manufacturing Quality”: A mechanical engineering medium-sized company that shares predictive quality models with supply chain partners via a GAIA-X connection. Archetype two, “Energy Optimization”: An energy supplier or industrial company that operates load forecasting and storage management models on sovereign infrastructure. Archetype three, “Health Compliance”: A medical technology medium-sized company that shares training data in a data-sovereign manner with research partners and deploys models in regulated environments in an audit-proof manner. All three archetypes are prioritized in the current IPCEI-AI design.
The realistic project duration from the first workshop to productive rollout is nine to 18 months, depending on the data governance maturity in-house. The total project cost ranges between 800,000 and 4 million euros, with the funding quota typically between 40 and 60 percent within the IPCEI framework. For your own budget planning, this means: those who start now should plan a net investment of 300,000 to 2 million euros for a pilot project with a clear reference to the report.
A pragmatic stumbling block we see in almost all projects: the coordination between the corporate parent (if present) and the medium-sized unit for budget co-financing. Often, only the medium-sized part is eligible for funding. Those who set up clear governance from the start (project steering with clearly defined decision-making bodies and signing authorities) save themselves two to three months in the application phase. Those who ignore this end up in escalation rounds that are severely penalized in the IPCEI timeline.
Interfaces to EU AI Act and NIS2
The report from April 17, 2026, is regulatory compatible in two directions. Firstly, with the EU AI Act: high-risk systems in industrial contexts (safety-relevant manufacturing robotics, energy infrastructure) are addressed by the report’s recommendations, reducing the coordination effort between industrial policy and the competencies of the responsible supervisory authorities. Secondly, with NIS2: data spaces and compute-sharing scenarios are classified as critical infrastructure where they are relevant for energy supply or healthcare. The operational consequence: those who complete the three tasks cleanly document NIS2 readiness and EU AI Act mappings in parallel, without it becoming double work.
Third Perspective: What Trade Unions and Works Councils Want to See
One dimension that has received little attention in the public discussion of the report is the role of works councils and trade unions in the implementation of industrial AI. In Germany, the involvement of the works council in AI introduction is mandatory under the Works Constitution Act, while in France, similar processes are handled by the CSE (Comité Social et Économique). SMEs that only involve employee representatives in their project governance at the rollout stage often end up in escalations with project delays of two to six months. The operational consequence: The data owner role and the use-case portfolio should be discussed with the works council or CSE in the first quarter of the project, not the last.
The report addresses this dimension on a meta-level (“industrial acceptance”), but does not provide model agreements or governance templates. It is the task of employer associations and trade unions to fill this gap in the coming months. For SMEs, this means: Establish your own agreements based on DGB/IG-Metall models or analogous FO/CFDT models before submitting the first funding application. A good project management should plan at least four weeks for this and involve the HR department early on. Those who underestimate this dimension may produce a nice application, but will fail to gain operational acceptance as soon as the notice arrives and implementation begins.
What the Report Does Not Achieve
Three criticisms from our conversations with SME customers regarding the report. Firstly, concrete funding volumes per SME project are still missing; these will be addressed in the next IPCEI calls. Secondly, the role of SME associations (BDI, BVMW, MEDEF in France) in the implementation is vaguely formulated. Thirdly, the IT security dimension is mentioned but not elaborated. Those who have these three gaps on their radar can smoothly integrate the report’s recommendations into their own roadmap without relying on promises the report does not make.
Frequently Asked Questions
Does the report only apply to companies with a French subsidiary?
No. The report is aimed at the entire European industrial mid-market, with a clear focus on German-French value chains. Mid-sized companies with partners or customers in France have a strategic advantage, but independent DACH participation in IPCEI-AI projects is explicitly provided for.
What are realistic funding amounts for mid-market pilot projects?
Based on the IPCEI-AI framework guidelines, we expect funding rates between 30 and 60 percent of direct project costs, depending on company size and degree of innovation. For a typical mid-market pilot project (500,000 to 2 million Euro project volume), this means a subsidy share between 150,000 and 1.2 million Euro. The specific rates will be defined in the respective national tenders.
Does a company have to participate in GAIA-X to receive funding?
Not formally, but increasingly operationally. GAIA-X integration significantly lowers the evaluation hurdles in tenders because sovereign data governance is considered a plus. Companies working without GAIA-X must demonstrate an equivalent sovereign data architecture, which is usually more complex.
How does the report relate to BMWE funding of AI in German companies?
The existing BMWE programs (Mittelstand-Digital-Zentren, KI-Innovationswettbewerb) will not be replaced but supplemented. The report provides a strategic framework; the BMWE programs are expected to include the report’s recommendations as evaluation criteria in upcoming tenders.
By when should companies complete the three tasks?
Our pragmatic target: The data owner role should be formally appointed by the end of June 2026, use case prioritization completed by the end of July, and the compute sharing strategy documented by mid-September. Companies adhering to this timeline will be positioned for the autumn tenders. Those starting later can aim for the spring tenders in 2027.
Who coordinates this operationally in Germany?
At the federal level, the BMWE with support from Fraunhofer; at the state level, the respective ministries of economic affairs. For mid-sized companies, the Chambers of Industry and Commerce (IHKs) and the respective Mittelstand-Digital-Zentren are the first points of contact to clarify specific tenders and application windows.
Source of title image: Pexels / Dom J (px:355948)
Network: Continue reading on MyBusinessFuture
- Our assessment of AI data readiness in the mid-market and the five tasks before the productive agent
- Analysis of the 80 percent AI failure rate and the structural question behind it
- Report review of the Deloitte State of AI 2026 Enterprise and the execution gap