France has drafted a position paper for the Council, with co-signatories still to be named, titled “Strengthening the CADA to Promote Value Creation within the European Union in the digital sector”, which was leaked and published by Euractiv. It wants a stronger European preference in the Cloud and AI Development Act (CADA), proposed by the Commission in June. It builds on the Franco-German joint paper on digital sovereignty of 17 June.
EuroStack backs its direction: public buying is the strongest tool Europe has to grow its own digital industry. As drafted, though, the paper would reward suppliers who move jobs to Europe more reliably than suppliers who give Europe control. Seven amendments would fix that.
What the paper gets right
- It gives the European criterion real weight. The Commission’s Article 32 lets buyers score a tenderer’s European contribution, then Article 32(2)(d) declares that score “ancillary and not decisive in the award of the contract”. The recitals cap it at 15 points out of 120. A criterion that can never change the winner changes nothing. France asks for a floor of 20% of the total score.
- It widens the scope. Article 32 covers only “innovative AI and cloud solutions”, which the text never defines. France wants the whole digital sector in, and public funding too.
- Two of its metrics ask the right question. Two thirds of the R&D on the core technology in the EU; engineers in the EU who maintain the service, fix it when it breaks and hold the source code. Both ask who can change the product.
Jobs in Europe, decisions abroad
The other metrics count where people work: half the workforce, half the R&D staff and half the subcontracting spend in the EU, plus a registered head office. France states the intent openly: market access “conditional on effective transfers of skills and know-how”, with non-European firms anchoring R&D in Europe, “leveraging external competencies rather than excluding them”.
Jobs in Europe matter. A supplier can still employ thousands of people here and decide alone, from elsewhere, next year’s price, which features stay, when updates stop and which law reaches the data. A European subsidiary reselling its foreign parent’s technology passes the head-office and headcount tests while the parent keeps every one of those decisions. Several “sovereign cloud” offers work exactly this way.
The paper never says which company the metrics apply to. Measured on a global group, a 50% EU workforce excludes most large non-European vendors. Measured on a local subsidiary, it excludes almost no one. The text has to make that choice.
The metric that would settle the control question most directly sits empty. The last row of the table reads “Open-source core software and active contribution”; its threshold box is blank. A buyer checks a licence in five minutes at no cost. An open-source licence lets the buyer hire someone else to maintain the software if the supplier fails or changes course; European contribution to the project makes that option real.
Nothing in the list looks at the cost of leaving
The Franco-German joint paper had six dimensions of sovereignty. The French paper builds on one of them, value creation. The joint paper also asked that systems “enable a change of provider and technology within reasonable time and financial expenditure”, and added that “open source solutions can and should play an important role in this regard”. That dimension has disappeared from the new paper.
A preference for European value added only bites on new contracts. Most public digital spending renews existing ones, with customers who cannot realistically leave: fewer than one European cloud or software customer in a hundred changes provider in a year. A supplier whose customers cannot leave can raise prices up to what leaving would cost them. French audit and parliamentary reports published this year record increases of 30% to 46% by several large vendors on the French state. A 20% weighting on new tenders leaves all of that untouched.
CADA has the same blind spot elsewhere. Article 30(4)© lets a buyer set the sovereignty framework aside whenever applying it “would require the contracting authority to procure services at disproportionate cost”, a term the text never defines. A comparison of this year’s prices that leaves out the cost of leaving the incumbent will find the European offer disproportionate almost every time. Buyers should compare costs over the whole life of the contract, exit included, and publish every use of the exception.
EuroStack therefore pairs European preference with interoperability. Open interfaces, portable data and a tested exit plan let a buyer move its systems to European providers piece by piece, each move making the next cheaper. Without them, European suppliers compete for the thin slice of demand that is new each year, however generous the preference.
A menu of self-declared metrics invites window dressing
France lets each buyer pick the metrics “most relevant” to each contract, checked by a sworn statement or, for larger contracts, an auditor’s certificate. A supplier will then show its strong metrics and hope the buyer skips its weak one. For a given service, the weak point decides. If the maintenance team, the source code or the right to leave sits outside Europe, scores on headcount and head office cannot make up for it. The joint paper makes the same mistake when it lets “strengths in one dimension … compensate deficits in others”. Every supplier should have to meet all the control tests; the menu should cover only the rest.
Declarations also age. A supplier that qualifies today can be bought tomorrow, as at least two dozen European technology companies have been in recent years. Framework contracts awarded on European credentials need a change-of-control clause that triggers a new assessment and lets the buyer terminate.
France already knows what “encourage” produces
France’s paper says nothing about Article 41, CADA’s one provision on open source. Its title promises “open source first”; its operative verb is “encourage”, with a broad exception attached.
France has run that experiment for ten years. Article 16 of its 2016 Digital Republic law tells public bodies to “encourage” free software and open formats. The largest public IT contracts have gone to the same few vendors ever since. French audit and parliamentary reports now find that the article changed nothing in what the state buys.
A rule works when someone enforces it. Make open source the default, require a written justification for each exception, costed over the life of the contract, and have a named authority read it before the decision and publish it. France, with ten years of evidence, is the Member State best placed to ask the Council to write “require” into Article 41.
Seven amendments France should table
- Keep the 20% floor and delete Article 32(2)(d), so European contribution can decide an award.
- Name the company the metrics apply to, and measure them at group level.
- Fill the open-source row and make it a test every supplier must meet.
- Make the control tests mandatory: EU maintenance with source-code access, core R&D, open-source licensing, a priced and tested exit.
- Price the exit in every framework contract, and judge the Article 30(4)© exception on lifecycle cost.
- Check declarations against licences, code repositories and company registers, and add a change-of-control clause to framework contracts.
- Replace “encourage” with “require, unless justified in writing” in Article 41, with an authority reading the justification before the decision.
France wants CADA to become “a genuine vehicle promoting value creation within the European Union”. A public buyer also needs to know whether it can change, fix or leave the systems it depends on. These seven amendments let CADA test for both.
