Skip to content
You can now search across every topic, entity and event.What's new
Russia-Ukraine War 2026
19JUL

Commission awards sovereign cloud slot to Google joint venture

5 min read
13:35UTC

The European Commission's €180m six-year sovereign cloud framework, named between 17 and 20 April, handed one of four provider slots to S3NS, a Thales-Google joint venture that cleared only the minimum SEAL-2 threshold.

ConflictDeveloping
Key takeaway

A sovereignty framework that treats SEAL-2 and SEAL-3 as interchangeable has collapsed its own distinction.

Between 17 and 20 April 2026 the European Commission named the four provider groupings selected for its €180m, six-year sovereign cloud framework . Post Telecom leads a Luxembourg consortium with CleverCloud and OVHcloud at SEAL-3; STACKIT, owned by the Schwarz Group, represents Germany at SEAL-3; Scaleway represents France at SEAL-3; and Proximus leads a Belgian consortium that includes S3NS, Clarence and Mistral AI, with S3NS rated at SEAL-2. The Commission's Sovereignty European Assurance Level framework grades contracts from SEAL-1 to SEAL-3, and SEAL-3 requires the operator to run the service without foreign technical dependencies. SEAL-2 requires only data-sovereignty minima. Three awardees cleared the higher bar; one did not.

CISPE's Francisco Mingorance told The Register the award was "clearly an own goal" and called the S3NS inclusion "sovereignty washing" . The objection is specific: S3NS is a joint venture between French defence group Thales and Google Cloud, and workloads running on Google infrastructure sit within reach of the US CLOUD Act, the 2018 law that lets American authorities compel disclosure of cloud data held anywhere in the world. Microsoft told the French Senate last year that it could not guarantee French customer data would never be disclosed under US legal orders; the same exposure attaches structurally to any Google-hosted workload regardless of where the data physically sits.

EU-native providers hold roughly 15 per cent of the European cloud market against around 70 per cent for the three US hyperscalers combined , and The Commission's counter-case rests on that arithmetic. Excluding every provider with any US linkage would have narrowed the tender to a field too small to be competitive, and the sovereign cloud market is forecast to triple to $23bn by 2027; a procurement logic that restricts the supplier base risks producing either no viable bidder or a single one. Hedged inclusion of S3NS at SEAL-2, alongside three SEAL-3 awardees, reads on that defence as a deliberate widening of the competitive field rather than a concession on sovereignty.

A €180m framework becomes the citation benchmark for every national cloud tender that follows, and member-state procurement Teams tend to cite The Commission's own designation as prima facie evidence of sovereignty compliance. Over six years the downstream contracts will be valued well above €180m, and a SEAL-2 provider sitting inside the same slot as a SEAL-3 one collapses two categories the framework was designed to separate. CISPE has the foundation for a legal challenge; Mingorance's keynote in Brussels this week on "Making Sovereignty Verifiable" suggests the challenge will come through an auditability argument rather than a tender-rules one.

Deep Analysis

In plain English

The EU runs its own computer systems on rented cloud infrastructure. This contract decides which companies can bid to provide that infrastructure for the next six years, with rules about how 'European' those providers must be. The controversy is that one of the four winners, S3NS, is a joint venture between French defence company Thales and Google. Critics say that because Google is an American company, EU data running on S3NS infrastructure could legally be accessed by US courts under a 2018 US law called the CLOUD Act. The Commission's response is that S3NS met the minimum sovereignty standard, and excluding it would have left too few suppliers in the competition.

Deep Analysis
Root Causes

Three structural factors make hyperscaler dependency difficult to escape. The managed services gap is the first: EU-native providers can match hyperscalers on raw compute and storage pricing but cannot yet replicate the depth of managed database, machine learning, networking and observability services that enterprise workloads depend on.

Migrating from Azure to OVHcloud or Scaleway requires rebuilding application dependencies against a shallower service catalogue, and the engineering cost is a real barrier for large organisations.

Network effects in procurement compound the problem: once a ministry or large enterprise runs on a hyperscaler, its staff develop platform-specific skills, its vendors integrate through that platform's APIs, and switching costs compound over time. Third, hyperscalers can cross-subsidise European market share from US revenue bases at a scale no European provider can match.

What could happen next?
  • Meaning

    The first consequence is precedential: member-state procurement agencies will cite the Commission's own SEAL-2 award as justification for purchasing from S3NS and structurally similar Thales-Microsoft or Thales-Amazon joint ventures. If those downstream contracts follow, the practical effect of the EU's sovereign cloud framework over six years will be to channel public money toward hybrid US-European structures rather than toward EU-native providers. The second consequence is political: CISPE's public challenge delegitimises the framework before it has generated a single call-off contract. If the Commission does not respond with a published SEAL audit methodology, the 'sovereignty washing' framing will become the dominant description of the award in procurement guidance discussions across Europe.

First Reported In

Update #3 · Sovereignty summit, minus the sovereigns

The Register· 23 Apr 2026
Read original
Different Perspectives
The United Kingdom
The United Kingdom
Starmer pledged £300 million in Kyiv on 16 July toward Ukraine's Gripen E squadron, adding to the PURL expansion Trump and Rutte had announced two days earlier. London is paying into a scheme built around a shortfall NATO's own published $4bn-plus pledge does not close against Zelenskyy's roughly $15bn stated need.
Brussels
Brussels
The EU's 21st sanctions package missed its Coreper vote on 15 July over Greek LNG re-export rights and an Austrian bank compensation demand, the same week Hungary stalled accession clusters on procedure rather than veto. Both processes run on unanimity, so a single national interest, not Russia policy, sets the pace either can move at.
Hungary's Tisza government
Hungary's Tisza government
Budapest refused to open EU accession Clusters 2 and 3 for Ukraine at COELA on 17 July, offering Moldova a standalone opening instead, and the question returns on 22 July. Having ended Orbán's blanket loan veto in May, it now blocks the narrower rule-of-law chapters where its own electorate is least comfortable.
Washington
Washington
Trump and Rutte expanded PURL on 14 July, letting allies fund the American interceptors and jets Washington will license but no longer gift outright. The same week, Lockheed Martin told allies it cannot guarantee PAC-3 MSE delivery timelines even after tripling output, so Washington now shapes Ukraine's air defence through a supply queue rather than a donation decision.
Moscow
Moscow
Novak ordered a study into cutting the diesel exchange quota to 10% within a week of his export ban, while June delivered Russia's first budget surplus of 2026 and National Wealth Fund liquidity above its own May forecast. Its own investors disagree: the Moscow Exchange has fallen for its longest losing streak since 1997.
Ukraine's government and its street protesters
Ukraine's government and its street protesters
Zelenskyy sacked Fedorov on 15 July, installed an acting SBU officer in his place, and did not move against three days of protest that followed across eight cities. He is betting that visible tolerance for dissent, timed to EU accession hearings on rule of law, outweighs whatever command dispute forced the reshuffle.