9APR

Sovereign cloud spend set to triple by 2027

3 min read

11:02UTC

European sovereign cloud spending is forecast to reach $23bn by 2027, up from $7bn in 2025. EU-native providers hold just 15% of the market.

← Back to Ceasefire redistributes the war, not ends it Jump to analysis ↓

ConflictDeveloping

Key takeaway

Sovereign cloud spending is tripling, but US hyperscalers and US-built AI models still dominate European infrastructure.

European sovereign cloud spending is forecast to triple from roughly $7bn in 2025 to $23bn by 2027 ¹. All 27 EU member states signed a digital sovereignty declaration in November 2025. European governments describe sovereignty as a "matter of national survival" ².

Budgets are growing, but the harder question is what runs on the infrastructure. AWS, Microsoft, and Google are all members of GAIA-X, Europe's flagship sovereign cloud initiative. The framework that was designed to reduce dependence on American providers now has the Americans inside the tent. GAIA-X's first multi-provider catalogue lists 600 services from 15 providers across four sovereignty tiers ³. Only the highest tier (Label Level 3) excludes companies subject to the US CLOUD Act. Uptake data for Level 3 is not publicly available.

Domestic providers (OVHcloud, Hetzner, Scaleway) account for roughly a sixth of European cloud revenue, with US hyperscalers commanding the rest ⁴. On price, the European alternatives win easily. But the vast majority of AI workloads on European cloud, sovereign or otherwise, use US-built models: OpenAI's GPT-4o, Anthropic's Claude, Google's Gemini. You can run a US model on a German server and call it sovereign. Genuine independence requires sovereignty at both the compute layer and the model layer. Europe has plausible compute alternatives. It has almost no enterprise-scale model alternatives.

Deep Analysis

In plain English

European cloud sovereignty is about who stores and processes your data, and whose laws apply to it. Three American companies; Amazon Web Services, Microsoft Azure, and Google Cloud; together control roughly 70% of the European cloud market. The remaining 30% is split between European companies (OVHcloud, Hetzner, Scaleway, T-Systems) and others. GAIA-X is a European initiative to create a framework for trusted cloud services, with different levels of sovereignty certification. Level 1 is basic compliance; Level 4 means the service is run by a company not subject to US law; which would exclude AWS, Azure, and Google. All 27 EU member states signed a digital sovereignty declaration in November 2025, signalling political commitment to using European cloud services for government data. But current procurement patterns have not changed: the €7 billion Europeans spent on sovereign cloud in 2025 is forecast to triple to €23 billion by 2027, though most of that growth may flow to American companies' European-branded products rather than genuinely European alternatives.

Deep Analysis

Root Causes

European sovereign cloud's persistent 15% market share despite years of policy attention reflects two structural constraints. First, enterprise IT procurement decisions have 5-10 year cloud provider lock-in timelines: organisations that migrated to AWS or Azure between 2015 and 2020 will not complete migration to European alternatives before 2025-2030, regardless of regulatory incentives. The procurement cycle is longer than any regulatory mandate cycle.

Second, GAIA-X's inclusion of US hyperscalers in its catalogue reflects a political compromise that undermines its market differentiation purpose. If AWS and Azure can achieve Level 1 and Level 2 GAIA-X certification by meeting basic data localisation standards, the GAIA-X brand loses its ability to signal European sovereignty to procurement officers.

The catalogue's 600 services from 15 providers includes providers that are CLOUD Act-subject; a structural inconsistency that sophisticated buyers will identify.

What could happen next?

Risk
GAIA-X's inclusion of US hyperscalers in its catalogue at lower sovereignty tiers risks making the GAIA-X brand a market confusion tool rather than a genuine sovereignty signal, undermining EU procurement differentiation.
Short term · 0.75
Opportunity
If EU member states implement mandatory domestic cloud preferences for regulated sectors (following South Korea's model), EU-native cloud providers could gain 10-15 percentage points of regulated-industry market share within 5 years.
Long term · 0.5
Consequence
The $23bn sovereign cloud forecast will disproportionately benefit AWS and Azure European Zone products and French/German hyperscaler sovereignty wrappers, not EU-native independent providers; unless DMA switching cost reductions materialise.
Medium term · 0.7

Source Landscape

This story draws on neutral-leaning sources

Primary parallel: Japan's attempt to build domestic internet infrastructure in the 1990s through the WIDE Project and domestic ISP ecosystem produced a robust local industry that briefly held significant market share before the global internet's winner-take-most dynamics concentrated hosting and cloud infrastructure in US providers.

Japanese internet sovereignty was real for a decade, then structurally eroded. Europe's GAIA-X faces the same trajectory if it cannot maintain regulatory differentiation that US providers cannot replicate.

Counter-parallel: South Korea's Naver Cloud Platform has maintained approximately 60% domestic market share against AWS and Azure in South Korea, supported by regulatory preferences for domestic cloud in public sector and financial services procurement, and by a domestic technology ecosystem that values local data residency.

If the EU were to implement equivalent mandatory domestic cloud preferences for regulated sectors; rather than voluntary sovereignty tiers; the Korean model suggests genuine market share maintenance is achievable.

The $7bn to $23bn sovereign cloud forecast growth represents a 229% increase in 3 years; roughly 95% compound annual growth rate. This is not market share gain by European providers but rather the rapid monetisation of Europe's existing preference for 'sovereign' framing in procurement. Much of this growth will accrue to AWS and Azure European Zone products and Thales/S3NS (Google Cloud wrapped in French infrastructure), not to EU-native independent providers.

Hetzner's 14.3x and Scaleway's 4.8x cost-performance advantage relative to AWS (per Holori/Callista pricing analysis) is real and measurable, but has not translated into market share gains because switching costs; not running costs; determine enterprise cloud provider selection. DMA interoperability mandates, if implemented, would directly address the switching cost constraint that prevents these price advantages from translating into market share.

Consensus view: Synergy Research Group and IDC both assess the 15% EU-native cloud market share figure as structurally stable over the past five years; it has not grown materially despite GAIA-X, sovereign cloud rhetoric, and national procurement initiatives, because enterprise switching costs and the AWS/Azure ecosystem premium have offset any price or regulatory advantage European providers offer.

The $7bn to $23bn forecast growth is an expansion of overall European cloud spend, not a rebalancing of market share.

Counter-view: OVHcloud CEO Jean-Marc Lazard argues the sovereignty-tier market is growing faster than aggregate figures suggest: the relevant metric is not total cloud market share but share of regulated-industry workloads (financial services, healthcare, public administration, defence) where GDPR and data localisation requirements create genuine preference for EU-native providers.

In that segment, OVHcloud, Hetzner, and T-Systems collectively hold approximately 35-40% of European regulated workload revenue; not 15%.

Key tension: Whether GAIA-X's four sovereignty tiers create a meaningful market segmentation that produces real procurement differentiation, or whether Level 3 and Level 4 (excluding CLOUD Act-subject companies) remain aspirational categories with no substantial uptake among European enterprise buyers.

First Reported In

Update #1 · Europe's chip ambitions meet reality

European Central Bank· 13 Apr 2026

Read original →

Causes and effects

This Event

Sovereign cloud spend set to triple by 2027

The spending surge demonstrates real demand for sovereign cloud infrastructure, but GAIA-X's inclusion of US hyperscalers in its catalogue raises questions about whether the framework delivers genuine independence or sovereignty in name only.

Led to

EU awards first sovereign cloud deal

The EC €180m sovereign cloud framework directly responds to the $23bn European sovereign cloud market forecast and the EU-native 15% market share gap.

Occurred 17 Apr 2026

Read story →

Different Perspectives

Civilians and prisoners inside Iran

Mojtaba Kian was hanged in under 50 days from arrest, the fastest wartime espionage case in Hengaw's record, as Trump announced a peace deal. Amnesty places Iran's 2026 execution count above 200 at its fastest pace in 44 years; the diplomatic track has not altered the internal enforcement tempo.

China

Beijing accepted a Pakistani civilian briefing mission on the same day OFAC's GL V expired, keeping itself inside the deal architecture without being a named signatory. How Chinese banks respond to Monday's Hengli dollar-clearing decision is the first real-world test of whether the verbal MOU carries any institutional weight.

Pakistan

Islamabad split its highest-level delegation: army chief Munir to Tehran on the security track, Prime Minister Sharif and Foreign Minister Dar to Beijing before Monday's GL V-driven bank compliance decision. The architecture routes the deal's hardest problem, IRGC buy-in, through the general-officer channel that has extracted every wartime concession.

Israeli government

An unnamed Israeli official told the Times of Israel that Trump privately told Netanyahu the deal will dismantle Iran's nuclear programme and remove all its uranium, terms incompatible with what Tehran and a Reuters source describe. If Netanyahu believes he was promised full dismantlement and the deal delivers less, Israel holds a sabotage veto before any signature.

Iranian Foreign Ministry

Spokesman Esmail Baghaei told state agency IRNA that nuclear issues are 'not in the current negotiations text' and the sequencing is: end the war first, then negotiate nuclear over two months. Baghaei's formulation preserves Khamenei's 21 May uranium-stay directive while letting the civilian diplomacy track continue.

Donald Trump / White House

Trump declared the Iran deal 'largely negotiated' on 23 May via Truth Social and signed nothing; the White House's only paper was a Memorial Day proclamation. The verbal-track method converts maximum political signalling into minimum legal exposure: no congressional notification, no Senate treaty ratification, no instrument for Iran to formally reject.