A policy analysis from June 2026
When Ursula von der Leyen stood before delegates at the Paris AI Action Summit in February 2025 and declared that global AI leadership was “within reach” for Europe, it was a statement of intent as much as strategy. Six weeks later, on 9 April 2025, the European Commission published the AI Continent Action Plan — its most sweeping attempt to turn that intent into policy. Structured around five pillars, backed by a €200 billion investment programme, and timed deliberately to signal that Europe could compete as well as regulate, the plan asked a simple question: would the EU be a maker of AI, or merely a manager of it?
Fourteen months on, with 19 AI Factories operational, a landmark regulatory simplification agreed, and a €1 billion adoption drive underway, the picture is more complex than either enthusiasts or sceptics predicted. Infrastructure is being built at genuine speed. But the adoption numbers remain stubbornly low, the AI Act’s simplification has left both industry and civil society dissatisfied, and the sovereign AI ambition faces the same structural challenge it always has: how do you turn public investment into globally competitive private companies?
This analysis examines what the AI Continent Action Plan actually set out to do, what has been delivered as of June 2026, where the gaps remain, and what the next phase looks like for Europe’s AI ambitions.
What the Plan Actually Set Out to Do
The AI Continent Action Plan was built on a frank diagnosis. Europe had world-class research talent, a unique industrial base, and — with the AI Act — the world’s first comprehensive AI regulatory framework. What it lacked was scale: in compute, in data access, in deployment, and in the domestic AI champions capable of competing with US hyperscalers and Chinese technology giants.
Rather than introducing new legislation, the plan pulled together concrete investment commitments, institutional reforms, and policy co-ordination across five pillars: scaling compute infrastructure through AI Factories and Gigafactories; creating a genuine single market for data; accelerating AI adoption in strategic sectors; building an EU AI talent base; and supporting trustworthy AI through practical implementation of the AI Act.
The ambition, as the Commission stated, was explicit: the EU should become not just a rule-setter but an AI continent — a place where models are trained, companies are built, and economic benefits accrue to European workers and citizens. The plan sat alongside the broader InvestAI initiative, which aimed to mobilise €200 billion across the EU economy, with €20 billion ring-fenced specifically for the flagship Gigafactory programme.
What Is Working: Building the Foundations at Scale
Compute: From Embarrassment to Ecosystem
Fourteen months ago, European public AI compute capacity was a persistent source of embarrassment in Brussels policy circles. The AI Continent Action Plan committed to changing that at pace.
The results have been substantial. By April 2026, the Commission confirmed that 19 AI Factories were operational across the EU’s world-leading supercomputing network, supported by 13 AI Factory antennas providing regional access points for researchers, startups, and industry who cannot afford their own infrastructure. The full list runs from LUMI-AIF in Finland and BSC-AIF in Spain to IT4LIA in Italy and MIMER in Sweden — a genuinely continental network rather than a Paris-or-Berlin affair.
Each AI Factory operates through the EuroHPC Joint Undertaking as a single-entry point for computing time, with access rules designed to prioritise EU startups and SMEs. Data Labs embedded within the factories gather and curate high-quality datasets for AI development — a deliberate attempt to solve data access and compute in the same institutional move.
Above the factories sits the Gigafactory programme, which represents the plan’s most ambitious and contested infrastructure bet. The Commission committed to establishing up to five AI Gigafactories — facilities equipped with over 100,000 advanced AI processors, roughly four times more powerful than current AI Factories — capable of training very large models at hyperscale. The InvestAI facility, launched in February 2025, provides the €20 billion financing mechanism, structured as a public-private partnership under which the EU contribution can cover up to 17 percent of capital expenditure, with member states required to at least match that contribution.
The call for expressions of interest attracted 76 responses across 60 sites in 16 different member states — a level of demand that exceeded Commission expectations. One confirmed site in Norway has already begun construction, running on hydropower with direct-to-chip liquid cooling, with first GPUs expected online by the end of 2026.
The plan also set a broader infrastructure target: tripling EU data centre capacity within five to seven years, supported by the proposed Cloud and AI Development Act to incentivise private investment. This sits alongside the EuroHPC Regulation amendment — Council Regulation (EU) 2026/150 — which creates the formal legal basis for EU co-ownership of Gigafactory computing time, to be redistributed to European startups.
Data: The Data Union Strategy Takes Shape
One of the plan’s most structurally significant elements has been the launch of the Data Union Strategy, published by the Commission in November 2025. The strategy aims to create a genuine internal market for data — the precondition, in the Commission’s view, for European AI developers to train competitive models without depending on data assets concentrated in the hands of US platforms.
The strategy focuses on interoperability, data availability, and cross-sector data sharing. Data Labs within AI Factories are being used to aggregate and curate datasets from diverse public and private sources. New access rules are being developed to prioritise EU-funded research and AI innovators, and the strategy explicitly links to the Apply AI programme to ensure sectoral data gaps are addressed where adoption is being targeted.
The Apply AI Strategy: Taking the Ambition Sectoral
Launched in October 2025, the Apply AI Strategy is the plan’s operational answer to the adoption problem. Rather than assuming that infrastructure investment would automatically translate into economic deployment, the strategy makes adoption itself a policy objective — structured around eleven sectoral flagships covering manufacturing, healthcare, agriculture, energy, mobility, and the public sector, among others.
The strategy promotes an “AI first” approach, encouraging organisations across Europe to treat AI as a default candidate solution in strategic and operational decisions. It also promotes preferring open-source European AI solutions in public procurement — a buy-European policy for AI that has generated both enthusiasm from domestic developers and scepticism from economists worried about procurement distortions.
With €1 billion in funding calls already earmarked, the Apply AI programme is backed with serious money. The initiative is complemented by a European network of AI-powered advanced screening centres and a frontier AI grand challenge — a competitive prize mechanism to stimulate the development of European AI capabilities in high-priority areas.
The Apply AI Alliance serves as the governance mechanism — a co-ordination forum bringing together AI providers, industry, public sector, academia, social partners, and civil society to participate in sectoral AI policymaking.
Investment Momentum Is Real
The commercial signals are also moving in the right direction. Since the plan’s publication, investment commitments across the EU AI ecosystem have grown meaningfully. The InvestAI umbrella targets €200 billion in total AI investment across the union — an ambition that dwarfs anything the EU has previously put behind a technology programme and reflects the Draghi Report’s central argument that the competitiveness gap with the US is partly a financing problem.
Where It Is Stuck: The Persistent Gaps
Despite genuine infrastructure progress, three structural problems continue to shadow the plan.
Adoption Is the Defining Challenge — and the Numbers Are Uncomfortable
The foundational problem the AI Continent Action Plan was designed to solve has not gone away. According to the Commission’s own Apply AI Strategy documentation, published alongside the plan in 2025, only 13.5 percent of EU businesses and 12.6 percent of SMEs were using AI technologies. The gap between large firms and small ones is stark: Eurostat data reveals a 38-percentage-point disparity in AI adoption between large enterprises and small firms across the EU.
The structural causes are familiar. SMEs face accuracy concerns, legal uncertainty, skills shortages, and an absence of governance frameworks that can be applied at small-company scale. The Apply AI Strategy addresses this directly — its second pillar is explicitly about cross-cutting challenges including SME enablement, workforce readiness, and trust-building — but the gap between policy intent and measurable deployment across millions of European businesses remains vast.
The public sector picture is similarly uneven. Adoption across European public administrations remains fragmented, with digital capability concentrated in larger, more resourced agencies rather than distributed across local government, healthcare systems, and schools. The Commission has launched four pilot projects for deploying generative AI in public administrations, and European Digital Innovation Hubs have been refocused as AI experience centres since December 2025. Whether that is enough to shift the dial at scale is an open question.
The AI Act: Simplified, But Not Settled
The EU AI Act was, and remains, the most comprehensive AI regulatory framework in the world. It was also — as its critics warned — a source of genuine compliance complexity for the businesses it was meant to enable. The AI Continent Action Plan recognised this tension from the start, and the Commission’s response has been the AI Omnibus, part of the broader Digital Omnibus simplification package.
After six months of negotiations conducted under an exceptionally tight schedule, EU legislators reached a provisional agreement on the AI Omnibus at 4:30 in the morning on 7 May 2026. The deal is structurally meaningful. High-risk AI rules — covering systems used in employment, education, biometrics, critical infrastructure, and migration — will now apply from December 2027 rather than August 2026. The national regulatory sandbox obligation has been pushed to August 2027. Obligations and certification procedures have been streamlined, SME privileges extended to small mid-cap companies, and the AI Office’s supervisory role has been clarified and centralised.
New prohibitions have also been added: on AI-generated non-consensual intimate imagery and on child sexual abuse material — substantive policy changes tucked into a simplification vehicle.
The core risk-based framework remains intact. But the deal has left stakeholders on both sides frustrated. Industry representatives, who lobbied hard for relief, see the outcome as partial: compliance costs for high-risk systems remain substantial, and SMEs still face governance requirements that are resource-intensive even after simplification. Civil society organisations — including Amnesty International — have characterised the omnibus as a deregulation exercise driven by Big Tech lobbying, warning that weakening high-risk protections could expose EU citizens to discrimination and surveillance harms. The Jacques Delors Centre has criticised the absence of comprehensive impact assessments given the speed of the process.
What is genuinely clear is that the August 2026 deadline for high-risk AI rules — widely described as unworkable before harmonised standards had even been finalised — would have been a compliance crisis. On that narrow but important point, the omnibus has done its job.
The Talent and Skills Gap Is a Long-Run Race
The plan set ambitious targets for talent: an AI Skills Academy launched in 2025, international recruitment partnerships, scholarships for female professionals re-entering the AI workforce, and an EU-India legal gateway office launched in February 2026 to facilitate talent movement in the ICT sector.
The pipeline is real. Europe produces strong AI researchers and maintains world-leading universities. But retaining that talent — and ensuring it flows into European companies rather than US platforms — remains a structural challenge that no single initiative resolves. The returnship schemes and fellowships are welcome interventions. Whether they can compete with the compensation structures, compute access, and career trajectory offered by US frontier labs is a harder question.
Regulatory Fragmentation Across 27 Member States
A persistent structural challenge for the AI Continent Plan is that it operates across 27 national regulatory environments with varying enforcement capacity, digital infrastructure maturity, and institutional appetite for AI deployment. The AI Office provides central co-ordination for frontier model governance, but most day-to-day AI adoption decisions involve national regulators, national procurement processes, and national skills systems that the Commission can influence but cannot control. The Apply AI Alliance is a useful convening mechanism. Whether it translates into genuine policy co-ordination at implementation speed remains to be seen.
The Regulation Question: Rule-Setter and Rule-Taker
The EU’s positioning on AI governance is its most distinctive contribution to the global landscape — and its most complicated competitive variable. The AI Act represents genuine intellectual leadership: a risk-based framework that takes seriously the societal implications of AI deployment in high-stakes domains. The AI Office has developed credibility as an enforcement and monitoring body. The international influence of the EU regulatory model, though contested, is real.
But the competitiveness cost of regulatory complexity is also real. The omnibus process revealed how poorly calibrated some of the Act’s original implementation timelines were for a market that includes millions of resource-constrained SMEs. The Tech Policy Press analysis of the deal noted that the outcome left stakeholders on both sides frustrated — industry wanting more relief, civil society wanting fewer concessions.
The Draghi Report’s central argument — that Europe’s competitiveness gap with the US partly reflects the cumulative burden of overlapping regulations applied to the same economic activity — has not been resolved by a simplification exercise that left both industry and civil society unhappy. The wider “digital fitness check” planned for the first quarter of 2027 will be a further test of whether the EU can sustain regulatory ambition alongside genuine competitiveness.
What Comes Next: The Deployment Phase
The April 2026 one-year review was clear about the transition underway. The founding phase of the AI Continent Plan — commissioning AI Factories, launching InvestAI, establishing the Apply AI framework, publishing the Data Union Strategy — is largely complete. The plan now has to prove that infrastructure translates into adoption, and adoption into the kind of broad-based productivity growth that justifies the investment.
Several directions are clearly in motion for the remainder of 2026 and into 2027.
The Gigafactories are the centrepiece of the next phase. With 76 expressions of interest received and construction beginning in at least one location, the programme is moving from design to implementation. Getting the first facilities operational — and ensuring that their computing time actually reaches European startups rather than being absorbed by larger incumbents — will be watched closely by anyone assessing whether the EU’s sovereign AI ambition has genuine industrial policy teeth.
The Apply AI sectoral flagships are where deployment ambitions will be tested most concretely. Healthcare, manufacturing, and energy are the sectors where the Commission’s rhetoric about AI transforming European industry will either produce measurable results or expose the limits of top-down adoption programmes. The €1 billion in earmarked funding calls needs to reach organisations capable of deploying effectively — which requires procurement reform, governance frameworks, and skills development running in parallel with investment.
The European AI Innovation Month, planned for October to November 2026, will be the Commission’s most visible opportunity to showcase tangible results before the end of the year.
The AI Act compliance calendar continues regardless. With high-risk AI rules now deferred to December 2027, businesses have breathing room — but not unlimited time. The AI Office will need to use the extra space to publish the harmonised standards and compliance tools that were missing when the original August deadline was set.
An Action Plan at Its Hardest Moment
The EU AI Continent Action Plan was ambitious when it launched. Fourteen months on, the infrastructure progress is genuine: 19 AI Factories deployed, a €20 billion Gigafactory programme in motion, a Data Union Strategy published, an Apply AI adoption programme funded at €1 billion, and an AI Act simplification agreed under severe time pressure.
But the harder yards remain ahead. Only 13.5 percent of EU businesses are using AI. The SME adoption gap is 38 percentage points against large firms. The AI Act omnibus has provided deadline relief without resolving the deeper tension between regulatory ambition and compliance capacity. And the Gigafactories — the most interventionist element of the plan — have not yet produced a single operational facility, let alone a European frontier AI model capable of competing globally.
The EU’s bet is that sovereign infrastructure, combined with a unified data market, regulatory clarity, and targeted sector adoption, can produce the conditions for European AI champions to emerge. It is a coherent industrial policy argument. Whether it can be executed at the speed and scale that global AI competition demands is the central question of the next chapter — and the European AI Innovation Month in autumn 2026 will offer the first meaningful public reckoning.
Key References and Further Reading
- AI Continent Action Plan (April 2025) — European Commission
- AI Continent Action Plan: One Year On (April 2026) — European Commission
- AI Continent Action Plan: Progress Dashboard — European Commission
- AI Factories and Gigafactories — European Commission / EuroHPC JU
- Apply AI Strategy (October 2025) — European Commission
- Data Union Strategy (November 2025) — European Commission
- AI Omnibus: May 2026 Agreement Analysis — Mishcon de Reya
- AI Omnibus Deal: What Changed — Tech Policy Press
- EU Digital and AI Omnibus: Heading in the Wrong Direction? — Jacques Delors Centre
- InvestAI and AI Gigafactories: Finance Architecture — European Economics
- EU AI Gigafactory Initiative: Infrastructure Analysis — STL Partners
- Apply AI Strategy: EUR-Lex Full Text — EUR-Lex
- AI Continent Action Plan: Open Future Observatory — Open Future
- EU AI Act: Legislative Train Schedule — European Parliament
