How Is Europe’s AI Strategy Turning Regulation into Industrial Policy?

Key Takeaways

• Europe is pairing AI rules with compute, data, cloud, skills, and sector adoption.
• AI Factories and gigafactories are meant to give startups shared access to compute.
• Implementation will depend on member-state coordination, energy supply, and procurement.

Europe’s AI Strategy Starts with a Concrete Date

As of June 2026, the European Commission’s Cloud and AI Development Act proposal had become part of a wider push to make artificial intelligence (AI) capacity less dependent on foreign cloud platforms and more useful to European industry. That timing matters because Europe’s AI strategy is no longer limited to rulemaking. It now combines the AI Continent Action Plan, the Apply AI Strategy, the European Data Union Strategy, AI Factories, planned AI gigafactories, and a phased timetable for the European Union Artificial Intelligence Act.

The European Union (EU) is trying to solve three problems at once. It wants firms and public bodies to adopt AI faster. It wants more of the underlying compute, data, cloud, and model capacity to sit inside Europe or under stronger European control. It also wants the AI Act to become a market-shaping rulebook rather than a brake on deployment. That combination is why the strategy now looks like industrial policy, digital sovereignty policy, and governance policy inside one package.

The AI Act remains the legal anchor. It entered into force on August 1, 2024, and its general application date is August 2, 2026, with phased dates for banned practices, general-purpose AI, and high-risk systems. Yet the Commission has spent 2025 and 2026 adding the missing implementation layer: access to supercomputers, data labs, sector projects, public procurement tools, guidance, sandboxes, and funding calls. New Space Economy’s discussion of AI governance in 2026 captures the broader shift from soft principles toward binding rules, procurement duties, and infrastructure choices.

The strategy is also a response to a visible gap. Europe has strong research institutions, large industrial firms, public-sector buyers, and regulated markets that generate valuable data. It has fewer globally dominant AI model developers and far less hyperscale cloud capacity than the United States. Its plan, as of June 2026, is to use public infrastructure and rule design to make AI adoption easier for small and medium-sized enterprises (SMEs), researchers, and public administrations that cannot build large model pipelines by themselves.

The AI Continent Action Plan Builds the Operating System

The AI Continent Action Plan, released in April 2025, provides the top-level design. It organizes Europe’s AI push around computing infrastructure, data, sector adoption, skills and talent, and regulatory support. That structure is practical. A firm cannot adopt AI safely if it lacks affordable compute, usable data, trained staff, procurement confidence, or legal clarity. A public body cannot put AI into health, energy, transport, or defense workflows without similar support.

The plan’s compute pillar centers on AI Factories. These are access points linked to EuroHPC supercomputing resources, support teams, and related services. The Commission says AI Factories let startups, SMEs, researchers, and public organizations train, refine, and integrate models without needing to buy their own advanced clusters. As of April 2026, the network included 19 AI Factories and 13 AI Factory antennas, with support tied to AI-optimized supercomputers and access services.

Gigafactories sit at the next scale. The Commission’s InvestAI initiative, launched in February 2025, targets €200 billion in AI investment, including a €20 billion European fund for large AI gigafactories. The goal is to support development of more complex models, including frontier AI and industrial AI. The Commission has described these sites as much larger than standard AI Factories, with the action plan indicating a plan to support up to five gigafactories.

Data forms another operating layer. AI adoption depends on access to high-quality data that can be used legally, securely, and at scale. The Data Union Strategy is designed to widen access to high-quality datasets, simplify data rules, expand common European data spaces, launch data labs, and support international data flows on terms aligned with EU values and interests. That matters for sectors such as health, mobility, energy, agriculture, media, and space, where model performance depends on sector-specific data rather than generic web text.

The action plan also tries to connect AI development to adoption. The Apply AI Strategy gives the Commission a sector route for deployment. Instead of treating AI as a standalone software market, it places AI inside healthcare, manufacturing, mobility, energy, climate, communications, culture, public administration, security, and space. New Space Economy’s review of the AI value chain offers a useful parallel because it separates chips, cloud, models, applications, data owners, and governance providers into different value layers.

The strategy has six linked execution lanes.

Infrastructure Plans Turn Strategy into Capacity

AI strategy becomes real only when compute is available, affordable, and matched to user needs. That is the reason AI Factories occupy such a central place in the Commission’s plan. A startup developing a medical imaging model, a robotics firm training a factory inspection system, and a climate research group working with satellite data do not need the same infrastructure. They do need access to secure computing, technical support, model-tuning capacity, and data pipelines that would be costly to build alone.

The EuroHPC Joint Undertaking gives the EU a vehicle for this shared infrastructure model. The AI Factories draw on supercomputing capacity that already had a scientific and industrial base. The change is access. The Commission and EuroHPC have turned supercomputers into service platforms for AI users, with priority for startups and SMEs. That approach suits Europe’s market structure because much of its industrial strength sits in manufacturing firms, specialized software vendors, laboratories, and regional clusters rather than a small group of hyperscale AI firms.

The planned AI gigafactories address a different bottleneck: very large model development. These facilities are meant to train and develop complex models that require far more computing power than standard access programs. The Commission has framed them as shared European infrastructure so that model development does not depend only on foreign hyperscalers. This is one of the sharpest sovereignty claims in the strategy. Compute is no longer just an input. It is a bargaining asset, a location decision, an energy question, and a procurement tool.

CADA adds the physical layer beneath compute. Its proposal seeks to simplify permitting, improve access to energy, land, water, and financing, and at least triple EU data center capacity within five to seven years. That plan will face local constraints. Data centers need grid connections, cooling, site approvals, and credible demand forecasts. They can support AI adoption, but they can also intensify competition for electricity and water in regions already managing industrial loads.

For space and Earth observation markets, compute is more than a back-office issue. Copernicus, Galileo, EGNOS, secure satellite communications, and commercial satellite imagery generate data that becomes valuable only after processing, validation, and integration into user workflows. New Space Economy’s discussion of the EU downstream space economy describes how open space data can still support commercial revenue through interpretation, integration, assurance, and workflow-specific services. AI Factories and data spaces could reinforce that model if firms can access both compute and trustworthy data.

The infrastructure plan also affects competition policy. Publicly supported compute can lower entry barriers, but only if access rules are transparent and allocation does not favor politically connected users. The EU will need to show that its shared facilities help firms build products, raise capital, and win customers. Otherwise, AI Factories could become useful research infrastructure without enough industrial pull.

Data, Cloud Sovereignty, and Semiconductors Set the Supply Base

Europe’s AI strategy treats data, cloud, and chips as linked supply constraints. The Data Union Strategy focuses on availability and legal clarity. CADA focuses on cloud and data center capacity. The EU’s wider tech sovereignty package connects those efforts to semiconductor policy. The message is direct: AI adoption cannot depend only on model access if the compute stack, data rights, cloud operations, and chip supply remain outside European influence.

The Data Union Strategy has three priorities. It seeks to scale access to data for AI, streamline data rules, and safeguard data sovereignty. Its proposed data labs would connect common European data spaces with AI users, pooling public and private resources so firms and researchers can use higher-quality sector datasets. The Commission also points to expansion of common European data spaces, including work on a defense data space, and to 30 million digitized cultural objects becoming available for AI training under its broader data measures.

The data agenda matters because many European strengths are domain strengths. A pharmaceutical company, railway operator, precision-manufacturing firm, public hospital, energy grid operator, or space-data company owns knowledge that generic models do not fully capture. Europe’s industrial AI opportunity depends on turning that data into safe, compliant, and usable training, testing, and deployment material. That requires data governance, not just storage.

CADA brings cloud sovereignty into the same frame. The proposal describes four assurance levels for cloud and AI sovereignty, ranging from EU-based processing and storage to higher levels that require independence from third countries, supply-chain transparency, and EU ownership and control. Public bodies would use the framework based on risk assessments. That design avoids a blanket foreign-provider ban, but it gives public procurement a structured way to favor stronger European control for sensitive uses.

This approach responds to a real dependency problem. U.S. cloud providers dominate much of the European enterprise cloud market, and cloud law, export controls, vendor policy, and platform terms can all affect public-sector and regulated-sector users. New Space Economy’s article on sovereign AI makes a related point: sovereignty depends on portability, standards, exit options, and procurement terms, not just the nationality of a model vendor.

Semiconductors complete the supply base. Training and running advanced models depend on accelerators, memory, networking, and power systems. Europe does not control the full AI chip chain, but it can use procurement, research funding, packaging capacity, chip design support, and data-center approvals to improve resilience. The European Chips Act gives the EU an existing policy base for strengthening parts of the semiconductor chain. That will not make the EU self-sufficient. It can reduce the risk that every industrial AI project depends on the same external hardware supply and cloud platforms.

The Apply AI Strategy Pushes Adoption into Real Sectors

The Apply AI Strategy turns the AI Continent Action Plan from a capability program into a deployment program. It identifies 10 industry sectors plus the public sector for targeted action. The list includes healthcare and pharmaceuticals, mobility and automotive, robotics, manufacturing, engineering and construction, climate and environment, energy, agri-food, defense, security and space, electronic communications, and cultural, creative, and media sectors. That breadth shows how the Commission sees AI as a general-purpose production tool rather than a narrow software category.

The strategy promotes what the Commission calls an “AI first policy,” meaning that organizations should consider AI as one possible tool when they make strategic or policy decisions. That phrase should not be read as an order to automate everything. In regulated sectors, AI adoption has to pass tests of safety, legality, reliability, human oversight, and cost. A hospital, transit operator, or border agency faces different risks than a marketing department. The value of the Apply AI Strategy is that it puts sector-specific deployment above abstract AI enthusiasm.

Healthcare provides a useful example. The Commission points to AI-powered advanced screening centers as a flagship. The policy idea is that AI can support earlier detection, triage, imaging review, workflow management, and research if health systems have data access, governance, validation, and procurement capacity. The difficulty is that health data can be sensitive, fragmented, and unevenly formatted. Adoption will depend on standards, clinical validation, liability management, and trust from patients and professionals.

Manufacturing and robotics show another path. European firms have deep industrial know-how, but many lack large AI teams. AI Testing and Experimentation Facilities, Experience Centres for AI, and AI Factories could help firms test models for quality inspection, predictive maintenance, process control, digital twins, and supply-chain management. New Space Economy’s AI taxonomy gives a useful structure for separating model types, workloads, applications, and market roles.

Space is a smaller sector, but it illustrates the strategy’s logic. Satellites, ground systems, launch operations, and Earth observation services already use automation, image analysis, anomaly detection, mission planning, and communications optimization. AI can help process Earth observation data, manage satellite fleets, support onboard autonomy, and reduce operations costs. New Space Economy’s article on AI and space exploration connects AI to mission operations, scientific analysis, satellite management, and human spaceflight, and its article on autonomous satellite operations explains how automation changes the ground segment cost curve.

The Apply AI Strategy will be judged by adoption rates, not press releases. Firms must see lower costs, faster testing, better compliance guidance, and easier procurement. Public bodies must avoid turning pilot programs into permanent demonstration projects. Europe has a strong record of funding research. The harder test is moving validated AI into daily operations, budget lines, and measurable service gains.

The AI Act Turns Trust into Market Access

The AI Act gives Europe’s AI strategy its governance foundation. It creates a risk-based framework for AI systems and sets duties for providers, deployers, importers, distributors, and public authorities. Some uses are banned. High-risk systems face duties tied to risk management, data governance, technical documentation, transparency, human oversight, accuracy, cybersecurity, and post-market monitoring. General-purpose AI model providers face separate duties, with stronger expectations for the most capable models.

The timetable matters. Prohibited AI practices and AI literacy duties began applying on February 2, 2025. Governance rules and general-purpose AI model duties became applicable on August 2, 2025. The AI Act’s general application date is August 2, 2026. The Commission’s 2026 simplification work and political agreement on the AI Omnibus clarified later timing for high-risk systems: certain high-risk areas apply from December 2, 2027, and product-integrated systems apply from August 2, 2028.

This staged timetable gives firms more time, but it also creates a demanding compliance runway. Organizations deploying AI in hiring, education, medical devices, biometrics, border control, public services, or infrastructure cannot wait until the application date. They need inventories of AI systems, data lineage, vendor documentation, risk controls, incident processes, human review procedures, and procurement clauses. General-purpose AI users need to understand which duties sit with model providers and which duties pass downstream to application vendors and deployers.

The European AI Office, national authorities, the AI Board, the Scientific Panel, and the Advisory Forum form the implementation structure. The General-Purpose AI Code of Practice adds a voluntary compliance tool for providers of general-purpose AI models. Guidance will matter because many organizations do not know whether their systems fall into high-risk categories, whether a general-purpose model has systemic risk, or how to document a modified model inside a product.

Trust is also commercial. A European AI firm that can prove compliance may win public-sector and regulated-sector contracts more easily. A vendor that cannot provide documentation, model information, testing evidence, or audit support may face slower sales. New Space Economy’s analysis of AI risks highlights issues such as model changes, version records, regression testing, and governance controls. These are no longer just internal quality practices. They are becoming market-access tools.

The risk is administrative drag. If compliance becomes too costly for SMEs, the AI Act could favor large incumbents with bigger legal and engineering teams. That is why the AI Factories, sandboxes, service desk, standardization work, and simplified guidance are part of the same strategy. Europe cannot claim an innovation model if its compliance system is too difficult for the firms it wants to support.

The main AI Act timing points are set out below.

Implementation Will Test Coordination Across Member States

Europe’s AI strategy has a coordination problem built into its design. The Commission can propose legislation, fund common infrastructure, organize EU-level programs, and set legal frameworks. Member states control many permits, grid connections, education systems, public procurement decisions, health systems, industrial policies, and regional development programs. That division can be a strength if national plans reinforce EU goals. It can become a weakness if countries compete for facilities without building interoperable systems.

AI Factories already show this mixed picture. A distributed network gives firms access points in many regions. It also requires common service quality, fair access rules, compatible procedures, and clear paths from prototype to production. A startup should not need to learn a completely different administrative process for each member state. Researchers and SMEs need predictable queues, usable technical support, transparent pricing after any subsidized access, and practical help with data access.

CADA raises similar issues. Tripling data center capacity is an EU-level target, but sites are local. A data center project can face delays because of electricity demand, water concerns, zoning, public opposition, supply-chain delays, or uncertainty about long-term demand. Energy policy will become part of AI policy. Regions with clean power, grid capacity, and fast permitting will have an advantage. Regions with fragile grids or public resistance may struggle to host capacity even if they have research talent.

Procurement is another test. The EU can promote a “buy European” approach for public bodies and open source AI solutions. Public buyers still need contract language, approved frameworks, vendor evaluation criteria, security controls, data rules, model documentation requirements, and staff who can manage AI projects. Without these tools, public-sector AI adoption could fragment into small pilots and isolated vendor contracts.

Skills could become the slowest constraint. The Commission has linked the wider strategy to talent programs, AI literacy, and skills measures. The need reaches beyond data scientists. Europe needs procurement officials who can evaluate AI contracts, lawyers who understand model duties, engineers who can manage data pipelines, public managers who can supervise automated systems, and sector specialists who can test whether AI outputs are useful. The hardest roles sit between domains, such as clinical AI validation, manufacturing data engineering, space-data product management, and compliance engineering.

The strategy’s success will also depend on whether Europe can hold together two goals that can pull apart. One goal is openness: foreign investment, trusted partners, research collaboration, and global standards. The other goal is autonomy: European control of sensitive data, procurement resilience, and reduced dependency in cloud and compute. CADA’s four assurance levels try to manage that tension. They let lower-risk uses remain open and reserve higher sovereignty expectations for more sensitive public-sector cases.

Europe’s AI Strategy and the Space Economy Share a Sovereignty Logic

The space economy shows why Europe’s AI strategy is larger than a digital-sector plan. Galileo, Copernicus, EGNOS, the IRIS² secure connectivity program, space situational awareness, and launch policy all reflect the same strategic pattern: build shared European infrastructure, reduce dependency in sensitive functions, and turn public systems into commercial opportunity. AI now follows that playbook.

Earth observation is a clear example. Copernicus provides open data, but firms create value by turning that data into flood models, crop analytics, insurance tools, maritime monitoring, carbon services, and public-sector decision support. AI can make those services faster and more scalable, but only if firms can access data, compute, validation methods, and customers. Europe’s Data Union Strategy and AI Factories could improve that pipeline if they connect with space-data users rather than remaining generic digital programs.

Autonomous satellite operations offer another example. Large constellations need automation for scheduling, anomaly detection, routing, collision-avoidance support, ground segment management, and mission planning. That creates demand for AI systems that must be reliable, auditable, secure, and compatible with safety rules. The same governance logic that applies to high-risk terrestrial AI can influence space operations, even where the AI Act does not directly prescribe every operational detail.

Europe’s space industrial base also has regional depth. New Space Economy’s profile of space industry economic centers in Europe describes a distributed industrial structure across countries and regions. That structure resembles the AI plan. Europe is not building one giant AI city. It is building a network of factories, antennas, data spaces, test facilities, hubs, and procurement channels. The network model can work if firms can move between nodes without friction.

A smaller but valuable point concerns dual-use demand. Defense, security, and space appear together in the Apply AI Strategy’s sector list. That does not mean every AI project is military. It means that satellite data, secure communications, autonomous systems, cyber defense, and infrastructure resilience now overlap. Public buyers will want AI tools that meet legal, security, and sovereignty expectations. Firms that can meet those expectations may win more long-term contracts in sensitive markets.

Europe’s AI strategy could give space companies better access to compute and data support. It could also raise compliance expectations for firms using AI in regulated or security-sensitive contexts. That trade is consistent with Europe’s model: public infrastructure helps firms innovate, and public rules shape how that innovation reaches the market.

The Main Risks Are Execution, Energy, and Speed

Europe’s AI strategy is coherent on paper. Its risks sit in execution. The largest risk is that infrastructure programs, legal duties, data projects, and adoption campaigns move at different speeds. Compute capacity can expand before enough firms are ready to use it. Guidance can arrive after procurement decisions. Data labs can exist without enough usable data. Sector pilots can succeed technically but fail to scale because budgets, liability rules, or staff capacity are missing.

Energy is another risk. CADA’s data-center capacity goal requires power, cooling, land, grid access, and political support. AI infrastructure has to compete with electrification, manufacturing, housing growth, and climate goals. Europe can reduce this conflict through energy-efficient data centers, better location planning, reuse of waste heat, and alignment with renewable power. Those measures help, but they do not erase the demand problem.

Speed adds another risk. U.S. and Chinese firms are moving quickly across model development, chips, cloud platforms, AI agents, robotics, and enterprise deployment. Europe’s plan relies on coordination among institutions, member states, industry, researchers, standards bodies, and regulators. Coordination can produce legitimacy and scale. It can also slow execution. The Commission’s challenge is to convert process into capacity before market positions harden further.

There is also a market-design risk. Publicly funded AI infrastructure must avoid becoming a substitute for private investment. The better model is catalytic: reduce entry barriers, validate early demand, support skills, and help firms reach paying customers. If companies depend too heavily on subsidized access without building revenue, the strategy will generate activity but not competitiveness.

Regulatory credibility creates one more pressure point. The AI Act can become a trust advantage if guidance is clear, enforcement is proportionate, and compliance tools work for smaller firms. It can become a drag if duties feel unpredictable or if national interpretations diverge. The Commission’s simplification work shows awareness of this problem. Firms will judge it by the paperwork, audit demands, and procurement outcomes they experience.

The strongest version of Europe’s AI strategy would make the EU a demanding but attractive market. Firms would build there because they can access compute, high-quality data, sector customers, skilled workers, and a trusted legal framework. Public bodies would buy AI with better contract terms and stronger oversight. Citizens would see AI in services only where performance, rights, and accountability have been tested. That outcome is possible, but it depends less on slogans than on permits, procurement, standards, training, and working systems.

Summary

Europe’s AI strategy has moved from a legal framework into a broader implementation program. The AI Act still anchors the model, but the Commission is now adding compute infrastructure, data access, cloud sovereignty, sector adoption, skills programs, and support tools. This makes the strategy a test of industrial execution rather than policy ambition alone.

The plan’s strongest feature is its recognition that AI adoption depends on the full stack. AI Factories and gigafactories address compute. The Data Union Strategy addresses data. CADA addresses cloud, data centers, and sovereignty. The Apply AI Strategy addresses sector deployment. The AI Office, General-Purpose AI Code of Practice, AI regulatory sandboxes, and standards process address legal implementation. These pieces fit together logically.

The weak point is timing. Europe has to align EU-level plans with member-state permits, grid capacity, procurement systems, private investment, and workforce development. It must also help SMEs comply with the AI Act without turning compliance into a barrier that favors the largest vendors. The strategy can work if it turns public coordination into market capacity. It will fall short if it produces programs that do not change where firms build, train, deploy, and buy AI.

Appendix: Useful Books Available on Amazon

• Artificial Intelligence: A Guide for Thinking Humans
• Human Compatible: AI and the Problem of Control
• The Coming Wave
• Atlas of AI
• Prediction Machines, Updated and Expanded
• Power and Prediction

Appendix: Top Questions Answered in This Article

What Is Europe’s AI Strategy?

Europe’s AI strategy is a combined policy program for AI rules, compute infrastructure, data access, cloud sovereignty, sector adoption, and skills. It includes the AI Act, AI Continent Action Plan, Apply AI Strategy, Data Union Strategy, AI Factories, planned AI gigafactories, and the proposed Cloud and AI Development Act.

What Is the AI Continent Action Plan?

The AI Continent Action Plan is the European Commission’s 2025 framework for building AI capacity across computing infrastructure, data, adoption, skills, and regulatory support. It links AI Factories, gigafactories, InvestAI, the Data Union Strategy, and AI Act support tools into one implementation program.

What Are AI Factories?

AI Factories are European access points linked to EuroHPC supercomputing resources and support services. They help startups, SMEs, researchers, and public organizations train, refine, test, and integrate AI models without needing to build large private computing clusters.

What Are AI Gigafactories?

AI gigafactories are planned large-scale computing facilities for training and developing complex AI models. The Commission’s InvestAI initiative includes a €20 billion European fund for these facilities, with the action plan indicating plans for up to five sites.

What Is the Apply AI Strategy?

The Apply AI Strategy is the EU’s sector adoption plan for AI. It targets strategic industries and the public sector, including healthcare, manufacturing, energy, mobility, communications, culture, security, and space. Its purpose is to move AI from experimentation into operational use.

What Is the Cloud and AI Development Act?

The Cloud and AI Development Act is a 2026 European Commission proposal designed to expand sustainable data center capacity, support cloud and AI investment, and create a sovereignty framework for cloud and AI services. It also addresses permitting, infrastructure, procurement, and public-sector resilience.

How Does the AI Act Fit Into Europe’s Strategy?

The AI Act sets the legal framework for trustworthy AI in the EU. It defines duties for providers and deployers, bans certain practices, regulates high-risk systems, and creates rules for general-purpose AI. The wider strategy adds infrastructure and support so firms can comply and deploy.

Why Is Data Central to the Strategy?

AI systems need high-quality data for training, testing, validation, and deployment. The Data Union Strategy seeks to expand data access, streamline rules, support data labs, and develop common European data spaces so firms and researchers can build domain-specific AI systems.

How Does the Strategy Affect SMEs?

SMEs benefit if shared compute, data access, test facilities, sandboxes, and guidance reduce the cost of AI adoption. They face risk if compliance duties remain expensive or confusing. The strategy’s SME value depends on practical access, simple procedures, and sector support.

Why Does the Space Economy Matter to Europe’s AI Strategy?

The space economy uses AI for Earth observation analysis, satellite operations, autonomy, anomaly detection, and secure data services. Europe’s space programs already show the sovereignty model that AI policy now follows: shared infrastructure, public demand, commercial services, and rule-based trust.

Appendix: Glossary of Key Terms

Artificial Intelligence (AI)

Artificial intelligence refers to software systems that perform tasks usually associated with human reasoning, prediction, perception, language, planning, or decision support. In this article, the term covers model training, generative AI, industrial AI, public-sector AI, and AI tools used in sectors such as space, health, and manufacturing.

European Union (EU)

The European Union is a political and economic union of member states that shares laws, institutions, funding programs, and single-market rules. In AI policy, the EU sets common legal frameworks and coordinates programs that member states implement through national authorities, funding bodies, and public buyers.

Small and Medium-Sized Enterprises (SMEs)

Small and medium-sized enterprises are firms below defined thresholds for employees, revenue, or balance sheet size. EU AI policy gives them special attention because they often lack the compute budgets, legal teams, and technical staff that larger technology firms use when adopting AI.

AI Continent Action Plan

The AI Continent Action Plan is the European Commission’s framework for turning Europe into a stronger AI developer and adopter. It links compute, data, sector deployment, skills, investment, and AI Act implementation into a common policy design.

AI Factories

AI Factories are EU-supported access points connected to AI-optimized supercomputing resources. They provide computing capacity and support services for startups, SMEs, researchers, and public organizations that need help training, refining, testing, or integrating AI models.

AI Gigafactories

AI gigafactories are planned large-scale facilities intended to support development of complex AI models. They sit above standard AI Factories in scale and are linked to the InvestAI initiative, which includes a European fund for gigafactory investment.

General-Purpose AI

General-purpose AI refers to models that can support many tasks rather than one narrow application. Under the AI Act, providers of these models face dedicated obligations, and the most capable models can face stronger expectations tied to systemic risk.

Data Union Strategy

The Data Union Strategy is the Commission’s plan to improve access to high-quality data for AI, simplify data rules, and protect European data sovereignty. It includes data labs, common European data spaces, and measures to support international data flows on trusted terms.

Cloud and AI Sovereignty

Cloud and AI sovereignty describes the degree of control, assurance, and resilience a public body or organization has over cloud infrastructure, data processing, software supply chains, and AI services. CADA proposes assurance levels to guide public-sector risk assessments and procurement.

High-Risk AI System

A high-risk AI system is an AI application covered by stricter duties under the AI Act because of its potential effects on safety, health, rights, public services, employment, education, border control, infrastructure, or regulated products. These systems require stronger documentation, oversight, and testing.