Understanding Digital Sovereignty in the Age of Artificial Intelligence

The Network Law Review is pleased to present a special issue on “Industrial Policy and Competitiveness,” prepared in collaboration with the International Center for Law & Economics (ICLE). This issue gathers leading scholars to explore a central question: What are the boundaries between competition and industrial policy?

Abstract: EU Digital Sovereignty has emerged as a key policy topic amid growing geopolitical tensions fuelled by an AI arms race in which corporate investment and government support appear mutually reinforcing. This essay examines the meaning of digital sovereignty and the trade-offs associated with the EU’s ability to act as a first-class participant in the global digital economy with the autonomy and independence required to control its destiny (i.e., digital future). This essay calls for more holistic analyses of digital sovereignty and more nuanced policies predicated on evidence-based decision-making rather than the appeal of simplistic or selective narratives.

1. Introduction

For the European Union (EU), digital sovereignty implies the ability to act as a first-class participant in the global digital economy with the autonomy and independence required to control its destiny (i.e., digital future).[1] To achieve this, EU businesses and citizens will need to be able to access best-in-class, resilient, secure, and trustworthy digital infrastructures. As we move toward a future in which almost everything may be augmented and connected to an evolving range of digital technologies, all imaginable social and economic contexts may be affected, directly or indirectly, with or without direct human attention or awareness (Stocker & Lehr, 2025a, b). The impact of these technologies on different stakeholders across all sectors depends on an evolving fabric of technical and business relationships among users and those who design, deploy, and operate the hardware/software that enable our networked telecommunications and computing products and services.

Softwarization is shifting the location of functionality, creating industry and firm boundaries that are more fluid and dynamic. Softwarization enables virtualization, which facilitates the dynamic separation and unbundling of resource use and control, with the promise of enabling “everything as a service” (XaaS).[2] These business capabilities depend on a panoply of technologies supported as a layered stack of digital platforms (Lehr et al., 2019). Although most of the digital infrastructure components predate current incarnations of AI, they are increasingly being integrated into the mix. With AI’s rise to prominence in mass consciousness, concerns over sovereign control of what has come to be known as the AI stack of components have come to the fore.

Moreover, although digital infrastructures are comprised of a mix of components provided and managed locally within the confines of a Member State, a significant proportion of digital components are provided by multinationals. This raises legitimate concerns regarding the role of a handful of large global players in a more consolidated Internet and AI ecosystem with respect to control over multiple core components of our digital infrastructures, ranging from cloud services to web hosting and the DNS (Domain Name System).[3] As this handful of non-EU global players has developed highly scalable and extensible technology substrates that underpin a large fraction of economic activity, they have become increasingly essential for social and economic activity.[4] Many of the resulting dependencies may not be apparent or visible, with the consequence that quantifying risks resulting from them is hard and requires context-aware assessments.[5] Finally, rising global tensions and foreign (external to the EU) threats, such as war, growing income disparities, and the rising power of economic rivals, are matched by the reduced trustworthiness of traditional alliances. Collectively, these forces have amplified legitimate concerns over the EU’s digital sovereignty. Concurrently and inextricably bound to those concerns is the growing need for an EU (digital) Single Market.

Amid growing geopolitical tensions fuelled by an AI arms race in which corporate investment and government support appear mutually reinforcing, the 2024 Draghi Report on EU Competitiveness (Draghi, 2024) and Draghi’s September 2025 speech (Draghi, 2025) have clarified that the EU is caught between a rock and a hard place. Beyond a variety of existing deficiencies and unresolved issues related to fragmentation and non-aligned Member State interests, AI has revealed profound, structural deficiencies that have caused the EU to fall behind other jurisdictions in terms of innovation, growth and, ultimately, competitiveness. The inability to control the EU’s digital infrastructure (quality and cost) or rely on trusted third parties (US) creates risks to digital sovereignty. What is true for the EU is even more true for Member States – these lack the scale and capabilities to self-provision the digital infrastructure needed for their economy, and their level of integration with neighbors (via trade, travel, etc.) is too great to be isolated.

2. The Sovereignty Conundrum

Digital sovereignty is not dichotomous. Instead, it exists on a continuum between self-sufficiency and total dependency. Reductionist views of digital sovereignty that aim to locate supply chains entirely within the EU are misguided and counterproductive.[6] It does not make sense to impose rules that aim to achieve complete independence from non-EU vendors, products, infrastructure, or algorithms. Although full self-sufficiency and a return to heavy-handed protection of infant industries is neither feasible nor desirable at the EU or Member State level, even tilting towards such a position may be dangerous. Such autonomy would be predicated on isolationism, without mutually beneficial technical, commercial, or political relationships. Even in milder forms, which are arguably more likely to emerge since trade-offs will vary across different domains and contexts, such developments would imply deviations from a global trading system where comparative advantage has enabled countries to benefit from the scale, scope and specialization enabled by a global economy.[7]

Inherent to isolationism is the replication of digital infrastructures by Member States (or potentially coalitions of Member States), fragmenting the Single Market and resulting in duplicate and wasteful investment that could have been directed towards other (more fruitful) areas. Although such an approach entails duplicated investment and socio-economic inefficiencies, it would provide a Member State (or coalition) with a degree of autonomy once it had developed its own capabilities. However, the development of those capabilities will take years once two related questions have been answered: what domestic capabilities should be developed, and which ones should be done first?

Equally damaging are shifts to protectionist trade policies. The proponents of protectionist policies seek to justify them on the basis of how they speedily refocus the provision of digital infrastructure. Inherent to these policies are inefficiencies. Fragmenting markets reduces opportunities to realize scale/scope economies, encourages sub-optimal investments, and leads to favoring a few selected actors. Beyond the undesirable outcomes of reinforced hierarchies and power structures among EU companies and the potential ossification of industry structures that will almost certainly increase intra-EU tensions between favored and non-favored Member States, it is likely that companies will be encouraged toward (sycophant) firm behavior to strategically seek regulatory support/protection[8] as well as to strategically market and label their products as sovereign.[9]

The essence of sovereignty is about having independence in control of one’s destiny. Self-sufficiency might seem to offer such independence; but it is both infeasible and undesirable in a global digital economy.[10] Inherent to choice is the ability to pick among alternatives which obviously depends on the availability of alternatives. With respect to digital sovereignty this would mean that the switching costs are practically and financially manageable and that viable and acceptable alternatives exist.

The implications of defining choice in this way are that if the EU or Member State aspires to become independent of, for example, large technology companies, then a replacement scenario emerges where EU technology infrastructures and services are developed to replace their non-EU counterparts. But in which contexts will this be reasonable? What does independence mean in the specific context? And who will these companies be? If you look at the current Single Market, it is not unreasonable to suggest that those who are currently the most dominant within the EU will continue to be. The continued dominance will generate tensions within the EU, in terms of the competitiveness of these companies vis-à-vis non-EU players on one hand and strategic autonomy and independence versus intra-EU competition on the other.

The link between less fragmented and more consolidated EU markets and more innovation, which we feel is often (axiomatically) assumed across several relevant contexts, will not necessarily hold in the longer run if incumbents with entrenched positions are under insufficient competitive pressures within the EU to innovate. If intra-EU competition is ‘sacrificed’ for EU competitiveness with the US and China, we cannot know what the long-term effects will be.

Regardless of the outcome of debates at the EU-level, the potential for an EU company to evolve into a world-class competitor and the goals of EU integration may be promoted by consolidation, but this may raise the specter of rent-seeking behavior among EU Member States and their national champions, thereby undermining Single Market goals. Those Member States who feel they have been marginalized in consolidation discussions may be unwilling to participate in EU-wide initiatives, preferring instead to either develop their own digital capabilities or continue using non-EU alternatives. In other words, the availability of an EU technology could be viewed as being unacceptable to some Member States if they regard it as being imposed. A technology solution that is imposed on a Member State threatens the Member State’s digital sovereignty since it provides it with neither choice nor control, continuing its reliance on technologies developed elsewhere (even if the provider would be considered “like-minded” and neither a Chinese nor US-based company).

3. The Scope of Digital Sovereignty

The digital economy is comprised of a complex and dynamic set of activities. At its most basic, it combines infrastructure with services. The liberalization of the telecommunications industry, coupled with widespread and sustained innovation, have created a global digital economy where its various components – infrastructure, devices and services – are intertwined but often provided separately by companies.

An extreme interpretation of digital sovereignty would mean different things at different levels of the digital stack. For example, striving to recreate the halcyon days of the late 1990s and early 2000s when Nokia dominated global handset markets would be extremely expensive and likely doomed to fail in light of the entrenched positions and resources available to Apple, Samsung, and a number of Chinese vendors. Moreover, the trend has been toward relying ever more heavily on low-cost, commodity hardware with limited margin potential. While ensuring that EU businesses are not overly dependent on any single source of commodity hardware, there is little economic advantage to the EU in trying to focus on meeting its needs from EU sources. Given the challenges associated with handsets and their relatively low value creation potential, services are clearly more attractive. However, not all services are similar. For example, in search, winner-take-all economics have long rendered attacks on Google’s dominance of search unsuccessful.

However, prospects for EU competitors emerging in AI are more complicated. First, the components of the AI stack including LLMs/foundation models and higher-level applications are subject to different scale and scope economies as well as innovation dynamics. Second, the inter-dependencies of AI on non-AI digital infrastructures such as the telecommunications, cloud, and data center services span both national and international components. EU-sourcing of the basic hardware (from chips to commodity digital servers) and operating systems is not available today and would require years and substantial investment to accomplish even it were feasible. EU prospects for producing world-class competitors for AI algorithms are better. Although that will also require significant investment, a number of EU companies are already active, and a spate of recent announcements make EU prospects for some of these look bright.[11]

At the level of AI end-user application development, the diversity of business contexts and potential for AI-innovation activity and the implications for user-generated innovations render the prospects for EU competitors emerging quite promising, so long as the ability of those EU businesses to have access to best-in-class digital infrastructure services (mostly not AI) exists. The growth of EU AI-native competitors that may emerge and hope to compete successfully in global markets will need to prosper in their local (national) and then EU-level markets, as well as to tap into global markets whenever possible. Thus, too much of an emphasis on self-sufficiency within a Member State or within the EU for AI components and services (especially lower-level components) may damage prospects for nascent home-grown AI-native companies drawn from across the entire EU economy becoming world-class global competitors.

The scale of the investment that is needed means that not all companies will receive the financial support that they need to grow and develop, which in turn suggests that not every Member State will be the home to a large AI company. If digital sovereignty is interpreted as the development of capabilities, then this will result in the EU developing the relevant capabilities to provide AI services, but these will be unevenly spread across the bloc. The development of EU capabilities will also enable Member States to choose which provider to adopt, with competition occurring between both internal and external providers. If digital sovereignty is viewed as the development of capabilities within a Member State, then it is unlikely that a full range of AI services will develop within each one. While there are EU companies that have demonstrated world-class competitive capabilities and the possibility of new competitors emerging in the EU is reasonable to expect, it is unreasonable to expect that best-in-class alternatives will and should be limited to EU sourcing.[12] Thus, Member States will combine national, EU and non-EU AI to fulfil their requirements.[13]

Reflecting the desire of the EU to develop its own AI capabilities, coupled with the more general growth in the market, alliances and partnerships between EU and non-EU companies have been established.[14] There are many benefits to these partnerships. The non-EU companies invest in the EU, tying the bloc to the wider global digital economy and creating jobs in the process. This investment brings with it new capabilities, enhancing the ability of EU companies to innovate and generate socio-economic value. These investments may be seen as threats to digital sovereignty – not only do they integrate the EU with the wider global economy, but they may also delay the development of the bloc’s indigenous AI capabilities. Having said that, they could be viewed as an acceptable means of kickstarting AI within the EU. Moreover, if the EU supported strategic alliances and partnerships, it could help hasten the development of AI capabilities and thus lessen the bloc’s long-term dependence on non-EU companies. Of course, deciding which alliances and partnerships to support raises issues around who picks the winners, the extent to which this process is political, and where they would be located.

If the approach is to combine non-EU with EU-based AI capabilities, which parts should be done within the EU and which should be ‘bought in’ from elsewhere? In many respects, this is a ‘make or buy’ decision, with the EU and/or Member State deciding the extent to which they integrate with the global AI market. In industries characterized by low levels of innovation, or whose scope or dynamics are well understood, determining what to make or buy is relatively straightforward. But AI is neither of these. As a consequence, decision-making challenges are fraught with difficulties.[15]

4. Concluding Thoughts

We should strive to converge towards a North Star vision of what digital sovereignty means and implies.[16] Dependencies, their nature, as well as their social and economic impact, vary considerably across domains and contexts. We should be skeptical of overly simplistic, reductionist narratives about dependencies and one-size-fits-all solutions. Moreover, as digital sovereignty is a moving target shaped by evolving dependencies, responses require careful, regular reassessment. Migration paths towards achieving digital sovereignty differ across contexts and evolve as technologies, dependencies, and other relevant factors continue to evolve.

While we recognize the need for EU-level industrial policy and regulatory action to address legitimate concerns over threats to digital sovereignty, a key consideration needs to be how those relate to the EU’s need for a Digital Single Market and the internal tensions that will arise naturally as a result of within-EU competition and concerns over Member State sovereignty. Make no mistake: the EU has reached the state we are in now while establishing a wide range of digital policies and regulations. As we have learned, these past interventions did not get us to the Digital Single Market or global competitiveness desired. We should learn from the past.

Regulatory reforms should focus on industrial policies that will contribute to nurturing an innovation ecosystem that is not unduly constrained by internal or external factors. The key to digital sovereignty will depend on private investment and entrepreneurial activity within EU Member States. Although public investment at the Member State and EU-level will be needed, it needs to be carefully targeted so as to complement rather than substitute for private investment. The emphasis should be on removing roadblocks for stronger EU competition to emerge, which will require ensuring that abuses of market power to stifle nascent competition are addressed, but without preference regarding the national identity of abusers. That sort of competitive policy neutrality will be needed if the EU is to achieve its goal to create an EU Digital Single Market. It is inconsistent with protectionist policies designed to protect EU competitors from international competition whether those come from within or outside the EU. However, coordinated EU-level trade policy responses may be called for to respond to protectionist policies by foreign governments. Thus, the capability for EU rather than Member State level trade and other policy responses is needed as credible threats despite the risk they pose to aspirations for open- and free-markets for global trade.

Overly simplistic, false dichotomies and wrong (i.e., outdated) mental models of (ICT and AI-driven) platforms, ecosystems, and markets and their dynamics must be avoided. Similarly, waterbed effects must be anticipated in the course of carefully crafted, holistic strategies. Lehr and Stocker (2024) and Stocker and Lehr (2025a), emphasized as a foundation requirement, the need for a multi-stakeholder measurement ecosystem to enable flexible, evidence-based decision-making. Public authorities in Member States and at the EU-level need to proactively promote the creation of that measurement ecosystem as the best bulwark against rent-seeking or panicked responses, both of which thrive in the face of unaddressed ignorance.[17]

One significant threat to good industrial policies will be a failure to adequately appreciate that the implications of promoting digital sovereignty will manifest itself through different success outcomes and complex feedback implications for EU businesses and society within and across various layers within the AI stack. Scale and scope economies, the structural nature of dependencies and feedback and the optimal migration paths to achieve digital sovereignty differ across layers, contexts, and time. In the box below, we summarize key considerations for a constructive debate about the EU’s digital future:

Sovereignty is a desirable goal, but hard to define and complex to achieve. While inherently hard to define, digital sovereignty should not mean full self-sufficiency at either the Member State or EU-level. With respect to any such goal, ‘Made in EU’ is likely to be better than ‘Made in Country X’, but sovereignty is about EU control over its destiny to the extent that it is feasible. In an interconnected world, individual destiny is collectively determined. Some dependencies may be avoidable, others not. The EU should aspire to control its destiny consistent with protecting the EU and – where they align – global interests. Even amid geopolitical tensions, a core aspiration should be to make the pie bigger for all, and to ensure the EU is a first-class player, not unduly subject to external third-party control.

Self-sufficiency and isolationism are neither feasible nor desirable. Isolation is an extreme solution that undermines the goal of the EU Single Market, which would weaken bloc competitiveness and fail to ensure the goal of a resilient/reliable digital infrastructure. It would undermine incentives encouraging a robust/dynamic industrial market base and runs counter to the Draghi Report. Picking winners and ‘Made in the EU’ requirements for critical infrastructure shrink the opportunities for the firms that the EU hopes to empower to develop its own AI infrastructure and capabilities. It may also prevent the formation of coalitions with ‘like-minded’ non-EU countries. Policies that restrict competition and shield incumbents will prolong stagnation and invite rent-seeking, especially without a clear industrial strategy.

The EU and its Member States should avoid relying on a single supplier for critical infrastructure components. Multiple EU suppliers should compete to provide infrastructure and services, but where bottlenecks exist then regulatory oversight is needed. The threat of competition from self-provisioning options may temper the market power of monopolists – the ability to self-provision need only be sufficiently robust to discipline market power; it need not fully replace a monopolist. The EU should not rely on essential inputs from third parties without developing realistic contingency plans.

Resilience and performance requirements vary across the different layers of digital infrastructures and time scales. Digital infrastructure is layered, interconnected and heterogeneous. Needs for resilience and performance vary across different layers and time scales for different components of digital infrastructure – from chips to equipment to platforms to applications, from hardware to software, etc. Digital systems are layered and interconnected, with each layer/element exhibiting different scale, scope, and learning effects, and evolving at different time scales. The pace of change for AI is particularly fast, complicating any assessment of this particular digital infrastructure to determine what policy(s) the EU and/or Member States should best adopt.

William Lehr*, Volker Stocker** & Jason Whalley***
* MIT CSAIL
** Weizenbaum Institute
*** Northumbria University

Volker Stocker would like to acknowledge funding by the Federal Ministry of Education and Research of Germany (BMBF) under grant No. 16DII141 (Weizenbaum-Institut für die vernetzte Gesellschaft – Das Deutsche Internet-Institut)

Citation: William Lehr, Volker Stocker & Jason Whalley, Understanding Digital Sovereignty in the Age of Artificial Intelligence, Industrial Policy and Competitiveness (ed. Thibault Schrepel & Dirk Auer), Network Law Review, Fall 2025.

References:

Arthur D Little (2022). Digital Economy Outlook 2022, Economic Study for Fédération Française des Télécoms, December. Available at: www.fftelecoms.org
Atomico (2025). State of European Tech 2025, November. Available at: https://www.stateofeuropeantech.com
Curwen, P., Sadowski, B., & Whalley, J. (2015). Where are the Europeans? A longitudinal analysis of TMT companies, info, 17 (5), 1-19.
Draghi, M. (2024). The Draghi Report: A Competitiveness Strategy for Europe. Available at: https://commission.europa.eu/topics/eu-competitiveness/draghi-report_en
Draghi, M. (2025). High Level Conference – One year after the Draghi report: what has been achieved, what has changed. Available at: https://commission.europa.eu/document/download/0951a4ff-cd1a-4ea3-bc1d-f603decc1ed9_en?filename=Draghi_Speech_High_Level_Conference_One_Year_After.pdf
Erixon. F. (2018). The Economic Benefits of Globalization for Business Consumers. European Centre for International Political Economy, 15 February. Available at https://ecipe.org/
European Commission (2025a). EU launches InvestAI initiative to mobilise €200 billion of investment in artificial intelligence, 11 February. Available at https://ec.europa.eu/commission/presscorner/detail/en/ip_25_467
European Commission (2025b). Cloud Sovereignty Framework, Version 1.2.1, October, Directorate-General for Digital services, Luxembourg
European Parliament (2023). Benefits of economic globalisation in Europe: facts and figures, 17 July. Available at https://www.europarl.europa.eu/topics/en/article/20190603STO53520/benefits-of-economic-globalisation-in-eu-facts-and-figures
Frias, Z., W. Lehr, & Stocker, V. (2025). Building an ecosystem for mobile broadband measurement: Methods and policy challenges. Telecommunications Policy, 49(5), 102905. ISSN 0308-5961, https://doi.org/10.1016/j.telpol.2025.102905
Garcia, D.C. (2025) Funding the AI economy: Strengthening Europe’s investment capacity, EU Start-ups, 17 November. Available at: https://www.eu-startups.com
Habib, R., Ruth, K., Akiwate, G., & Durumeric, Z. (2025, September). Formalizing Dependence of Web Infrastructure. In Proceedings of the ACM SIGCOMM 2025 Conference (pp. 1132-1153). Available at: https://kcruth.com/papers/2025-Centralization.pdf
Haeck, P. & Van den Hove, A. (2025, December). Backlash builds against EU Commission’s ‘Buy European’ push. POLITICO, 5 December. Available at: https://www.politico.eu/article/backlash-builds-against-commissions-buy-european-push/
La Rocco, N. (2025). Nvidia, Telekom, SAP: Deutsche KI-Cloud mit 10.000 GPUs geht Anfang 2026 online. Computer Base, 4 November. Available at: https://www.computerbase.de/news/internet/nvidia-telekom-sap-deutsche-ki-cloud-mit-10-000-gpus-geht-anfang-2026-online.94903/
Lehr, W., D. Clark, S. Bauer, & Claffy, kc (2019). Regulation when Platforms are Layered. TPRC47: Research Conference on Communications, Information and Internet Policy, Washington DC, September 2019. Available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3427499
Lehr, W.H., & Stocker, V. (2024). Competition Policy over the Generative AI Waterfall. In: A. Abbott & T. Schrepel (eds.) AI and Competition Policy, Concurrences (pp. 335-358).
Madiega, T. (2020). Digital sovereignty for Europe. European Parliamentary Research Service, July. Available at www.europarl.europa.eu
OpenAI (2025). AP and OpenAI partner to launch sovereign ‘OpenAI for Germany’. 24 September. Available at: https://openai.com/global-affairs/openai-for-germany/
Panenkov, D. (2025). Cloud sovereignty is now fashionable. But most such offerings are anything but. Tech Monitor. 5 November. Available at: https://www.techmonitor.ai/comment-2/cloud-sovereignty-deals-false/?cf-view
Rona, S. & Levy, S. (2025). The state of AI industry trends in Europe: Talent drives success, but US funding still crucial, Silicon Valley Bank, 15 April. Available at: www.svb.com
SAP (2025) AWS and SAP expand collaboration to advance digital sovereignty across Europe, 24 September. Available at: https://news.sap.com/2025/09/aws-sap-expand-collaboration-advance-digital-sovereignty-europe/
Stocker, V., & Lehr, W.H. (2025a). The Growing Complexity of Digital Economies over the GenAI Waterfall: Challenges and Policy Implications. Network Law Review. Available at: https://www.networklawreview.org/stocker-lehr-ecosystem/
Stocker, V., & Lehr, W.H. (2025b). Principal-Agent Dynamics and Digital (Platform) Economics in the Age of Agentic AI. Network Law Review. Available at: https://www.networklawreview.org/stocker-lehr-ai/
UNCTAD (2025) World Investment Report: International Investment in the Digital Economy, United Nations Conference on Trade and Development.
Whalley, J. & Curwen, P. (2024) Creating value from 5G: The challenge for mobile operators, Telecommunications Policy, 48(2), 102647.

Footnotes:

[1] In a research note for the European Parliament, Madiega (2020) states “digital sovereignty refers to Europe’s ability to act independently in the digital world” (page 1). In forthcoming work, we unpack further the implications of more nuanced views of Digital Sovereignty. For example, concerns over data sovereignty (i.e., EU governance over EU data) motivate calls for infrastructure sovereignty (e.g., capability to host EU data on EU platforms to protect it from third party access) (see Panenkov, 2025). The U.S. Clarifying Lawful Overseas Use of Data (CLOUD) Act (2018) provides a substantive basis for such concerns by implementing an avenue for U.S. authorities to access private data stored on cloud platforms, regardless of where the relevant servers are located physically. (See https://www.justice.gov/criminal/cloud-act-resources for additional information; visited: 9-12-2025).
[2] XaaS enables greater business flexibility for just-in-time, make-vs-buy decision-making across value chains, both within and across the tech sectors that provide the digital products and services and the non-tech sectors that use those digital technologies. Within the technical architectures, it makes it feasible for centralized and decentralized, dedicated and shared, private and public systems to compete with varying and changing degrees of success across business contexts.
[3] See, for example, Stocker & Lehr (2025a, b) and Habib et al. (2025).
[4] See Stocker & Lehr (2025a) and note that many of these large players offer devices like smartphones but also IoT devices that are deployed deep within home networks and are increasingly compute and AI assisted. Some players are also active in the industrial IoT space.
[5] Undertaking such evidence-based assessments requires access to techno-economic measurement data that requires the existence of a measurement ecosystem (see Frias et al., 2025).
[6] No serious scholar believes that true autarky is a viable or desirable solution for addressing Digital Sovereignty concerns, yet like “perfect competition” (which also exists only as a theoretical construct), it may serve as a standard against which the current state and alternatives are judged.
[7] The benefits of globalization to the EU are significant, with 38 million jobs across Member States being dependent on exports to non-EU countries (European Parliament, 2023). For an illustration of the broad array of benefits associated with globalization see, among others, Erixon (2018). The economic significance of the digital economy is highlighted by UNCTAD (2025), where international investment grew unlike that of broader global foreign direct investment in 2024.
[8] The EU Commission’s reported plans for a ‘Buy European’ initiative, which would include an EU preference for public procurement, has reportedly created pushback from within the EU and also calls for exemptions for “like-minded” non-EU countries like Japan (Haeck and Van den Hove, 2025).
[9] We already observe EU players selling their solutions as EU alternative or “made in EU” while critically relying on inputs from non-EU countries. A recent example is the partnership between German tech company SAP and OpenAI to launch “sovereign ‘OpenAI for Germany’” (OpenAI, 2025). The recent joint venture between DT and NVIDIA for building an “Industrial AI Cloud” was presented recently. The presentation slides made a statement about sovereignty at different layers and levels, distinguishing between a “data” (Germany), “operational” (Germany + EU), and an upstream “technology” (Germany + US) layer (La Rocco, 2025). Panenkov (2025) refers to sovereignty washed products as “Trojan horses – appearing compliant on the surface while leaving critical levers of control outside local reach” and that ‘sovereign’ cloud solutions may be “creating an illusion of independence while maintaining deep dependencies on dominant global players.”
[10] While independence and autonomy at the EU-level are observable themes in the European Commission’s recent cloud sovereignty framework (European Commission, 2025b), the framework hints at the complexity and scope of achieving this in practice and thus the difficulties of achieving it.
[11] See, for example, Atomico (2025), European Commission (2025a), Garcia (2025) or Rona & Levy (2025) for illustrations of the acknowledgment that significant investment is required and where it is being directed towards.
[12] For example, ASML is a leading player and among the most valuable companies in the EU, but is one among a handful of companies. The paucity of EU-based companies is evident in, among others, Arthur D Little (2022), Curwen et al. (2015) or Whalley and Curwen (2024).
[13] It is worth remembering that today, firms across the EU currently use cloud services from AWS, Google Cloud, and Microsoft Azure and consumers are regularly using social media and messaging services provided by leading non-EU tech firms which often have embedded consumer-facing AI tools, providing non-EU firms with access to data stocks and flows via end-user interfaces and contributing to the training of major AI models and development of AI tools (see Stocker & Lehr, 2025b). Those entanglements do pose sovereignty risks at multiple levels, but severing those relationships is neither practical nor desirable. Providing viable EU alternatives where those are feasible and reducing reliance on single providers everywhere are reasonable responses to enhance sovereignty, but the risks will remain.
[14] For example, the recent alliance between AWS and SAP (SAP, 2025).
[15] For example, as the EU and/or Member State moves towards digital sovereignty one decision that it will need to grapple with is how to diversify from where and whom it obtains AI related activities. This diversification facilitates choice, but at the same time complicates the decision-making processes: it increases the likelihood that too much time will be spent deliberating what to do with the consequence that the EU and/or Member State is playing ‘catch-up’ or that mistakes are made. As mistakes only appear with time, there is a need to ensure that the EU and/or Member State has contingency plans in place and that it is not reliant on a single company or country.
[16] This is necessarily not a complete roadmap, but an agreement on the direction of progress. Navigating by the North Star may allow travelers to differentiate between northern and southern paths yet offer insufficient guidance to differentiate between north-northeast and north-northwest directions.
[17] Asymmetric private information renders both deterrence of bad behavior and coordination of good behavior more difficult, but it is also key to innovation and competition. Transparency and information sharing can help address asymmetric information problems but needs to be balanced with privacy and security concerns and recognition that addressing ignorance is costly and the future is always uncertain.