48.8566° N · 2.3522° E
BRIEFING 01 — ACTIVE 14 PP.

The Geopolitics of Artificial Intelligence.
A field report on digital sovereignty.

Fewer than twelve computers in the world are large enough to train frontier AI models. Roughly 80% of nations lack sufficient AI infrastructure within their borders. This is the architecture of dependence — and the strategy of those negotiating its terms.

90%
RARE-EARTH REFINING CONTROLLED BY CHINA
70%
GLOBAL CLOUD HELD BY AWS, AZURE, GOOGLE
70%
OF FOUNDATIONAL AI MODELS BUILT IN THE US
BEGIN BRIEFING
REF · DS-2026-Q2 · TRANSMISSION OPEN
01 THE CONDITION

Digital dependence in the age of AI.

Artificial intelligence is assuming a growing importance in geopolitics. Its role is evolving to the point where it directly impacts how states promote and develop their national and foreign policies. As discussed in the course, AI capability for a state is ultimately a production function of distinct inputs:

PRODUCTION FUNCTION
A = f(C, E, D) × T × η
CCOMPUTEChips, GPUs, fabs
EENERGYBaseload for data centres
DDATATraining corpora & rights
TTALENTResearchers & engineers
ηEFFICIENCYHow inputs combine

The states and firms that control these critical inputs of AI are accumulating structural advantages over those that do not, making digital dependence a pressing security concern.

This is not a prospective concern. The United States and China already dominate most layers of the AI infrastructure stack, from chip design and fabrication to the cloud platforms upon which models are trained and deployed. China controls around 90% of global rare earth refining capacity, essential for the hardware underpinning all digital systems, while Amazon, Microsoft, and Google alone account for roughly 70% of the global cloud market. The models themselves follow the same pattern: 70% of foundational AI systems have been developed in the United States and 15% in China (Bria et al., 2025)0. For most nations, this means that the chips, data centres, and platforms their economies depend on are designed, owned, and governed elsewhere. This creates deep structural dependencies and clear imbalance in the distribution of AI power.

For those who do not control all these inputs, the question is no longer whether to engage with foreign systems, but how to do so, with whom, on what terms and at what cost. These choices are being made now, driving innovation, growth and the modernization of public services. This strategic context is what leads to the question: is it possible for middle power countries to achieve digital sovereignty?

02 DEFINING THE TERM

Sovereignty under conditions of constraint.

Digital sovereignty is a state's domestic capacity to exercise meaningful, informed, and politically accountable control over its digital infrastructures, data regimes and governance rules — while managing international partnerships to overcome unavoidable external dependencies in line with its own legal frameworks, public interests and strategic priorities.

This definition's framing deliberately focuses on states. The infrastructure it describes is largely built and operated by private actors whose decisions carry geopolitical consequences. That dimension is not ignored here but sovereignty, as a legal and political concept, remains a state function.

No major intergovernmental body has adopted a single binding definition of digital sovereignty, and none explicitly uses the term as a formal concept. The institutions chosen represent broad and authoritative cross-sections of intergovernmental thinking on digital governance. As such, they offer a set of adjacent frameworks that can help address the problem from different entry points.

OECDGovern in a digital environment, not just digitise bureaucracy.1STATE CAPACITY
UNCTADData governance and structural terms of participation.2MATERIAL INPUTS
UN GDCOpen, secure, rights-grounded digital future for all.3GLOBAL COOPERATION
WORLD BANKShared, open foundations reducing proprietary lock-in.4INFRASTRUCTURE

Takeaway: These frameworks do not provide a single agreed definition of digital sovereignty, but they do point in the same direction. All converge on the same underlying concern: ensuring that states retain enough agency to make meaningful choices about the digital systems that structure public and economic life, rather than simply inheriting arrangements designed by and for others.

Because the term lacks a universally accepted definition, it is contested and justified in a myriad of ways across the policy board globally. The term of digital sovereignty is often invoked by actors pursuing fundamentally incompatible projects — such as states asserting infrastructure control, firms marketing compliance products, and movements resisting corporate surveillance — with no shared metric to determine when any of them has succeeded5. Pohle and Thiel (2020) observe that the sovereignty discourse tends to appear most forcefully precisely where authority is weakest, making it as much a political speech act as a policy programme6. Santaniello (2025) argues that the resulting gap between rhetoric and implementation is an inherent attribute of sovereignty claims7. States can keep invoking digital sovereignty even when their policies fail to deliver it because the claim itself serves political functions — mobilising support, signalling intent, and framing national identity — which are independent of whether the underlying condition is actually achieved. The concept is politically self-sustaining even in the face of its own shortcomings.

The sophistication of the question lies in whether a state can still make meaningful choices about where data flows, whose infrastructures shape public life, what legal protections apply and how dependencies are managed. Digital sovereignty is less about total independence than about governance under conditions of constraint. It is precisely because these inputs are finite, unevenly distributed, and increasingly contested that digital sovereignty has become an operational priority for states. Because states must operationally position themselves and define their AI strategies in response to concrete physical constraints, digital sovereignty retains meaning regardless of the theoretical ambiguity that surrounds it.

03 THE STRATEGIC POSITION

Middle powers: room to act, not enough to escape.

"When we [middle powers] only negotiate bilaterally with a hegemon, we negotiate from weakness… This is not sovereignty. It's the performance of sovereignty while accepting subordination… In a world of great power rivalry, the countries in between have a choice: compete with each other for favour, or combine to create a third path with impact."

MARK CARNEY · WORLD ECONOMIC FORUM, DAVOS 20268

While great powers can absorb the costs of technological rivalry more easily because they control key firms, infrastructures, and strategic chokepoints, smaller states often have little choice but to adapt to their external circumstances. Middle powers sit in between these two poles. They are integrated enough in the global digital economy to be a commercially meaningful market that deploys this infrastructure, yet they lack the resources to fully develop them at home.

States that fall under this definition of a middle power are typically shaped by varying levers, resources, and constraints, despite navigating similar strategic choices. With enough room to act strategically but not enough to combat all dependencies, their strategies are a result of their specific combination of assets and vulnerabilities.

Those assets can be organised into five broad categories:

01 · MARKET POWER
02 · REGULATORY & NORMATIVE AUTHORITY
03 · DIGITAL PUBLIC INFRASTRUCTURE
04 · PHYSICAL ENDOWMENTS
05 · DIPLOMATIC POSITIONING

Market power considers the size and affluence of the domestic market — to what extent a state can use access to its market as a compliance lever over foreign firms. Regulatory and normative authority focuses on state capacity to set rules that others follow as a condition of market access, which can substitute for hardware control when enforced at sufficient scale. Digital public infrastructure (DPI) refers to states that have built open, interoperable DPI at scale — consisting normally of identity, payment and data exchange systems — on which both public and private applications can run, reducing dependence on closed proprietary systems and the lock-in that accompanies them. Physical endowments considers energy availability, renewable capacity, critical mineral reserves, and geographic positioning, which determine which parts of the AI stack a state can credibly claim to support or supply. Diplomatic and institutional positioning focuses on the state's ability to use multilateral forums, standard-setting bodies, and bilateral partnerships to shape governance norms and diversify dependencies, even in the absence of independent productive capacity.

Their dilemma is how to manage strategic dependencies. They must decide which partnerships to pursue, which vulnerabilities to reduce and where to build leverage. The strategies they adopt are analytically understood as a result of the specific combination of assets and vulnerabilities of each middle power.

04 FIELD STUDIES

Unlock the implications of digital sovereignty.

The following case studies examine how middle powers operationalise digital sovereignty under conditions of structural dependence. Rather than treating digital sovereignty as a single policy model to be replicated, the cases are analysed as different configurations of the same underlying problem: how states that lack full control over the AI stack manage dependence, build leverage where possible, and prioritise specific layers of the digital ecosystem according to their economic, developmental, and geopolitical objectives.

Each case mobilises a distinct combination of assets while accepting persistent vulnerabilities. Read comparatively, the cases illustrate that digital sovereignty for middle powers is not a question of absolute autonomy, but of strategic choice: deciding where sovereignty matters most, which dependencies are tolerable, and how leverage can be accumulated over time within the constraints of an unequal global digital order.

01LOCKED
🔒
European Union
STRATEGIC AUTONOMY
02LOCKED
🔒
Kenya
LEADERSHIP THROUGH CREDIBILITY
03LOCKED
🔒
India
ATMANIRBHAR BHARAT
04LOCKED
🔒
Brazil
SMART INTERDEPENDENCE
50.85° N · 4.35° E · BRUSSELS
European Union — Regulating power in a world it does not build
The EU's strategy has long relied on its ability to leverage its large, affluent single market to compel foreign firms to comply with European regulatory standards, a mechanism known as the Brussels Effect (Bradford, 2012)9. Instruments such as the GDPR, the DMA, and the DSA extend beyond consumer protection alone to function as tools of normative influence, turning the EU the world's foremost regulatory power in the digital domain. However, the EU's share of global GDP has fallen from 30% in 2008 to 21% today and is projected to reach 16% by 2050 (Draghi, 2024), directly eroding the cost-of-bypass calculation. The AI Act has not yet achieved equivalent reach1011. Its enforcement is inconsistent across member states and jurisdictions that adopted GDPR-aligned frameworks are now taking a wait-and-see approach before following Europe's lead again. Overall, the EU's approach reveals an appetite for governing technologies it does not fully control — a paradox at the heart of the European digital project1314. Now, the bloc is trying to move beyond regulation alone, using industrial policy and external partnerships to reduce critical dependencies and strengthen its position across other layers of the digital stack. This tension makes the EU a particularly instructive case for understanding digital sovereignty not as a fixed objective to be attained, but as a dilemma to be continuously managed.
LEVERAGE POINTS
01
REGULATORY MARKET POWER & STANDARD-SETTING
A market of 450 million people and around €18 trillion in GDP is too large to ignore15. Any firm interested in EU market access must meet compliance obligations. The GDPR became a global privacy standard; the AI Act is now shaping AI governance conversations far beyond Europe16. By embedding values (privacy, transparency, interoperability) into legally binding frameworks, the EU shapes what "legitimate" AI looks like internationally.
02
ASML SEMICONDUCTOR CHOKEPOINT
The Dutch firm ASML holds a near-monopoly on extreme ultraviolet (EUV) lithography machines — the tools needed to manufacture the world's most advanced chips. No country, including the US or China, can produce cutting-edge semiconductors without them. This is Europe's single most powerful hardware chokepoint in the global AI supply chain.
03
ENERGY ASSETS
Although the EU is far from energy independent, it does have meaningful sources of green energy, the importance of which increases as geopolitical volatility continues to put pressure on oil and gas supplies. France's nuclear fleet provides large volumes of firm-power electricity attractive to AI infrastructure, with the country exporting a record 90 TWh of decarbonised electricity in 20241718. These assets are increasingly strategic as global data centre electricity demand surged 17% in 2025 and is projected to grow from 460 TWh in 2024 to over 1,000 TWh by 20301920. However, these assets are national and not collectively pooled, with no common procurement mechanism to convert them into shared digital infrastructure.
04
CAPITAL MOBILISATION
The AI Continent Action Plan (2025) commits €200 billion for AI development, including €20 billion via the InvestAI facility for up to 5 gigafactories and 19 AI factories already operational, targeting a tripling of EU data-centre capacity within 5–7 years212223. The financing architecture has been confirmed by the European Investment Bank24.
05
MISTRAL AI — PROOF OF CONCEPT
Mistral AI has grown to a valuation exceeding $14 billion and raised $830 million in March 2026 to build a 44 MW data centre near Paris, with a total European infrastructure budget of €4 billion25. It functions as Europe's only operational frontier AI company with dedicated sovereign compute infrastructure under construction26. It holds important European significance, but its primary identity and regulatory home is French.
06
HUMAN CAPITAL
Europe boasts strong universities with 37.22% of the population having completed tertiary education27. However, only a little over a quarter pursue degrees in STEM fields28.
07
DIVERSIFICATION THROUGH PARTNERSHIPS
The EU's 2025 International Digital Strategy seeks to combine regulation, public-private investment and digital cooperation through a "Team Europe" approach2930. Global Gateway uses infrastructure finance and digital partnerships to project influence. The EU is pursuing digital trade agreements with Canada, mutual data adequacy with Brazil, and deepening digital economy collaboration with India313233.
DEPENDENCIES & LIMITS
01
CLOUD IN FOREIGN HANDS
European cloud providers hold less than 15% of their domestic market, with that share shrinking34. AWS, Azure and Google Cloud are US companies subject to the US CLOUD Act35.
02
FRAGMENTED INTERNAL MARKET
With 24 official languages and 27 member states with divergent industrial strategies, energy policies and risk tolerances, collective action is slow and costly.
03
SHALLOW CAPITAL MARKETS
US-listed companies achieve valuations 3.32 times higher than European counterparts36. Between 2014 and 2023, VC investment amounted to EUR 89 billion in the EU, compared with over EUR 1,000 billion in the US37. The Draghi Report identified this as a structural weakness38.
04
ASML'S CHOKEPOINT IS NOT PERMANENT
China is investing heavily in domestic alternatives to EUV lithography. The lead Europe holds through ASML is real but time-limited.
05
ENERGY DEPENDENCY
In 2024, the EU still imported 57% of its energy needs39, with major dependencies on oil and gas suppliers, presenting an indirect but real vulnerability for further infrastructure build-out.

Regulatory-industrial balancing — using legal weight to shape the rules of the game while simultaneously trying to diversify and build enough productive capacity to play. Brussels is simultaneously trying to do multiple things at once: retain its normative authority through regulation and partnerships while preventing regulatory complexity from becoming a competitive liability and innovation barrier through simplification initiatives and asset leveraging. Whether it can do both simultaneously without losing rule-making credibility in the process is the central strategic question.

WHY THIS STRATEGY? USAGE OF ASSETS

Since the EU lacks domestically dominant hyperscalers and frontier AI firms, its most credible source of influence lies not in production but in rule-setting. This structural constraint explains why the EU prioritises regulatory and normative authority and market power, while starting to expand its diplomatic and institutional positioning, as substitutes for hardware and platform control, rather than attempting immediate technological parity with the United States or China.

1.2921° S · 36.8219° E · NAIROBI
Kenya — The sovereignty paradox of the Silicon Savannah
Kenya is a small, lower-middle-income East African economy of 55 million people that has become the most advanced digital economy in East Africa and one of the continent's clearest cases of digital sovereignty pursued through visibility rather than ownership. Kenya's National AI Strategy 2025–2030 calls for AI development centres, a national dataset initiative, localised AI models, and governance reform40. Kenya is effectively converting the credibility of its DPI ecosystem into geopolitical positioning, while remaining structurally reliant on the foreign capital and platforms that make that ecosystem run.
LEVERAGE POINTS
01
M-PESA AND DPI AS GEOPOLITICAL CURRENCY
M-Pesa reaches over 90% of Kenyan adults and processes over 70% of the country's digital payment transactions. The government has developed eCitizen, hosting 16,000 public services, and NOFBI, a national fibre backbone connecting all 47 counties4142.
02
RENEWABLE ENERGY COMPARATIVE ADVANTAGE
In 2023/24, 83% of electricity came from renewable sources, with geothermal alone meeting 41.71% of demand43. Microsoft and G42 announced a new East Africa cloud region in Kenya, to be powered by geothermal energy. This is one of Kenya's clearest structural advantages for data-centre development.
03
PAPSS INTEGRATION & REGIONAL POSITIONING
AI for development is central in Kenya, where digital systems are closely tied to public service delivery and economic transformation. The February 2026 integration of PesaLink into PAPSS connects over 80 Kenyan financial institutions to 160+ institutions across Africa in real-time local-currency settlement, bypassing dollar-dominated correspondent banking44. Afreximbank estimates $5 billion in continental savings from full PAPSS rollout.
04
ICT SECTOR MOMENTUM
Kenya's ICT sector has grown at 10.8% annually over the past decade45. Kenyan startups attracted close to $800 million in funding in 2023. The country attracts intra-African tech talent from Nigeria and Uganda.
05
DIPLOMATIC POSITIONING & AU ROLE
Kenya has claimed the role of Africa's "digital capital" by hosting PAPSS headquarters, occupying the AU Champion on Institutional Reform role, and parlaying its DPI track record into a seat at global AI governance conversations.
DEPENDENCIES & LIMITS
01
COMPUTE IS FOREIGN-OWNED
The Olkaria campus, announced in May 2024 as Kenya's flagship AI infrastructure project and to be built by Microsoft and UAE-based G42 on KenGen's geothermal land in Naivasha, has stalled due to grid capacity constraints and the absence of National Treasury financial approval. Kenya's only domestic semiconductor initiative, the STL nanofabrication facility at Dedan Kimathi University in Nyeri, produces legacy chips for testing and research rather than the advanced processors that AI infrastructure requires. A $1.3 million US feasibility grant awarded to STL in May 2024 under the Biden administration's CHIPS Act to study a potential fabrication facility has an uncertain future, as the Trump administration has redirected international semiconductor funding toward domestic US production46474849. Kenya currently relies on import of advanced semiconductors to power its AI ambitions. Supporters of Kenya's semiconductor ambitions may lean on the country's mineral wealth as a pillar for semiconductor manufacturing. However, most of the value in the semiconductor industry lies far beyond extraction, in processing and fabrication.
02
STATE DATA ON FOREIGN CLOUDS
Despite the Cloud-First Policy's localist framing, sensitive government databases sit on Microsoft Azure, AWS, and Huawei Cloud50. The National Cloud Data Centre at Konza Technopolis remains underutilised — symbolising the gap between aspiration and operational reality.
03
WEAK DATA GOVERNANCE
The Data Protection Act of 2019 established oversight but administrative fines are capped at approximately $50,000, low enough that compliance costs often exceed penalty risk51. The government has simultaneously expanded surveillance powers via the October 2025 Computer Misuse and Cybercrimes Amendment Act52.
04
NO BINDING AI LEGAL FRAMEWORK
As of early 2026, no AI-specific legislation is in force. The Robotics and AI Society Bill remains pending. The National AI Strategy functions as a political document, not an enforceable framework5354.
05
TALENT DRAIN WITH NO RETENTION MECHANISM
Kenya produces strong engineers but lacks binding retention mechanisms, diaspora equity programmes, or return-migration incentives. Silicon Valley, London, and Gulf hubs consistently absorb Kenyan talent at compensation levels the domestic market cannot match.

DPI-as-diplomatic-leverage: use the credibility of operational public digital infrastructure to claim governance authority and attract foreign investment, while managing rather than resolving the underlying structural dependencies. The gap between hosting infrastructure and owning it remains unresolved.

WHY THIS STRATEGY? USAGE OF ASSETS

For Kenya, digital sovereignty is shaped by the constraints of a small market and limited capital, which make ownership of frontier AI infrastructure unrealistic in the short term. The state prioritises visibility, credibility, physical endowments and diplomatic positioning through operational DPI, using these assets to attract investment and governance influence rather than attempting costly technological self-sufficiency.

28.6139° N · 77.2090° E · NEW DELHI
India — The greenfield sovereign that still imports its stack
India frames digital sovereignty through an explicitly techno-nationalist lens. The "Atmanirbhar Bharat" (self-reliant India) doctrine links technological independence to national security, economic growth, and an anti-colonial narrative drawing explicit parallels between historical data extraction by US tech firms and the resource extraction of British colonial rule55. India's DPI has delivered measurable results, raising the country to the top of the digitalisation index of revenue administrations among emerging market economies565758. This "greenfield" model — designing systems digitally from the outset rather than patching legacy ones — has been India's most reproducible export.
LEVERAGE POINTS
01
OPERATIONAL HOMEGROWN DPI
Aadhaar (biometric identity for over 1.3 billion people), UPI (the world's fifth-largest payment network, accounting for 70% of India's digital payment transactions), and DigiLocker596061. UPI is live in the UAE, Bhutan, Oman, and Singapore; pilot-tested in France, Nepal, Mauritius, and Sri Lanka. MOSIP, a non-profit initiative hosted by the International Institute of Information Technology, Bangalore (IITB), is adopted by over twenty countries with 121 million+ active users62.
02
SCALE OF DOMESTIC MARKET
With 1.4 billion people, India offers a market large enough to sustain domestically-oriented platforms. The economic value added from DPIs to India's GDP is estimated to reach 2.9–4.2% by 2030 (from 0.9% in 2022)6364.
03
PUBLIC COMPUTE — INDIAAI MISSION
Launched in 2023 and spearheaded by the Ministry of Electronics and Information Technology (MeitY), the IndiaAI Mission focuses on developing AI applications to address societal challenges in healthcare, education, agriculture, and smart cities while promoting responsible and ethical AI development65. It is responsible for deploying more than 34,000 GPUs by 2025, creating one of the largest public AI compute networks globally. This reduces dependence of Indian startups on private US hyperscalers.
04
BHASHINI — MULTILINGUAL DATA SOVEREIGNTY
Develops NLP models for India's 22 official languages using open-source frameworks66. This addresses a structural gap — most frontier models perform poorly on non-English languages — while building an asset foreign models cannot easily replicate.
05
TALENT DEPTH & GLOBAL POSITIONING
96% of Indian professionals use AI tools at work, significantly above the 81% US rate67. India's IITs and IISc actively contribute to global open-source AI research.
06
STRATEGIC POSITIONING FOR SEMICONDUCTORS & ENERGY
India is developing its own semiconductor manufacturing capacity under the India Semiconductor Mission (2021)68. Its long-term carbon-neutrality roadmap (net-zero by 2070) is driving investment in nuclear baseload power via the SHANTI Bill, specifically designed to support AI infrastructure by supporting India's long-term vision of expanding clean, reliable energy69. India has set an ambitious target of achieving 100 GW of nuclear capacity by 2047.
07
SUPPLY-CHAIN PARTNERSHIPS
India signed MoUs with Australia, Argentina, Brazil and the DRC for lithium and cobalt70717273. It joined the G7-led Mineral Security Partnership in 2023 and has an EU Trade and Technology Council7475.
08
DPI AS GLOBAL NORM EXPORT
India used its G20 presidency in 2023 to globalise the DPI concept by launching the Global Digital Public Infrastructure Repository (GDPIR). It has signed bilateral MoUs with eleven countries including Kenya, Colombia, and Tanzania.
DEPENDENCIES & LIMITS
01
SEMICONDUCTOR IMPORT DEPENDENCE
India imports over 95% of its semiconductors76. When Microsoft temporarily blocked services to Nayara Energy in 2025, a single corporate decision made abroad disrupted a major Indian company with no domestic recourse. India remains 100% import-dependent for energy transition minerals such as lithium, cobalt and nickel, even though it has sizable reserves of cobalt (44.9 Mt ore), nickel (189 Mt), copper (163.9 Mt), and graphite7778.
02
FOREIGN CLOUD INFRASTRUCTURE
Microsoft, Amazon, and Google collectively announced over $67 billion in India investments but they bring capacity, not sovereignty. Data centres in Hyderabad, Mumbai, and Pune remain subject to US jurisdiction, including the CLOUD Act.
03
TALENT DRAIN TO SILICON VALLEY
India produces exceptional AI engineers and exports a disproportionate share. The Atmanirbhar rhetoric has not been matched by policies capable of retaining the talent needed to make it real.
04
DATA PROTECTION LAGGING
India has pushed for data localisation through the DPDP Act, but its data protection standards remain significantly weaker than the EU's GDPR or China's Data Security Law79. The gap between asserting data sovereignty and protecting data is a credibility problem.
05
THE OPEN-SOURCE DILEMMA
The practical path to AI capability runs through fine-tuning Meta's open-source Llama models rather than building from scratch80. This is faster and cheaper but means India's AI "sovereignty" rests on a foundation it did not build — undermining the Atmanirbhar narrative.

Greenfield DPI with selective sovereignty: build from scratch where strategically viable (identity, payments, language infrastructure) and accept managed dependence where it is not (chips, cloud, foundation models). The goal is not to escape dependence but to accumulate enough domestic capability to negotiate its terms.

WHY THIS STRATEGY? USAGE OF ASSETS

India's prioritisation of digital public infrastructure is driven less by industrial competition with frontier AI powers than by the need to scale public service delivery across a vast population with uneven state capacity. Sovereignty at the application and governance layers matters more politically than sovereignty over chips or foundation models. This explains India's willingness to accept managed dependence at the hardware layer, leveraging its market power to attract foreign investment and using selective diplomatic positioning to secure inputs and extend the reach of its DPI model.

15.7975° S · 47.8919° W · BRASÍLIA
Brazil — The Interdependent Brownfield State that Reuses While Diversifying Partners
Rather than asserting independence from foreign infrastructure or building from scratch, Brazil has pursued what its government explicitly calls a "brownfield" strategy: upgrading, connecting, and repurposing existing legacy systems into a unified digital public infrastructure, while simultaneously using multilateral diplomacy to diversify external dependencies. This is tied to a broader agenda of state transformation, in which digital infrastructure is linked to social, economic and environmental missions defined through interministerial plans, with the Ministry of Management and Innovation in Public Services (MGI) playing a coordinating role. The National Strategy for Digital Government 2024–2027 formalises this direction by emphasising inclusion, interoperability, secure data reuse, and simpler access to services81.
LEVERAGE POINTS
01
OPERATIONAL DPI AT SCALE
Gov.br (160 million users, 12,000+ integrated public services), Pix (150 million+ users, Brazil's dominant payment method), and Conecta Gov.br form the most operationally complete DPI ecosystem outside of India and China828384.
02
CLIMATE DATA AS A SOVEREIGN ASSET
CAR, a georeferenced registry covering 7 million rural properties and 700 million hectares, has been relaunched as Brazil's first official Digital Public Good during COP3085. Combined with INPE's deforestation monitoring systems (DETER, PRODES)86, Brazil has a data stock that is genuinely difficult to replicate elsewhere.
03
RENEWABLE ENERGY ADVANTAGE
Over 80% of Brazil's electricity comes from renewable sources87. This gives Brazil a comparative advantage for powering AI infrastructure with low carbon intensity — directly relevant to data centre investment.
04
CRITICAL MINERALS POSITIONING
Brazil holds significant reserves of lithium, nickel, copper, graphite, rare earths, and niobium. The Ministry of Mines and Energy's 2026 investor guide explicitly links these reserves to higher value-added processing ambitions88.
05
ELLALINK CONNECTIVITY
The 2021 EllaLink submarine fibre-optic cable connecting Brazil directly to Europe reduces dependence on North Atlantic data routes and lowers latency for transatlantic digital exchange89.
06
MULTILATERAL DIPLOMATIC POSITIONING
Brazil leveraged the 2024 G20 presidency, 2025 BRICS presidency, and COP30 to project its climate-relevant systems outward through Climate DPI and ClimateStack90. It signed a DPI MoU with India, participates in 50-in-5, partnered with Norway on CAR reform, and deepened EU cooperation through the EU-LAC Digital Alliance and the BELLA Programme9192939495.
DEPENDENCIES & LIMITS
01
FOREIGN CLOUD & ADVANCED CHIPS
Brazil's 205 data centres include AWS, Google Cloud, and Microsoft Azure, but supercomputing runs on imported NVIDIA GPUs and AMD/Intel CPUs. Brazil has no domestic advanced semiconductor production. The ReData tax-incentive regime incentivised local investment but not local ownership96.
02
FRAGMENTED IMPLEMENTATION
The brownfield strategy depends on interoperability between federal, state, and municipal systems, but implementation is uneven. Remote and rural areas face weak connectivity and high infrastructure costs.
03
AI TALENT SHORTFALL
Brazil has significant institutional capacity (INPE, LNCC)9798 but faces shortages in advanced digital and AI skills. Brasscom estimated a projected deficit of over 500,000 ICT professionals99. Public-sector AI capability is still being built: MGI and national agencies launched a civil servant AI training track in 2024, especially in procurement, implementation, auditing, and long-term governance of complex digital systems.
04
ENVIRONMENTAL TENSIONS
As digital infrastructure expands, its escalating energy and water demands are coming into sharper tension with the adverse effects experienced by local traditional communities.
05
POLICY WITHOUT STRUCTURAL TRANSFORMATION
The PBIA 2024–2028 and ReData regime are real commitments, but they stop short of addressing deeper hardware dependency100. PBIA proposes a new supercomputer and AI excellence centres, yet the effort remains oriented toward expanding access and national capability rather than becoming a domestic producer of cutting-edge chips or server infrastructure. ReData itself was enacted by Provisional Presidential Measure, expired in February 2026, and is currently awaiting Congressional approval.

Brownfield DPI with diplomatic diversification: build what is strategic domestically, reuse what already works rather than rebuilding from scratch, and diversify external ties to avoid lock-in by any single partner. The brownfield logic generates leverage from existing systems rather than requiring the massive capital outlays that greenfield approaches demand.

WHY THIS STRATEGY? USAGE OF ASSETS

Brazil's strategy reflects the political and administrative realities of a large federal state with extensive legacy systems and uneven territorial capacity. Rather than pursuing a greenfield overhaul, Brazil prioritises upgrading and integrating existing systems into its digital public infrastructure (DPI), avoiding the fiscal and political costs of rebuilding from scratch. Simultaneously, it leverages its market power and physical endowments to attract investment, while benefiting from its diplomatic positioning to diversify partnerships and avoid dependencies.

05 COMPARISON MATRIX

A posture of smart interdependence.

🇪🇺 EUSTRATEGIC AUTONOMY
HOW IT FRAMES SOVEREIGNTYRule-setting power and strategic autonomy: shaping the conditions under which digital technologies operate, while diversifying and building productive capacity
MAIN LEVERAGEBrussels Effect; ASML chokepoint; nuclear energy; standard-setting abroad; large market and capital mobilisation; Global Gateway partnerships
MAIN DEPENDENCIESUS cloud infrastructure; no hyperscalers; talent drain; fragmented market; capital gap; weak frontier AI production; energy dependence; exposure to US strategic shifts
DOMINANT STRATEGYRegulatory-industrial balancing: buy time with normative power while building compute, cloud and AI capacity
MAIN LIMITATIONConfused standard-setting with capability-building for too long; risks dismantling regulatory leverage before alternatives are ready
🇰🇪 KenyaLEADERSHIP THROUGH CREDIBILITY
HOW IT FRAMES SOVEREIGNTYVisibility, credibility, and governance influence built through operational DPI and attraction of foreign investment without ownership of the full stack
MAIN LEVERAGEM-Pesa / eCitizen; NOFBI; renewable electricity; PAPSS hub role; diplomatic visibility as East Africa's digital hub
MAIN DEPENDENCIESForeign-owned compute (Microsoft/G42); imported chips; state data on US/Chinese clouds; weak data governance; no binding AI law; external capital; talent drain
DOMINANT STRATEGYDPI-as-diplomatic-leverage: convert operational credibility into continental and global positioning
MAIN LIMITATIONHosting foreign infrastructure is not owning it; energy advantage attracts foreign capital but no sovereign productive capacity yet
🇮🇳 IndiaATMANIRBHAR BHARAT
HOW IT FRAMES SOVEREIGNTYTechno-nationalist sovereign AI framed as anti-colonial self-reliance: build national DPI and AI capability to control key governance and application layers
MAIN LEVERAGEIndia Stack DPI (Aadhaar, UPI, DigiLocker); Bhashini; IndiaAI Mission public compute; global DPI diplomacy; huge domestic market; talent depth; expanding strategic positioning for semiconductors
MAIN DEPENDENCIES95%+ semiconductor import dependence; foreign hyperscalers dominate cloud; talent drain; weak data protection; reliance on foreign open-source foundation models
DOMINANT STRATEGYGreenfield DPI with selective sovereignty: build where strategically valuable, accept managed dependence elsewhere
MAIN LIMITATIONAtmanirbhar rhetoric outpaces structural reality — sovereignty claimed at the application layer while the hardware layer remains foreign
🇧🇷 BrazilSMART INTERDEPENDENCE
HOW IT FRAMES SOVEREIGNTYDomestic DPI assets built reusing existing systems ("brownfield approach"), while diversifying external partnerships rather than accepting lock-in
MAIN LEVERAGEBrownfield DPI at scale (Gov.br, Pix, CAR); EllaLink connectivity; renewable electricity; climate data; critical minerals; multilateral diplomacy (G20, BRICS, COP30)
MAIN DEPENDENCIESForeign cloud and advanced chips; uneven territorial implementation; AI talent deficit; politically contingent legislation; environmental tensions
DOMINANT STRATEGYBrownfield DPI with diplomatic diversification: upgrade existing systems, link to social/climate missions, project model internationally
MAIN LIMITATIONPolitical contingency and implementation unevenness — coherent strategy that depends on administrative continuity and cannot yet close the compute sovereignty gap
06 CONCLUSION

A posture of smart interdependence.

The four cases examined here each represent a different configuration of smart interdependence. The EU wields regulatory market power to shape the rules governing technologies it does not build, buying time to develop productive capacity while risking the erosion of its normative leverage before industrial alternatives are ready. India builds public digital infrastructure from the ground up at the application layer while remaining structurally dependent on foreign chips, cloud, and foundation models at the hardware layer. Kenya converts the operational credibility of its public digital systems into diplomatic positioning and foreign investment, building the institutional profile of a digital leader while the underlying compute remains foreign-owned. Brazil upgrades existing systems, diversifying external partnerships and using critical minerals and renewable energy as strategic assets in what its own government terms a brownfield approach, but also without controlling the hardware and cloud layers beneath it and with uneven state capacity across its territory.

What emerges from this analysis is a family of approaches, each shaped by local capabilities and constraints, yet all recognisable as variations of the same underlying posture: smart interdependence — understood as the deliberate identification, prioritisation, and governance of dependencies in ways that build leverage over time101. Not all dependencies carry the same strategic risk. Dependence on foreign semiconductors carries fundamentally different risks than dependence on foreign governance norms.

Keohane and Nye (1977) identified this logic decades ago, arguing that what matters in any relationship of interdependence is not the dependency itself, but who bears the higher cost of walking away102. The middle powers examined here are not trying to escape dependence. What the cases reveal is a pattern of prioritisation, with states investing in the layers where domestic capacity is feasible and leverage is buildable, while accepting the structural dependencies they cannot yet address103.

THREE OPERATING PRINCIPLES
01BUILD LEVERAGEWhere capacity, energy, data, or rules give a real edge.
02ACCEPT DEPENDENCEWhere structural reality makes alternatives illusory.
03INVEST TO RENEGOTIATEIn the institutional muscle to revise terms over time.

None of these cases has achieved digital sovereignty in any absolute sense. What the cases collectively demonstrate is that digital sovereignty is not a condition to be reached but a capacity to be continuously built, defended, and renegotiated across five strategic levers (market power, regulatory authority, digital public infrastructure, physical endowments, and diplomatic positioning). Compute and semiconductor dependencies still stand out as the hardest to address and the most politically consequential, with none of the middle powers examined here meaningfully resolving them. However, they are accumulating the leverage needed to negotiate these dependencies rather than simply inheriting them.

Smart interdependence is therefore less a policy model than a strategic orientation — moving beyond both the illusion of full sovereignty and the passivity of unmanaged dependence, toward a posture in which states build leverage where they can, accept dependence where they must, and invest continuously in the capacity to renegotiate both. Recent debates frame this less as self-sufficiency than as a more collaborative form of sovereignty, achieved through structured cooperation rather than isolation.

Digital sovereignty retains political force precisely because of the physical nature of dependencies today. States that treat interdependence as something to organise, secure, and negotiate — rather than something to either deny or surrender to.
07 SOURCES & FOOTNOTES

[0] Bria, F. et al. 2025. Foundational AI systems distribution data.

[1] OECD 2020. Digital Government Policy Framework

[2] UNCTAD 2024. Digital Economy Report

[3] UN Global Digital Compact 2024. unglobalcompact.org

[4] World Bank 2025. DPI and Development Report

[5] Couture & Toupin 2019. doi.org/10.1177/1461444819865984

[6] Pohle & Thiel 2020. doi.org/10.14763/2020.4.1532

[7] Santaniello 2025. doi.org/10.1080/14650045.2025.2521548

[8] Mark Carney, WEF Davos 2026. weforum.org

[9] Bradford 2012. The Brussels Effect, OUP

[10] Bradford 2020. The Brussels Effect. Oxford University Press.

[11] EU AI Act national implementation. artificialintelligenceact.eu

[13] EC AI Continent Action Plan. commission.europa.eu

[14] EC International Digital Strategy. digital-strategy.ec.europa.eu

[15] EU Facts and Figures. european-union.europa.eu

[16] Science Direct 2025. sciencedirect.com

[17] Data Center Dynamics. France AI data centre build-out

[18] IEA 2025. Data centre electricity use surged in 2025

[19] IEA 2025. Data centre electricity use

[20] IEA. Energy supply for AI

[21] EC AI Continent Action Plan. commission.europa.eu

[22] AI Factories. digital-strategy.ec.europa.eu

[23] AI Action Plan factpage. digital-strategy.ec.europa.eu

[24] EIB Group & European Commission. EIB press: AI Gigafactories

[25] Mistral AI. mistral.ai

[26] Mistral Europe. europe.mistral.ai

[27] OECD. Population with tertiary education

[28] EC. STEM in tertiary education

[29] EC. International Digital Strategy 2025

[30] Team Europe Initiatives. international-partnerships.ec.europa.eu

[31] EC press: EU-Canada digital trade. commission.europa.eu

[32] EC press: EU-Brazil data adequacy. commission.europa.eu

[33] EU-India FTA summary. trade.ec.europa.eu

[34] CNBC 2026. Europe's reliance on US digital infrastructure

[35] CLOUD Act impact. alstonprivacy.com

[36] ECB 2024. ECB working paper op373

[37] Banque de France. Europe VC market

[38] Draghi Report 2024. commission.europa.eu

[39] Eurostat Energy 2026. eurostat

[40] Kenya AI Strategy 2025–2030. ict.go.ke

[41] M-Pesa Africa. m-pesa.africa

[42] eCitizen Portal. ecitizen.go.ke

[43] EPRA Kenya Energy Statistics Report FY 2023-2024. epra.go.ke

[44] PesaLink. pesalink.co.ke

[45] Kenya Digital Economy. trade.gov

[46] Semafor 2026. Kenya Microsoft data centre stalled

[47] Dedan Kimathi University STL. dkut.ac.ke

[48] USTDA Kenya semiconductors. ustda.gov

[49] Manufacturing Dive. CHIPS Act future

[50] Kenya Cloud Policy 2024. ict.go.ke

[51] Kenya Data Protection Act 2019. kentrade.go.ke

[52] Computer Misuse and Cybercrimes (Amendment) Act 2025. parliament.go.ke

[53] Kenya Robotics and AI Society Bill 2023. dataguidance.com

[54] Kenya AI Strategy 2025–2030. ict.go.ke

[55] Atmanirbhar Bharat. tribal.nic.in

[56] DIKSHA. indiastack.global

[57] ABDM. india.gov.in

[58] VISTAAR. vistaar.da.gov.in

[59] Aadhaar. uidai.gov.in

[60] UPI. npci.org.in

[61] DigiLocker. digilocker.gov.in

[62] MOSIP. mosip.io

[63] Nasscom DPI Report. nasscom.in

[64] Institut Montaigne. institutmontaigne.org

[65] IndiaAI Mission. indiaai.gov.in

[66] Bhashini. bhashini.gov.in

[67] Emeritus 2025. Global Workplace Skills Study

[68] India Semiconductor Mission. ism.gov.in

[69] SHANTI Bill. pib.gov.in

[70] IEEFA: India critical minerals strategy. ieefa.org

[71] MEA India Annual Report. mea.gov.in

[72] India-Brazil MoU. mea.gov.in

[73] IEEFA: India critical minerals. ieefa.org

[74] Mineral Security Partnership (MSP). state.gov

[75] EU-India TTC. digital-strategy.ec.europa.eu

[76] Bluekraft: India's Semiconductor Age. bluekraft.in

[77] IEEFA: India's Hunt for Critical Minerals. ieefa.org

[78] IEEFA: Critical minerals import dependency. ieefa.org

[79] DPDP Act analysis. pib.gov.in

[80] Meta Llama. llama.com

[81] Brazil Digital Government Strategy. gov.br

[82] Gov.br. gov.br

[83] Pix. bcb.gov.br

[84] Conecta Gov.br. gov.br

[85] Amazon Fund: CAR. amazonfund.gov.br

[86] PRODES/DETER systems. infoamazonia.org

[87] EPE: Brazilian Energy Balance 2025. epe.gov.br

[88] Brazil Mining Investor Guide 2026. gov.br/mme

[89] EllaLink activation. gov.br/mcom

[90] ITS Rio: DPI for Climate Report. itsrio.org

[91] India-Brazil DPI MoU.

[92] 50-in-5 Campaign. 50in5.net

[93] 50-in-5 Brazil profile. 50in5.net

[94] EU-LAC Digital Alliance. eeas.europa.eu

[95] BELLA Programme. eeas.europa.eu

[96] ReData regime. gov.br/mme

[97] INPE. gov.br/inpe

[98] LNCC. gov.br/lncc

[99] Brasscom: ICT talent deficit. brasscom.org.br

[100] PBIA 2024-2028. gov.br/mcti

[101] GovTech Intelligence Hub: Digital Sovereignty & the EuroStack. govtechintelhub.org

[102] Keohane, R.O. and Nye, J.S. (1977). Power and Interdependence. Little, Brown.

[103] ECDPM: Brazil to the World. ecdpm.org