Sovereign AI has become a national priority, replacing the old dream of a borderless internet with state-backed control over computing hardware. The digital economy is fragmenting along geopolitical lines, as nations reshape global digital infrastructure around competing tech alliances. Power in this new landscape no longer comes from controlling territory alone. Instead, it flows through critical nodes in the AI ecosystem where technical, material, and institutional dependencies converge: semiconductor supply chains, cloud platforms, biometric systems, and technical standards.
In the AI era, sovereignty means the ability to grant, deny, or control access to the infrastructure of governance itself. This control operates through three layers: compute (processing data), connectivity (moving data), and storage (holding data), all of which depend on energy, water, hardware, and capital. The scale is massive. Between 2010 and 2024, global investment in AI infrastructure exceeded $600 billion, and by 2030, annual spending is expected to surpass $400 billion. This spending is concentrated among a few powerful players. Nvidia controls 92 percent of the market for data-center graphics processing units, while TSMC produces roughly 65 to 70 percent of the world’s advanced chips. This concentration of control is not accidental. It results from deliberate investment cycles, intellectual property laws, and geopolitical strategies that favor established powers over newcomers.
This reconfiguration does not follow a single logic. Instead, sovereignty emerges in three distinct forms: State, Corporate, and Indigenous. Each creates its own patterns of power, governance, and vulnerability.
The Infrastructural Shift
Stay ahead of the geopolitical week.
MD Briefing delivers expert analysis across five global fronts — the Indo-Pacific, energy, geoeconomics, European security, and the Middle East — every Monday morning. Free.
Traditional sovereignty, based on the Westphalian model, assumed that state power works through exclusive control over territory. AI changes this logic because sovereignty now flows through layered infrastructure controlled by a few corporate and geopolitical actors. These include semiconductor production, hyperscale cloud platforms, energy-intensive data centers, proprietary software, and international standards bodies.
The key bottlenecks in AI infrastructure are not natural features of technology. They are products of concentrated investment, intellectual property regimes, and geopolitical strategies that systematically benefit established players. Authority is exercised not primarily through laws but through the power to grant, deny, or control access. Nvidia’s dominance, TSMC’s manufacturing control, and the CUDA software standard have become de facto rules that entrench unilateral decision-making. Foundation models deepen this dependency, because training them requires specialized computing power that most nations cannot replicate on their own.
This infrastructural shift extends colonial extraction into the digital age. Just as colonial railways were built to move resources from colonies to empires, today’s digital infrastructure reproduces unequal power relations. Surveillance capitalism captures individual behavior, but this lens is too narrow. The concept of digital colonialism better captures the collective dependency, dispossession, and subordination that result from this system. Algorithmic systems embed colonial power relations within automated decision-making, a process experts term algorithmic colonisation.
However, when states declare digital sovereignty, they often echo old territorial claims, even as their power is mediated by corporate actors that operate across borders with significant independence from state control. This reveals a paradox: state techno-sovereignty seeks to reinforce national autonomy but relies on international supply chains, foreign platforms, and proprietary technologies, creating new dependencies. The result is a three-way contest among state, corporate, and Indigenous forms of techno-sovereignty, each operating through distinct modes of control and resistance. Frameworks like CARE and OCAP ground Indigenous data governance in collective stewardship rather than extraction.
State Techno-Sovereignty
State techno-sovereignty is a nation’s capacity to independently design, develop, and govern artificial intelligence through domestic infrastructure. It requires three building blocks: compute (processing power), connectivity (data networks), and storage (data centers), each depending on energy, water, land, hardware, and capital. Key tools include building domestic computing clusters, developing national language models (LM), requiring data to be stored locally, and using export controls to protect domestic AI ecosystems.
Japan illustrates the strategic push for chip independence. Through Rapidus, Japan is advancing 2nm chip prototyping. The government has designated it as an official business operator under a Ministry of Economy, Trade and Industry (METI) program to ensure stable semiconductor production. Japan recognizes that it cannot safely rely on chips made in Taiwan by TSMC, nor can it depend entirely on the United States.
The European Union pursues regulatory sovereignty through the AI Act’s risk-classification framework and its reach beyond EU borders. This aims to create a Brussels effect, exporting EU standards globally through market power.
The U.S. embraces AI mercantilism. Executive Order 14110 declares the policy to sustain and enhance America’s global AI dominance, while the AI Diffusion Rule tiers access to advanced chips and cloud services based on alignment with U.S. national security interests.
Yet state action faces a structural problem. While claiming autonomy, governments remain dependent on giants like Nvidia, which supplies 92 percent of data-center GPUs, and TSMC, which produces 65 to 70 percent of global foundry output, with ASML monopolizing the extreme ultraviolet lithography machines essential for advanced chip making. For the Global South, this dependency is structural rather than transitional. Territorial regulatory authority cannot replace material access to chip design, fabrication equipment, or cloud capacity.
Corporate Techno-Sovereignty
Corporate techno-sovereignty refers to the sovereign-like authority exercised by AI developers over digital infrastructure. Firms such as OpenAI, Google, and Nvidia govern through proprietary models and concentrated compute ownership. This private power arises because corporations design, produce, sell, and maintain core technologies. They exploit their knowledge advantage over states, a gap widened by the innovative potential of general-purpose technologies, to establish de facto rules in new regulatory fields while shaping market-based technical standards.
The pattern is visible in cloud computing. Amazon Web Services, Microsoft Azure, and Google Cloud centralize data storage and processing, limiting the digital sovereignty of developing nations by making them dependent on foreign infrastructure.
This arrangement crystallizes in what may be called Sovereign AI as a Service. States essentially purchase access to their own technological capability through cloud contracts, giving private firms the authority to set the technical standards that shape national AI ecosystems.
The state-corporate relationship is neither purely adversarial nor purely cooperative. It is bidirectional. States enhance corporate capability through subsidies, procurement, and creating markets to overcome coordination problems arising from network effects and economies of scale. At the same time, states constrain corporate power through antitrust enforcement, as seen in recent litigation against major tech firms and European directives like the General Data Protection Regulation (GDPR).
This interplay takes different forms across political economies. The U.S. tends toward co-optation through defense procurement. The European Union pursues ex ante regulation, as exemplified by GDPR. China enacts state-capitalist integration in semiconductors and frontier AI, treating AI as a high-tech product with an irreducibly material basis.
The sovereignty in question is therefore rarely full independence. It is more often partial technological autonomy, achieved through a calculated division of labor.
Consequently, accountability lags behind corporate velocity, creating severe regulatory fragmentation. The legal asymmetry between the strong regimes of wealthy nations and the weaker frameworks of the Global South is actively exploited. Multinational firms test AI and surveillance technologies in less regulated environments before rolling them out in jurisdictions with stricter oversight.
Neither states nor corporations, however, fully account for a third claimant to AI sovereignty: Indigenous peoples. Their data sovereignty claims derive from prior political existence as free and independent polities and remain structurally absent from both corporate governance and state regulation.
Indigenous People Techno-Sovereignty
Indigenous data sovereignty is the inherent right of Indigenous peoples to govern the collection, ownership, and application of their data. It is grounded in self-determination and the decolonial struggle against extraction. It emerges from centuries of colonial data practices that treated Indigenous communities as objects of study rather than agents of knowledge production. The contemporary ingestion of cultural material into AI training sets reproduces these dispossessive logics at industrial scale.
The principle is fundamentally collective. Data sovereignty is a derivative right of self-determination rather than a technical preference. Its authority extends across all Indigenous nations regardless of formal state recognition.
Two normative frameworks anchor this mode globally. The CARE Principles, Collective Benefit, Authority to Control, Responsibility, and Ethics, center Indigenous rights and community benefit. They challenge the open-data assumption that maximum accessibility equals maximum social good.
Parallel frameworks operate across continents. Māori tino rangatiratanga (sovereign authority) and kaitiakitanga (intergenerational stewardship of data as taonga, or treasure) guide New Zealand’s approach. Canadian OCAP – Ownership, Control, Access, and Possession principles are operationalized by the First Nations Information Governance Centre. Aboriginal Australian frameworks align with OCAP principles, and Arctic Council-style frameworks ground data governance in collective rights rather than individual consent.
Frontier AI is overwhelmingly developed without Indigenous oversight. Datasets are scraped indiscriminately from cultural materials, languages, and images belonging to communities that neither consent nor benefit. Large language models (LLMs) privilege Western scientific and cultural narratives, embedding epistemic hierarchies into global technical standards. In 2024, $252.3 billion in global AI investment flowed within a small cluster of advanced economies and corporations, virtually excluding Indigenous representation from governance.
The result is a new colonialism. Ancestral heritage, language, and relational knowledge are digitized and enclosed without Free, Prior, and Informed Consent protocols adapted to digital environments.
The alternative is stewardship rather than command. Federated architectures, data trusts, and community-controlled repositories embed governance within social relations rather than market or state imperatives. Resistance becomes part of the infrastructure itself. Examples show this is possible. Te Hiku Media’s Te Reo Māori speech recognition model achieved 92 percent accuracy under community leadership. The First Languages AI Reality Project demonstrates algorithmic sovereignty in practice.
Integrating Indigenous knowledge systems is therefore not an ethical add-on but a structural reorientation of how AI relates to ecological and collective governance. It requires decision-making authority rather than consultation, permanent Indigenous representation in standards-setting bodies, and institutionalizing pluralistic participation.
Interactions and Multi-Sovereign Governance
State and corporate techno-sovereignties are locked in a mutually reinforcing antagonism. The Silicon Sovereignty Paradox shows that states increasingly depend on advanced AI for strategic purposes, from military applications to digital public infrastructure, while private firms control the upstream compute and foundational models on which those capabilities rest.
Outdated power hierarchies fail to capture this fluidity. States and AI corporations simultaneously reinforce and undermine each other’s power through financial, infrastructural, and political mechanisms. The result is that state nationalism and platform capitalism contest rule-making authority over data flow, technical standards, and chip-export thresholds.
The Nvidia-TSMC-ASML chokepoint is the structural site of this antagonism. Nvidia supplies 92 percent of data-center GPUs, TSMC produces 65 to 70 percent of foundry output, and ASML monopolizes extreme ultraviolet lithography.
State-Indigenous relations are equally contested. National development agendas regularly override Indigenous consent. For example, administrative data systems undermine Māori tino rangatiratanga by extracting data without the Crown consultation that Te Tiriti requires. Yet co-governance remains possible when states recognize Indigenous data sovereignty in national law, implement consent-based frameworks, and institutionalize permanent Indigenous representation in standards-setting bodies.
Corporate-Indigenous encounters constitute the sharpest extractive frontier. Global AI investment of $252.3 billion in 2024 flowed within a small cluster of actors who scraped cultural materials and languages without Free, Prior, and Informed Consent. At the same time, Indigenous-led projects such as Te Hiku Media’s Te Reo Māori speech recognition model and an Indigenous-planned Alberta data center show algorithmic sovereignty under community control.
The Core Conflict and Path Forward
The tripartite core conflict, among states, multinational corporations, and publics, is infrastructural. Regulatory frameworks focused downstream, on chatbots, algorithmic outputs, or application-layer risks, leave the upstream ownership of foundational compute, hyperscale cloud, and training data untouched.
Voluntary corporate ethics are insufficient. The path forward requires enforceable architectures that reconfigure ownership and access upstream through two principles:
- Subsidiarity and digital federalism: data trusts and federated systems operate locally, state coordination works nationally, and pluralistic standards-setting functions globally. This establishes the general procedural architecture for distributed governance.
- Indigenous data sovereignty frameworks should be legally integrated into global AI treaties, not merely as a sub-category of local governance, but as a binding recognition of pre-existing collective rights over ancestral knowledge, biocultural heritage, and genomic data. Injecting a justice-based, historically grounded carve-out that generic federalism alone cannot provide.
A genuinely hybrid, distributed model, one that captures economic value of AI while distributing infrastructural power across multiple stakeholders, is the only structural alternative capable of prioritizing democratic legitimacy over corporate or state monopoly. That is the imperative this moment demands.
