AI Sovereignty

AI Sovereignty is the geopolitical doctrine that a nation or bloc must possess autonomous control over the full AI value chain — from semiconductor fabrication and datacenter infrastructure to foundation models, training data, and regulatory frameworks — in order to maintain strategic, economic, and cultural autonomy. While Sovereign AI describes the specific technology programs nations are building (models like Mistral, Falcon, and Bhashini), AI Sovereignty is the broader political and economic framework that motivates those programs. It asserts that AI is not merely a technology sector but a layer of national infrastructure as critical as energy, finance, and defense — and that dependency on foreign AI systems constitutes a strategic vulnerability comparable to dependency on foreign oil.

The Geopolitics of the AI Supply Chain

AI supply chains are irreducibly global, which makes true sovereignty extraordinarily difficult to achieve. Advanced chips are designed in the United States, manufactured in Taiwan and South Korea by TSMC and Samsung, using lithography equipment from the Netherlands' ASML — the only company on Earth capable of producing extreme ultraviolet (EUV) machines. The U.S. has weaponized this supply chain through semiconductor export controls, first under the Biden administration's AI Diffusion Rule in January 2025 (imposing license requirements on AI chip shipments to over 100 countries), and then through ongoing policy adjustments under the Trump administration. China has responded by accelerating domestic semiconductor self-sufficiency, increasing from 15% in 2019 to roughly 25% by 2025, with firms like Biren and Moore Threads anchoring a push toward full-stack AI independence. This chip-level competition is the foundation upon which all other dimensions of AI sovereignty rest: without access to advanced GPU computing, no nation can train frontier models.

From Sovereignty to Resilience

A growing consensus among policy analysts, including work from BCG and MIT Technology Review in early 2026, holds that for most countries, absolute AI sovereignty is an illusion. The realistic alternative is AI resilience: the ability to use, adapt, and govern AI domestically at scale while minimizing strategic dependencies on any single foreign supplier. This reframing is significant because it moves the conversation from autarky (building everything domestically) toward risk management — diversifying chip suppliers, negotiating data-sharing agreements, deploying open-weight models on domestic infrastructure, and building regulatory capacity to govern AI systems regardless of where they originate. The EU's approach exemplifies this: combining the AI Act's regulatory sovereignty with compute investments through the European High-Performance Computing Joint Undertaking, without attempting to replicate Silicon Valley's model-building ecosystem from scratch.

AI Sovereignty and the Agentic Economy

The rise of the agentic economy has transformed AI sovereignty from a government concern into a business imperative. When agentic AI systems don't just generate text but take autonomous actions — executing transactions, managing supply chains, operating critical infrastructure — the question of who controls those agents and under whose jurisdiction they operate becomes urgent. A 2026 survey of enterprise executives found that 95% consider building sovereign AI and data platforms a mission-critical priority within three years. The agentic transition raises the stakes because AI agents operating across borders create novel jurisdictional conflicts: an agent trained in the U.S., deployed on European cloud infrastructure, acting on behalf of a company in Singapore, processing data from Brazilian citizens, must somehow satisfy the sovereignty requirements of all four jurisdictions simultaneously. Governance, visibility, and compliance become exponentially more complex when AI doesn't just advise — it acts.

The U.S. Framework: Preemption and Competition

In March 2026, the White House released its National Policy Framework for Artificial Intelligence, recommending federal legislative action across seven areas including child protection, intellectual property, innovation, workforce development, and preemption of state AI laws. The framework represents the U.S. approach to AI sovereignty: rather than building state-owned AI infrastructure, it aims to maintain American dominance through private-sector innovation, uniform federal regulation, and export controls that limit adversaries' access to frontier capabilities. This contrasts sharply with the approaches of the EU (regulatory sovereignty), China (state-directed self-sufficiency), and Gulf states like Saudi Arabia and the UAE (sovereign wealth fund-backed infrastructure investment). Each model reflects different assumptions about the relationship between government, industry, and technology — and each carries distinct risks as AI becomes embedded in every layer of economic and social life.

The Semiconductor Bottleneck

At the hardware layer, AI sovereignty ultimately bottlenecks at semiconductor fabrication. The global AI economy — projected to reach $16.5 trillion and capture 17% of global GDP by 2028 — runs on a chip supply chain controlled by fewer than a dozen critical companies. Governments worldwide plan to invest $1.3 trillion in AI infrastructure by 2030, but infrastructure without chip access is an empty building. This is why the U.S.-China chip war, TSMC's geopolitical significance, and the race to build domestic fabrication capacity (the U.S. CHIPS Act, the EU Chips Act, China's Big Fund) are not peripheral to AI sovereignty — they are its most consequential battleground. Nations that secure their position in the semiconductor supply chain will define the terms of AI sovereignty for the next decade; those that don't will find their AI ambitions constrained by the export policies of others.