The global AI race has entered a phase that no longer belongs exclusively to its two dominant players. For the past three years, the dominant narrative positioned artificial intelligence as a bilateral contest — U.S. model capabilities against Chinese cost leadership, American chips against Chinese open-source strategy. That framing is now giving way to something more complex. A third force is emerging, composed of nations that lack the scale to build comprehensive AI stacks from scratch but retain enough strategic leverage to shape how AI reaches the two-thirds of humanity that neither Washington nor Beijing has yet fully captured.
This is the age of AI sovereignty — and the middle powers are no longer merely spectators.
The Bilateral Competition That Created the Opening
U.S.-based large language models continue to dominate in terms of global use, likely because of a first-mover advantage and superior model capabilities. But that dominance should not be taken for granted. As China’s DeepSeek R1 showed, competitive alternatives can rapidly erode U.S. market share.
A January 2026 RAND Corporation report confirms Chinese models operate at roughly one-sixth to one-fourth the cost of comparable U.S. systems, driven by open-source strategies, efficient architectures such as DeepSeek’s Mixture-of-Experts designs, and domestic infrastructure. Following DeepSeek R1 in January 2025, Chinese LLM global usage share surged from approximately 3% to 13% within months.
Chinese models captured more than 10% penetration in 30 countries and 20% market share in 11 countries. Gains were most pronounced in developing countries and countries with close political and economic ties to China.
The price differential is not a minor tactical advantage. At one-sixth the cost, Chinese models restructure the economics of AI adoption for governments and enterprises in the Global South in ways that U.S. alternatives cannot match at their current price points. The geopolitical implication is direct: countries that adopt Chinese AI infrastructure may embed structural dependencies that outlast any individual product cycle.
What Middle Powers Are Actually Doing
Middle powers that fail to secure influence over the development, deployment, and governance of artificial intelligence will likely forfeit control over their economies, societies, political systems, and positions in the global economy. Global dependencies on U.S. and Chinese technology are unavoidable, but increased sovereignty over the deployment of AI will allow smaller countries to develop their own technological paths that can prioritize the needs of their populations.
Chatham House’s February 2026 analysis identifies four pragmatic pathways: specializing in a particular part of the global AI supply chain; aligning with one of the AI superpowers; sharing sovereignty with other countries to amplify influence; or hedging against instability by using a range of AI capabilities from different countries. Most middle powers are doing some combination of the last two — building targeted domestic capabilities while hedging exposure across multiple foreign providers.
Middle powers are moving from policy to deployment, leveraging distinct advantages rather than attempting to build comprehensive AI stacks. Gulf states, backed by abundant low-cost energy and sovereign capital, are seeking to operationalize computing infrastructure. Japan and South Korea have leveraged their established semiconductor and advanced manufacturing capabilities to deepen domestic AI investment.
India: Rewriting the Terms of the Global South’s AI Access
The India AI Impact Summit 2026, held in New Delhi in February, was the first summit in the global AI series to be hosted by a Global South nation. Union Minister Ashwini Vaishnaw outlined India’s “whole-of-nation” AI strategy, describing plans to build a “frugal, sovereign and scalable” AI ecosystem. The government announced plans to add more than 20,000 GPUs to India’s existing base of 38,000 under the IndiaAI Compute Portal.
Reliance Industries announced plans to invest $110 billion over seven years to build India’s sovereign AI infrastructure, including data centres and a nationwide edge compute network. Government estimates suggested that AI-linked investment commitments associated with the summit could exceed $200 billion in the coming years.
Rather than dwelling on frontier model development or the existential risks that have dominated Western AI discourse, Prime Minister Modi anchored the summit around “impact”: equitable access, climate resilience, and inclusive growth. For countries like India, the question is not whether AI will become too powerful, but whether its near-term benefits will be captured by a narrow band of wealthy nations.
India’s domestic model development reflects this orientation. Sarvam AI launched a 105-billion-parameter foundational LLM alongside speech, multimodal, and text-to-speech systems designed for Indian languages. The government-backed BharatGen Param2 model supports 22 Indian languages with multimodal capabilities. The strategic logic is clear: AI that does not operate in the language of its users is AI that serves someone else’s interests.
The Gulf States: Energy as Compute Leverage
Gulf sovereign wealth funds are translating energy abundance into AI infrastructure in ways that no other regional bloc can replicate. Smaller regional groupings such as Gulf states might leverage energy abundance to build compute infrastructure.
Abu Dhabi-based AI firm G42 revealed plans to build an 8-exaflops supercomputer in India in partnership with U.S. chipmaker Cerebras Systems and local institutions including Mohamed bin Zayed University of Artificial Intelligence. The sovereign AI supercluster will allow India to train advanced AI models domestically under Indian governance.
The Gulf model is distinctive because it does not require the same population-scale data advantage that China holds, or the engineering talent base that India is cultivating. Low-cost energy, sovereign capital, and strategic positioning between East and West create a different form of leverage — one focused on becoming indispensable infrastructure rather than building frontier models.
The Fragility of Full-Stack Sovereignty
The aspirations of middle powers are real, but so are the constraints. The resource intensity and rapid pace of AI development means that only a select few superpowers and middle powers have the scale and breadth of capabilities to sustain a sovereignty strategy over time. Even for most of these countries, AI sovereignty conceived as full-stack autarky remains an illusion.
Europe made a significant deal about investing some $47 billion in AI infrastructure, but in 2026 alone, U.S. firms plan to make at least $650 billion in capital expenditures related to AI. The Gulf powers, India, Japan, and South Korea also lag well behind — despite substantial investments relative to historical levels.
The gap in absolute scale between what individual middle powers can commit and what U.S. hyperscalers spend in a single year means that full independence from either superpower’s AI infrastructure is not a realistic near-term outcome for most nations. The Brookings Institution argues that full-stack AI sovereignty is structurally infeasible for most countries and proposes “managed interdependence”: mapping AI stack dependencies, prioritizing feasible interventions, diversifying suppliers and partners, and embedding interoperability through standards, procurement, and governance.
Why the Race’s Outcome Still Matters to Everyone
The AI race in 2026 is still defined by a multipolar order. The United States and China will continue to yield the greatest influence. In 2026, expect China to double down on its open-source AI strategy to influence the world’s AI infrastructure.
The practical implication for the two-thirds of the world that is neither U.S.- nor China-aligned is that the choice of AI provider is increasingly a geopolitical decision with decade-long consequences. Infrastructure decisions made now — which cloud provider hosts a nation’s health records, which LLM powers its government services, which chip architecture runs its universities — will shape both economic dependency and information sovereignty for years to come.
The goal is not independence — full autonomy remains unrealistic — but strategic flexibility: the ability to switch providers, adapt to disruptions, and avoid coercion. Middle powers that act strategically can secure durable influence over the technologies that underpin their political systems, economies, and societies.
The competition for global AI market share is, at its core, a competition for that strategic flexibility. And the middle powers are no longer content to let it be decided for them.
