This essay is part of a series from Carnegie’s Digital Democracy Network, a diverse group of thinkers and activists engaged in work on technology and politics. The series is produced by Carnegie’s Democracy, Conflict, and Governance Program. The full set of essays is scheduled for publication in summer 2026.

In February 2026, on the sidelines of the AI Impact Summit in New Delhi, the United Arab Emirates (UAE) and India signed a consequential AI infrastructure deal, marking a new phase of sovereign AI development. The Emirates’ state-owned technology group G42, in partnership with the Mohamed Bin Zayed University of Artificial Intelligence and U.S. technology firm Cerebras, agreed to establish a national-scale AI supercomputer in India. The 8-exaflop supercomputer will have the power to accelerate AI research and innovation in India and will rank among India’s most powerful AI compute systems. The deal was subsequently formalized during Prime Minister Narendra Modi’s state visit to Abu Dhabi in May. Two elements of the deal make it noteworthy. First, the supercomputer will be funded by Emirati capital, using American chips and hardware expertise from the UAE and United States. Second, the agreement will be governed by India’s governance framework and data localization requirements, which mandate that data be stored domestically, while contributing to the country’s IndiaAI Mission.

The deal offers a useful perspective on AI sovereignty and the strategies middle powers are adopting in its pursuit. The quest to attain AI sovereignty has become a defining feature of contemporary geopolitics. Currently, a small number of companies, largely based in the United States and China, dominate frontier AI models and the infrastructure that powers them. The rest of the world is struggling to keep up. This raises two critical questions: How are middle powers pursuing AI sovereignty? And what does a realistic vision of sovereignty look like for them?

India and the UAE have emerged at the forefront of AI innovation, offering important insights. As their experiences suggest, attaining complete sovereignty is unrealistic for most nations, but that doesn’t mean they must revert to dependence on the United States or China either. Embracing “strategic autonomy,” or “managed interdependence,” offers a third way.

The Appeal of AI Sovereignty 

Sovereign AI generally refers to a nation’s ability to independently design, develop, and govern AI systems using its domestic infrastructure, nationally held data, and a homegrown talent base. Rather than thinking of AI sovereignty as a binary between independence and dependence, analysts increasingly understand it as a spectrum, with countries achieving different capacities across the AI stack, including semiconductors, cloud platforms, data centers, foundation models, and supporting digital infrastructure.

Countries have differing motivations for pursuing AI sovereignty. For many, national security is a top-level consideration, as they are concerned with safeguarding sensitive national data from foreign control and preserving access to advanced defense and intelligence capacities. Other relevant factors include economic competitiveness—nations look to AI to drive domestic investment, employment, productivity, and long-term growth—and cultural drivers, with governments prioritizing AI systems that reflect their own linguistic and cultural contexts.

These motivations point to a deeper structural concern: dependence. Countries cannot fully safeguard sensitive data, capture the economic value of AI, or ensure culturally and linguistically appropriate systems if the underlying AI stack is owned, operated, or governed by foreign actors. Almost every country exhibits dependencies across the AI stack. Companies based in either the United States or China continue to supply the bulk of the components powering systems worldwide. For instance, a recent report from the Washington, DC–based think tank Center for a New American Security found that NVIDIA alone supplied graphics processing units (GPUs) for 52 percent of all tracked infrastructure projects around the globe, from South Korea to Canada.

Nonetheless, countries are finding ways to navigate their dependencies. Some are prioritizing building out domestic infrastructure, such as data centers, GPU clusters, and access to compute infrastructure. A smaller number are focusing on developing or adapting foundation models. Only a few countries are actively investing in national datasets or data-sharing platforms.

Insights from India and the UAE

India and the UAE are instructive, varied cases—one a lower-middle-income democracy that has a large talent base and over 1.4 billion citizens, the other a small, capital-rich Gulf monarchy, with a centralized government and sovereign wealth to deploy at scale. Both offer lessons about the tricky balance between pursuing technological sovereignty while still drawing from advanced technology from the United States and China.

India’s government is betting heavily on domestic AI innovation, linking sovereign AI to its development goals. Delhi views AI innovation as a key means to address internal inequality gaps and raise its population’s standard of living. In 2024, the government launched its IndiaAI Mission, a flagship project to increase investments in AI. Since then, the country’s sovereign AI sector has raised over $5.5 billion spread across 1,700 firms, seeking to leverage the scale of its population, large talent pool, and state-backed digital public infrastructure.

India’s approach to building out domestic AI capacity plays out in two key areas: developing national foundation models and expanding compute infrastructure. The country has invested heavily in sovereign large language models (LLMs), such as Sarvam AI and BharatGen, in an effort to reach a significant portion of its population that remains underserved by leading frontier models. The country has also devoted sizable funding to domestic semiconductor manufacturing and data centers, although it has experienced limited results so far. As of early 2026, the IndiaAI Mission’s compute facility consisted of 38,000 GPUs, all supplied by American firms. This points to a persistent dependency: India continues to rely on U.S.-supplied infrastructure, including semiconductors from NVIDIA and cloud-computing services from Amazon Web Services, Microsoft Azure, and Google Cloud.

In light of these dependencies, India has prioritized investments in its domestic LLMs and made efforts to set strong data governance standards at home. India’s SarvamAI is particularly noteworthy and has allowed the country to cultivate domestic capacity in the development and use of LLMs. As a government-supported sovereign AI initiative, Sarvam was created to advance India’s multilingual AI ecosystem and support all twenty-two national languages. It incorporates text-, voice-, and vision-based data, and is trained on India-centric datasets. The model is now being deployed across several public sectors, including education and healthcare, where it is enabling rural patients to “access medical advice, schedule appointments and consult doctors through WhatsApp and low-bandwidth interfaces.”  Analysts note that these efforts to make national and sector-specific AI models may also be more financially feasible for other countries, like Saudi Arabia, over the next decade.

In comparison, the UAE has reportedly spent an astounding $148 billion on AI investments domestically and abroad since the start of 2024, backed by its sovereign wealth funds. Across the Gulf states, including Saudi Arabia and Qatar, the UAE has been at the forefront in both building data center capacity and attracting foreign investment. For Abu Dhabi, economic diversification is a key motivator driving its sovereign ambition, as it seeks new sources of long-term growth and influence in the face of declining hydrocarbon revenues. In addition, its leadership is responding to increasingly securitized global AI supply chains, recognizing the importance of preventing dependencies, and limiting exposure to the “compliance demands and alignment pressures” imposed by U.S.-China competition.

The primary vehicle for the Emirates’ sovereign ambitions is G42, a private AI company chaired by the UAE’s national security adviser, Sheikh Tahnoun bin Zayed, that is valued in the billions. Despite G42’s significant investments and global expansion, the UAE’s sovereign ambitions have yet to fully take off. Initially, the country sought to leverage both U.S. and Chinese expertise to build up its AI sector. That strategy unraveled in 2023, when the Joe Biden administration placed the Emirates on an export restrictions list because of its close ties to Chinese tech firms. Subsequently, the UAE reduced its reliance on Chinese technology and deepened cooperation with American companies (though the UAE retained several key partnerships with China). But these efforts have not translated into independence. Rather, they have deepened the UAE’s reliance on Washington in lieu of Beijing and pushed the country to align its national security standards and protocols with America’s.

Since then, the UAE’s sovereignty efforts have hinged on significant state-led investments and its open weight models, such as G42’s Falcon, Jais, and K2 Think, that promote Arabic culture. At the same time, the country mitigates its dependence on foreign chips and frontier models through long-term “negotiated access, co-production agreements, and US-aligned security partnerships.” Although the UAE has downgraded its ties with China after being put on the U.S. export control list, it has managed to increase its leverage with the United States by hosting American supercomputers and investing in American firms, securing reliable access to frontier capacities and using financial commitments as a form of geopolitical insurance. The UAE’s strategy shows how countries can gain advantages in specific sectors of AI development despite larger dependencies, affording them more bargaining power with their partners.

A Model of Managed Interdependence

It is increasingly clear that full AI sovereignty is beyond the reach of even the most capable middle power actors. A more realistic path toward sovereignty involves adopting a hybrid approach that allows countries to exercise a degree of autonomy and incrementally build their sovereign capacity in specific sectors, while forming partnerships with like-minded countries that reduce their exposure to risks, such as vendor lock-in and asymmetric bargaining power, across the AI stack. Analysts describe this strategy as “managed interdependence.”

The partnership between the UAE and India through the G42 deal demonstrates the promise of managed interdependence and offers an avenue through which countries can partner up, while simultaneously exercising greater autonomy over domestic deployment. Through the deal, India receives the compute infrastructure needed to power its domestic AI capacities, hosted within its borders under domestically defined rules. Similarly, through G42, the UAE gains a strategic foothold in one of the world’s largest digital markets and advances its position as an AI infrastructure partner of choice. To that end, the Emirates are already looking to scale up this model; recent reporting indicates that G42 is establishing a global network of AI facilities that it builds, owns, and operates for national governments, with India the first to sign on.

Taken together, the cases of India and the UAE underscore the value of managed interdependence as a guiding framework for AI sovereignty. With the reality that complete AI independence remains out of reach for many countries, a better approach is for governments to reduce their exposure to foreign dependencies, invest strategically in sovereign capacity, and preserve flexibility as the global AI landscape shifts.