“The AI race is not just a technological competition. It is a contest over who gets to write the rules of the future global order.” — Carnegie Endowment for International Peace, 2025

In 1946, Winston Churchill delivered a speech warning the world of an “iron curtain” dividing two political camps in Europe. Eight decades later, the world is witnessing a different kind of divide. One, which is not drawn by physical borders, but it, is drawn by the lines of code, chip architectures, and data-governance regimes. Artificial Intelligence (AI) has become the defining technology of the twenty-first century, and the race to control it, is splitting the global order into two rival tech ecosystems. One, under the leadership of the United States and its allies; while the other under the leadership of China. This division, sometimes referred to as “AI bifurcation,” is far from a foreseeable event.

The Two Blocs Taking Shape:

The clearest sign of AI bifurcation is visible in the area of semiconductor export controls. Since October 2022, the Bureau of Industry and Security (BIS) under the United states Department of Commerce  has issued a series of sweeping regulations to prevent China from accessing advanced AI chips. The January 2025 Framework for Artificial Intelligence Diffusion, published in the Federal Register, went further than any previous measure. It established new worldwide licensing requirements for advanced computing chips and for the first time in history regulated AI model weights, the core parameters that give an AI system its intelligence.

The framework sorted countries into tiers: a privileged group of nineteen close US allies, including Australia, Japan, South Korea, the United Kingdom, and members of the European Union, received near-unrestricted access. Countries outside this group, including India, received capped allocations. Countries classified as “adversaries,” primarily China, Russia, and North Korea, faced a presumption of denial for export licences.

In May 2025, the Trump administration rescinded the Biden-era framework, calling it “divorced from commercial reality,” but simultaneously tightened chip controls on Chinese companies, including issuing specific guidance warning against the use of Huawei’s Ascend AI chips. The core logic remained unchanged with keeping advanced AI infrastructure out of Chinese hands. This reflects a broader trend. According to the Stanford University 2025 AI Index Report, US federal agencies introduced fifty-nine AI-related regulations in 2024 alone which is more than double the number in 2023 and issued by twice as many agencies. The regulatory pace is accelerating.

China, for its part, has responded with equal resolve. By June 2024, China had built 246 EFLOP/s of total computing capacity across public and commercial data centres, aiming to reach 300 EFLOP/s by 2025, according to China’s own 2023 Action Plan for the High-Quality Development of Computing Power Infrastructure. In September 2025, China banned its domestic companies from purchasing American H20 chips and ordered data centres less than thirty percent complete to remove all foreign chips. The message was unambiguous: China was building its own AI stack, from chips to models, independent of the West.

DeepSeek and the Myth of Decoupling:

In January 2025, a Chinese start-up named Deepseek released Deepseek-R1, which challenged the underpinning assumption of the US export control strategy. DeepSeek-R1 was able to perform as well as some of the best western frontier models in the United States, but it was built at a much lower cost and offered as an open source model. The fact that this was a huge advancement because the entire US chip control strategy rested on the idea that cutting off China’s access to advanced hardware would slow its AI progress. Deepseek suggested otherwise.

According to the Observer Research Foundation’s analysis, a study of  Stanford University  in 2025, positioned China as the top nation for AI patent applications and the second-largest contributor to global AI research behind the United States. Meanwhile, the RAND Corporation noted in May 2025 that despite export controls, Chinese firms had exploited loopholes including a major incident in September 2024 in which TSMC, the Taiwanese chipmaker, was found to have inadvertently produced advanced AI chips for Huawei through a Chinese proxy company. The legal and commercial walls between the two ecosystems, it turns out, are easier to declare than to enforce.

The Global South Caught in the Middle:

One of the most significant, and unconsidered, aspects of the AI bifurcation is its effect on the Global South countries. The other countries in Asia, Africa and Latin America are being urged, implicitly or explicitly, to take sides in a technology race in which they were not directly involved. The US export controls established a hierarchy. Advanced chips would be available to the 19 “AI Authorisation Countries” without restrictions. All others, including India, had a quota system. The May 2025 revisions brought a measure of relief for India as it was removed from the restricted list, but the fact that the underlying architecture of technology gatekeeping was intact, was a different story. China has rushed in to fill the void.

China has been promoting the use of Deepseek and other open-source AI capabilities as a means to achieve digital sovereignty for developing countries. If the US does not allow open-source AI models, said the Observer Research Foundation in its brief on the DeepSeek moment 2025, China may be able to dominate the world in this area, as many businesses and governments will be using models that reflect Chinese norms and values rather than Western ones. At the 2025 BRICS summit in Brazil, the issue of “inclusive and responsible AI governance” was put at the heart of the agenda, marking the start of a different governance approach in developing countries.

The European Union has sought to forge a third way. The EU’s groundbreaking AI Act, the world’s first comprehensive AI regulation, has taken a risk-based approach to governance, rather than a geopolitical one. The EU has also been active in backing the UN Secretary-General’s idea for a Global AI Governance Agency and has supported UNESCO’s work on universal AI ethics standards. In September 2025, the China-CEE Institute pointed out that the EU has a deep internal contradiction: it is ahead of the curve in the area of governance norms, while it is far behind in the area of investments, infrastructure and number of frontier companies.

Governance without a Global Framework:

AI ecosystems are becoming more fragmented, and international governance is struggling to catch up. The number of legislations featuring AI increased by 21.3 percent in 75 countries from 2023 to 2024, up from ninefold from 2016 to 2024, according to the Stanford 2025 AI Index Report. However, the regulatory measures taken are largely national or regional. No international framework for the governance of AI like as the Non-Proliferation Treaty or Chemical Weapons Convention.

Washington and Beijing publicly agreed in November 2024 to ban the use of AI in nuclear command and control systems, in a rare moment of agreement, as the Brookings Institution noted. However, there is no implementation or verification. In May 2024, the US-China talks on AI safety in Geneva broke down at an early stage, with the two sides failing to reach an agreement even on the most fundamental questions of what constitutes AI safety. The UN backed Global Dialogue on AI Governance will likely give the first truly multilateral space for these conversations in 2026 and it is far from certain that the big powers will agree to binding commitments.

What This Means for International Relations:

“Technology alliances are replacing traditional security alliances. Export controls are now functioning as instruments of foreign policy more targeted than any sanctions regime in history.”                                                                                     – RAND Corporation May 2025

The global AI order is a dichotomy with far-reaching consequences for international relations in the near future. Technological partnerships are taking the place of or changing conventional security partnerships. The US has entered agreements with Saudi Arabia and the UAE on AI and has cautioned allies to avoid Chinese AI infrastructure. Export controls are being used as foreign policy tools and in a more specific and nuanced way than traditional sanctions.

It is especially problematic for nations such as Russia and India, which are key strategic partners of the US, as well as significant trading partners of China and a vocal supporter of the Global South interests. The third option is to build domestic AI capabilities, and India is trying to do just that by introducing its own sovereign large language model at the AI Impact Summit in February 2026. How viable this route is on a mass scale and whether other middle powers can do the same will be a major issue of the coming decade.

Summary:

The digital iron curtain is not a metaphor borrowed from Cold War history. It is an emerging structural reality in the global order. The world’s most consequential technology is now being developed and deployed along geopolitical fault lines. Different standards, different chip architectures and different governance philosophies define each side of this divide. Unlike the Cold War’s physical wall, this one is more porous and more technically complex. It is also far harder to govern.

The chips on a server, the weights inside an AI model and the export rules that permit or deny their use have all become instruments of statecraft. Nations are no longer competing only over territory or trade. They are competing over the infrastructure that will power the next century of economic and military power. Middle powers like India face a defining choice: build sovereign capacity or remain dependent on ecosystems shaped by others. For scholars and policymakers of international relations, this new terrain demands serious attention. Understanding AI bifurcation is no longer a niche technical concern. It is a core question of global power politics and it will only grow more urgent with time.

References:

Bureau of Industry and Security, US Department of Commerce. (2025, January 15). Framework for artificial intelligence diffusion. Federal Register, 90 Fed. Reg. 4,544. https://www.federalregister.gov/documents/2025/01/15/2025-00636/framework-for-artificial-intelligence-diffusion

Bureau of Industry and Security, US Department of Commerce. (2025, May 13). Rescission of Biden administration AI diffusion rule [Press release]. https://www.bis.gov/press-release/department-commerce-announces-rescission-biden-era-artificial-intelligence-diffusion-rule-strengthens

Stanford University Human-Centered AI. (2025, April). AI index report 2025. https://hai.stanford.edu/ai-index/2025-ai-index-report

RAND Corporation. (2025, May). China’s AI models are closing the gap — but America’s real advantage lies elsewhere. https://www.rand.org/pubs/commentary/2025/05/chinas-ai-models-are-closing-the-gap-but-americas-real.html

RAND Corporation. (2025, July). Full stack: China’s evolving industrial policy for AI. https://www.rand.org/pubs/perspectives/PEA4012-1.html

Observer Research Foundation. (2025, May). Global AI’s ‘DeepSeek moment’: Impact and implications. https://www.orfonline.org/research/global-ai-s-deepseek-moment-impact-and-implications

Nandi, A., & Balasubramani, S. (2025, November). DeepSeek and global AI innovation: Sovereignty, competition, and dependency (ORF Issue Brief No. 846). Observer Research Foundation. https://www.orfonline.org/research/deepseek-and-global-ai-innovation-sovereignty-competition-and-dependency

Observer Research Foundation. (2025, December). Geopolitical fallout from the US-China AI competition. https://www.orfonline.org/expert-speak/geopolitical-fallout-from-the-us-china-ai-competition

Brookings Institution. (2025, October). How will AI influence US-China relations in the next 5 years? https://www.brookings.edu/articles/how-will-ai-influence-us-china-relations-in-the-next-5-years/

Atlantic Council. (2026, January). Eight ways AI will shape geopolitics in 2026. https://www.atlanticcouncil.org/dispatches/eight-ways-ai-will-shape-geopolitics-in-2026/

Carnegie Endowment for International Peace. (2025, July). China’s AI policy at the crossroads: Balancing development and control in the DeepSeek era. https://carnegieendowment.org/research/2025/07/chinas-ai-policy-in-the-deepseek-era