How is the resurgence of industrial policy reshaping global cooperation in AI and related technologies? Ayantola Alayande and Aleksei Turobov argue that industrial policy activism in response to advances in AI and semiconductor technologies has led to an increased focus on “digital sovereignty”, encouraging bilateral and trilateral agreements over multilateral cooperation.
The artificial intelligence (AI) value chain is characterised by geopolitical and geo-economic rivalries. History shows that earlier general-purpose technologies like electricity and the steam engine revolutionised the domestic economic structures of the 18th century (the Industrial Revolution) while also reshaping international trade and politics, ultimately laying the foundation for modern globalisation. Similarly, AI and other emerging technologies of the 21st century are disrupting the international balance of power, but driving a new wave of protectionism rather than fostering global cooperation.
This protectionism manifests in two major ways. First, states hesitate to prioritise multilateral AI governance because they fear falling behind in what they see as a zero-sum game of global competition. This reluctance can be seen in the proliferation of national AI policies and a shift from multilateral frameworks like those established by the OECD (Organisation for Economic Co-operation and Development) or the UN (United Nations) to more focused bilateral and trilateral agreements. This potentially undermines the realisation of a unified international governance regime for AI, such as the proposal for a UN AI Agency, similar to the International Atomic Energy Agency (IAEA).
Second, the race to develop advanced AI systems for economic, political, and military superiority, particularly between the US and China, is intensifying. Both nations dominate AI research, talent, and data, viewing success in this ‘great tech rivalry’ as crucial to their future economic and political strength. This has spurred demand for advanced AI chips and led to protectionist policies aimed at safeguarding domestic industries. However, these policies clash with the globalised nature of the semiconductor industry itself, which relies on a complex, international supply chain involving the US, Netherlands, Taiwan, and South Korea. While advantages in the AI value chain are concentrated among a few frontier nations, the predicted run-out in compute and infrastructure in these countries will bring middle powers such as the United Arab Emirates (UAE), Brazil, Singapore, and India who already excel in talent, data, and physical infrastructure into the geopolitical mix, intensifying the race to the bottom.
How will this geostrategic competition impact the international governance systems that are in place to ensure equitable distribution of progress in global public goods such as AI? These developments suggest that the evolution of industrial policy in response to AI and semiconductor advances is leading to increased protectionism, fundamentally altering the global cooperation landscape. In particular, the rise of digital sovereignty will ultimately lead to favouring bilateral and trilateral agreements over multilateral cooperation.
Industrial policies for AI
Policy efforts to stay ahead in the AI race have primarily focused on investments in advanced semiconductors and AI-enabling infrastructure, such as data centres. According to the IMF’s 2024 report on industrial policy, high-tech sectors like semiconductors and the critical minerals powering them were the most active sectors of industrial policy intervention in 2023. The US has taken the lead by implementing export controls on advanced chips and offering billions in subsidies to major semiconductor companies like TSMC and Samsung, which have established production facilities domestically. Similar initiatives are underway in Europe and middle powers such as India.
European countries, such as Ireland, Sweden, Denmark and the UK have instituted a range of tax reliefs and R&D grants for companies working on energy-efficient data centres, advanced chips, and AI enabling physical infrastructure. For example, between 2015 and 2024, the US built seven new data centres and relocated over 200 from abroad back to domestic soil, reflecting a significant investment in computing infrastructure. Some countries, including the UK, have emphasised R&D as a key strategy. Earlier this year, the UK government announced over £100m in funding for nine new AI research hubs, adding to an earlier £54m investment in AI safety research. Similar initiatives have been launched in South Korea and India.
China, constrained by US (United States) export restrictions on advanced chips and stricter data exchange controls, has adopted a different strategy. Unable to match the capabilities of US models like OpenAI’s GPT-4, it has focused on three key areas. First, the government promotes the widespread application of AI models for industrial and commercial use by fine-tuning open-source LLMs (large language models) developed in the US, mirroring the notion that China’s technological advancements of the 2010s were built on the back of US innovations. Second, there is increased financial support for frontier AI firms, particularly in less developed regions of China. Third, leading universities such as Tsinghua receive strong R&D backing, which has helped spin out the most prominent AI start-ups in the country.
AI governance is also emerging as a distinct area of intervention. The rise in government-designated AI safety institutes worldwide reflects a growing trend toward nationally tailored AI governance approaches. While governments continue to publish national AI strategies and facilitate intergovernmental frameworks like the OECD AI Principles, the European Union (EU) remains committed to setting the global standard in AI regulation. Scholars suggest that the ‘Brussels effect,’ in which EU regulations influence other countries, could significantly impact the global AI market, highlighting the geopolitical dimension of AI rule-making.
The erosion of multilateral norms
Industrial policy can be a powerful driver of technological innovation, boosting domestic industries and fostering global competition, but it often veers into protectionism. This leads countries to restrict trade and investment and results in the increased localisation of global processes, which can stifle subsequent innovation. Over the past few years, industrial policy, particularly in emerging technologies, has increasingly been shaped by national security concerns and the perceived threat from ‘foreign entities of concern‘. Major global players like the US, China, and the EU are focused on building national champions.
Given the central role of AI and semiconductors in technological progress, several nationalist policies are emerging in that industry. Legislative measures like the US CHIPS and Science Act and the Infrastructure Investment and Jobs Act signal a shift towards security-oriented new forms of protectionism. Similarly, Europe’s initiatives, including the European Chips Act and the EU AI Act, which aim to solidify the EU’s leading position in regulation and strengthen its digital sovereignty are also characterised by protectionist logic.
Enhancing these challenges is the erosion of trust in leading international institutions. The World Trade Organization (WTO)’s dispute settlement mechanism has faced significant setbacks, undermining its ability to effectively manage trade conflicts in an increasingly digital economy. Similarly, the UN is facing a cascade of challenges, prompting growing worries about a systemic crisis within the organisation, raising concerns about its capability to manage the impact of new technologies on global governance. Meanwhile, Russia’s invasion of Ukraine and escalating global instability have exposed NATO’s limitations in crisis prevention and deepened political tensions within the alliance.
In this context of global disorder, marked by security threats, climate change, and the COVID-19 pandemic, there is a perception that multilateralism is in crisis. Key global actors are increasingly turning towards bilateral and trilateral agreements, which bypass the need for broad international consensus. While multilateralism traditionally promotes broad cooperation, bilateralism offers more targeted and adaptable solutions, enabling quicker responses to specific issues and bridge-building in challenging times.
Recent US strategic initiatives, such as the I2U2 Group and strengthened security ties between the Republic of Korea and Japan, reflect a shift in US foreign policy from multilateralism to localised agreements. However, the perceived ineffectiveness of these institutions in responding to global challenges has prompted the US to reorient its strategy towards more localised agreements in AI, where it is deepening ties with countries such as the UK, India, and the UAE.
This approach mirrors the current landscape of industrial policy. Faced with deadlock in international institutions, bilateral, mini-lateral, and regional agreements are becoming more prominent. China, too, is increasingly focusing on localised agreements, forging AI partnerships with countries like Saudi Arabia, France, and Singapore. The evolving nature of these agreements, such as Singapore’s dual engagements with the US and China, underscores the complexity of these dynamics. Despite the challenges, the US and China continue to engage on AI issues through bilateral channels, while the US deepens its semiconductor partnerships with key allies like the Netherlands and Japan. These targeted collaborations reflect a broader strategic shift towards pragmatic, focused agreements.
Given the rise of bilateralism and the strategic importance of emerging technologies, a balanced approach to industrial policy is essential — one that encourages collaboration while guarding national interests. This approach would foster international cooperation on shared challenges, such as cybersecurity, climate change, and ethical AI development while protecting critical industries through carefully calibrated measures that avoid excessive protectionism. Instead of broad trade restrictions, nations could adopt targeted export controls and investment screening mechanisms that address specific security concerns without stifling innovation or disrupting global supply chains.
Countries should engage in dialogue about their competitive approaches, ensuring a clear understanding of the benefits of economic and technological competition while preventing new ‘arms races’. A strategic balance between competition and cooperation is vital for sustainable AI governance and technology development. Moreover, restoring multilateral frameworks with reforms that enhance their agility and relevance in the face of rapid technological change is necessary. This dual strategy — combining targeted international cooperation with improved global collaboration — would enable nations to navigate the complex landscape of industrial policy, fostering global stability while protecting vital national interests.
The views and opinions expressed in this post are those of the author(s) and not necessarily those of the Bennett Institute for Public Policy.
