The future of AI depends on a supply chain that is simultaneously advanced and fragile – built on US chip design, Dutch lithography machines, and Taiwanese manufacturing capacity. As export controls tighten and trade rivalries deepen, governments and companies alike must confront difficult choices about their technological (in)dependence, economic security, and international influence. This is a highly concentrated supply chain, and the unfolding developments will co-determine who builds the most capable AI systems, and who sets the terms for digital competition.
AI holds an important place in the geopolitics of digital technologies, with competitors such as the US, China, and the EU engaged in what has been called an AI “global race.” European Commission President Von der Leyen emphasized the importance of the AI race to the EU and said the EU should lead it by “harnessing its power for productivity and people.” In the United States, AI concerns center on national security. For businesses everywhere, this competition creates massive opportunities alongside serious supply chain risks. On all sides, it is clear that immediate, strategic thinking is required.
AI runs on highly specialized chips designed for specific mathematical operations. Therefore, behind the surge in AI investment is a crucial competition over who makes these chips. Developing AI chips involves multiple stages requiring specialized knowledge and expertise. Three of these segments are particularly critical: chip design, manufacturing equipment, and high-precision fabrication. Each is dominated by just a handful of players and this is what creates significant chokepoints, which have strategic consequences. While other components like critical raw materials are also involved, the competitive dynamics involved are related to these three high-value links in the chain.
At the design stage, the United States leads through firms like Nvidia, whose Graphics Process Units (GPUs) were initially developed for gaming but proved ideally suited for AI’s intensive calculations. The company has an estimated global market share as high as 84%, propelled by its introduction of CUDA technology that allowed GPUs to be used for parallel computing in fields including research and innovations such as its Tensor Cores.
Producing AI chips requires highly complex manufacturing equipment. Chief among them are lithography machines, which use light to pattern the ultra-fine circuit features that make such chips so powerful. This segment is dominated by just a few players. The Netherlands-based ASML has a monopoly on so-called Extreme Ultraviolet (EUV) lithography – the technology that produces the smallest nodes and most advanced AI chips. Canon and Nikon, both Japanese companies, compete in the lower-tier Deep Ultraviolet (DUV) technology segment. As a result, the pace of AI hardware innovation effectively depends on the progress and output of a single Dutch company.
Even with the right designs and equipment, few countries are able to produce AI chips at scale. Most rely on highly advanced contract manufacturers, so-called foundries that specialize in the production of advanced chips for other firms. Taiwan-based TSMC dominates this market, with a global share of around 60% of all chips and more than 90% of the most advanced ones. The strategic importance of Taiwan’s position was underscored recently when Nvidia announced plans to build a new headquarters in Taipei and construct an AI supercomputer using 10,000 of its latest chips, despite ongoing US pressure for companies to reduce their dependence on the island. As demonstrated by recent US and EU efforts to onshore TSMC plants, other countries struggle to create similar conditions on their own. There is also a limited number of integrated device manufacturers that both design and produce chips, such as Intel, but their market share lags behind that of fabless design companies, like Nvidia, which design their chips but outsource the manufacturing to foundries.
AI continues to raise complex security risks, notably concerning military applications. In response to concerns about China’s assertive stance in the Indo-Pacific, the United States imposed extensive export controls on the semiconductor industry in October 2022, aiming to prevent Beijing from developing its own AI capabilities. The Netherlands and Japan followed with their own export controls on the sector in 2023. These measures were designed to block companies such as ASML, Canon, and Nikon from exporting their best machines to China.
The impact of these measures, however, remains contested – and is rapidly evolving. Claims that China’s leading semiconductor company, SMIC, has already manufactured 5nm electronic circuits, challenge both Taiwan’s TSMC dominance and the assumption that such production requires the latest Western technology and machines. The policy landscape is evolving under the second Trump administration, as exemplified by the case of Nvidia’s China-specific H20 chip. During a recent Taiwan visit, Nvidia CEO Jensen Huang said the company is evaluating how to address the Chinese market under current controls.
Geopolitically, the various technology holders in the semiconductor supply chain in Europe, Asia, and the US depend on each other for the production of advanced AI chips. Confronted with China’s significant strides in the development of chips and AI – exemplified by the emergence of SiCarrier, SMIC and DeepSeek – and the confrontational US stance under President Trump, the EU needs to develop its own strategy to manage its technology policy.
Some argue that EU member states would do well to ease export restrictions on chips and chip-manufacturing equipment. Doing so could allow ASML to expand market share in China and potentially ease pressure from state-supported Chinese competitors. Such a move, however, requires careful consideration of how much strategic alignment the EU wants to maintain with Washington and Beijing. After all, beyond the stated concerns about military use, export controls are a tool to slow China’s AI progress and preserve a technological edge for the US and its allies.
For European firms, the tradeoff is clear: strategic alignment with US policy could mean restricted access to one of the world’s fastest-growing chip markets, while loosening controls may open revenue streams that also raise political risk. Meanwhile, the EU could further invest in projects through its Global Gateway program to engage countries like India or Indonesia, which are also looking to develop chip supply chains – albeit in less advanced technologies. Governments will need to engage the private sector more strategically and account for geopolitical risks across supply chains. Companies in the chips industry could benefit by supporting infrastructure and development in Global Gateway partner countries, especially when coordinated with partners from countries such as Japan and South Korea, which have their own established semiconductor industries.
The EU’s Chips Act (and forthcoming Chips Act 2.0) has already allowed consortiums of European companies to work on onshoring chip manufacturing. These often include Taiwanese TSMC, whose participation brings critical expertise. A recent example is Infineon’s chip-making plant in Dresden. Additionally, nine EU member states recently took a step towards expanding the local semiconductor industry by forming a Semicon Coalition to explore and coordinate opportunities for Europe in the long term.
Taken together, these efforts demonstrate a growing ambition to secure Europe’s position in the global semiconductor value chain. Investments in infrastructure, local manufacturing, research, and supply chain resilience, in collaboration with the private sector, offer a powerful formula of economic and strategic leverage. A stronger domestic ecosystem will help Europe handle disruption, create jobs, and reduce dependency on global chokepoints. And just as important, the EU should continue to support semiconductor champions like ASML while fostering the next generation of chip innovators – ensuring it will always have something to bring to the geopolitical bargaining table, particularly with a more transactional United States.
© IE Insights.
