- AI and trade are no longer separate policy domains. AI growth depends on globalized supply chains for semiconductors, computing infrastructure and digital services while trade is increasingly shaped by who controls AI infrastructure, data flows and cloud capacity.
- Trade openness is a structural precondition for AI-driven productivity gains. Open economies benefit disproportionately from cheaper inputs, faster innovation diffusion and AI adoption spillovers. Trade openness accounts for 23% of the variation in AI adoption across countries with highly open economies, such as Singapore, UAE and Ireland, leading in diffusion. However, while AI can significantly boost growth, its benefits are unlikely to be evenly distributed.
- Export volumes of AI-enabling goods have surged from USD1trn in 2014 to USD3.8trn in 2025 (+280%), accounting for 15% of global trade and far outpacing the 40% growth in goods trade overall. Asia dominates the supply side, accounting for 65% of global AI-related exports and seven of the top ten exporters, led by China (18% of AI-related exports), Taiwan (12%) and Hong Kong (11%). The composition remains concentrated in intermediate inputs (76%) and equipment (23%), reflecting deep dependence on semiconductors and data-center infrastructure. On the demand side, the US has tripled its AI-related imports since 2023, underpinned by 5,427 operational data centers, 45% of the global total. Europe's import growth of just +40% underscores a widening infrastructure gap.
- Services imbalances could scale with AI. ICT services trade reached USD900bn in 2024 (11% of global services trade), with Ireland alone exporting USD173bn, exceeding both the rest of the EU (USD142bn) and the US (USD108bn), and reflecting its role as a billing hub for US multinationals. However, it masks structural dependence, with the EU excluding Ireland running a USD45bn ICT services deficit (mainly with Ireland), highlighting its reliance on US digital ecosystems. And things could get worse. Hypothetically, if the AI services subscription penetration rate of US providers such as ChatGTP Plus and Claude Pro increases from 3% to 50% in a high adoption scenario in the Eurozone, annual payments to US providers could reach EUR34bn (currently EUR2.7bn), equivalent to 20% of the current EU–US goods services deficit. AI thus acts as a scalable, recurring channel that exacerbates structural EU–US digital imbalances and has the potential to lower the US overall trade deficit.
- Beyond the headline data, three structural dynamics are reshaping the global AI trade order:
- 1. Supply chain concentration as single points of failure
- The AI supply chain is concentrating rapidly at the technological frontier. Taiwan, South Korea, the US and the Netherlands dominate the production of advanced chips, high-bandwidth memory and semiconductor equipment, creating critical single points of failure with no near-term redundancy. Unlike broader goods trade, which has partially reoriented along geopolitical lines, AI-related goods remain largely unfragmented across geopolitical blocs, with semiconductors and key components still flowing globally except at the technological frontier (notably advanced US export controls) due to deep interdependencies, challenging the narrative of technological decoupling.
- 2. Infrastructure as geopolitical power
- Control over data centers, cloud infrastructure and computing capacity is emerging as a primary source of geopolitical leverage, distinct from traditional goods trade dynamics. Europe's strategic exposure is acute. With less than 10GW of operational data-center capacity, four times below the US (60GW), and a development pipeline six times smaller, the gap is not expected to narrow. US hyperscalers already control 35% of European computing capacity and account for nearly half of the upcoming pipeline, consolidating a 70% cloud market share. Structural constraints, fragmented regulation, complex permitting processes, grid connection delays, no domestic hyperscaler and limited VC or state-backed funding reinforce this dependency. Finally, reliance on Asian hardware inputs complements the US dependence that will increase without sufficient EU investment in the area. Under that backdrop, Europe is permanently under the threat of a US “kill switch” on cloud data. In the meantime, this effective expansion of this key infrastructure is dependent on inputs sourced in Asia, notably in China (14% import ratio for AI datacenter components) and Taiwan for semiconductor, and as a result exposed to current supply-chain disruption risks along new geopolitical tensions in Middle East. Regaining control and strengthening sovereignty on AI good production but also service delivery is critical for the region to avoid suffering a twin external dependence.
- 3. Industrial policy: from subsidies to protectionism
- Global tariffs on AI-enabling goods have fallen from 5.6% in 2015 to 2.8% in 2025, well below the 7.8% average for manufacturing overall, but non-tariff measures have surged, driven primarily by the US and China, reflecting a strategic shift from financial support toward technology protectionism. Over 3,600 AI subsidy measures targeting critical materials, semiconductors, GPUs, computing equipment and optical fibers are now in force globally, with China having nearly doubled its trade coverage over five years, followed by South Korea, Malaysia, the US and Japan. The overlap of national and multilateral governance regimes is concentrating production, infrastructure and modelling capabilities in the hands of a small number of economies. Regulatory fragmentation is becoming a binding constraint on AI services trade, with ICT trade restrictiveness already explaining 5% of the cross-country variation in AI services flows. The EU and US are advancing incompatible regulatory frameworks, creating compliance friction that reinforces existing structural imbalances.