Hsinchu: The science park north of this Taiwanese city produces a share of the world’s advanced logic chips that no other single location comes close to matching, which explains why European officials keep finding reasons to visit an island the Union does not formally recognise.
The European Commission and the European Economic and Trade Office in Taipei convened the second Europe-Taiwan Semiconductor Industry Dialogue on 31 August, timed to the margins of SEMICON Taiwan. Companies from across both value chains attended alongside research and technology institutes. The agenda centred on the Chips Act 2.0 proposal and on the demand that artificial intelligence and data centre construction now place on European chip supply.
The venue itself carries meaning. Brussels maintains no diplomatic relations with Taipei and channels everything through a trade office, so a Commission-branded industry dialogue held on the island represents about as far as the Union’s One China policy stretches. Beijing notices each increment. European officials keep the framing strictly commercial for that reason, and the strategy has held so far.
Europe’s first Chips Act set a target of twenty percent of global semiconductor production capacity by 2030, a figure the European Court of Auditors and most independent analysts have long treated as unreachable. Global capacity grew faster than European capacity, so the denominator defeated the policy before the projects finished. The successor proposal abandons the pure volume ambition and shifts emphasis toward design, advanced packaging, equipment and materials.
That shift plays to genuine European strengths. No advanced fab anywhere operates without lithography from the Netherlands, optics from Germany and deposition tools from a handful of European and Japanese suppliers. Brussels sells access to that leverage, and it buys manufacturing presence with it.
TSMC’s Dresden joint venture with Bosch, Infineon and NXP demonstrates the trade in practice. The plant targets mature automotive and industrial nodes rather than the leading edge, which reflects what European customers actually consume and what Taipei will actually export. GlobalWafers and Foxconn have expanded on the continent along similar lines.
Taiwan’s own calculation runs in the opposite direction. Taipei has pushed for years for a bilateral investment agreement with the Union and has not got one, because member states with heavy exposure to the Chinese market decline to absorb the political cost. Taiwanese officials therefore trade industrial cooperation for incremental political normalisation, one dialogue at a time, while Brussels trades market access for fabs and keeps the political ledger closed.
Artificial intelligence demand changes the bargaining position more than either side admits publicly. Europe is building data centres faster than it is building the capacity to supply them, and the advanced packaging that high-bandwidth memory and accelerator modules require sits overwhelmingly in Taiwan. A packaging line in Europe would matter more to supply security over the next five years than another mature-node fab.
The dialogue produced no binding commitments, and industry dialogues rarely do. Its value lies in the standing channel, which lets European regulators understand Taiwanese capacity plans before they legislate rather than after. Whether Chips Act 2.0 turns that understanding into packaging capacity on European soil will define the relationship far more than the next communique does.
