Southeast Asia has been assembling chips for thirty years, and the AI wave amplifies the significance of the region’s front-end manufacturing deficit, without alone proving a complete shift in hierarchy. Of 64 new fabs planned in Asia by 2029, six are planned for Southeast Asia; the other 58 are planned outside Southeast Asia. Investment follows technology, and technology remains heavily concentrated in East Asia, but this concentration has shifted historically. Hanoi, Kuala Lumpur and Bangkok are seeking to move beyond assembly even as barriers to entry have risen considerably.
The essentials
- Asia accounts for more than half of global semiconductor manufacturing capacity, driven by massive investments, but production is heavily concentrated in Asia, with major hubs in Taiwan, South Korea, Japan and China (SEMI, June 2026).
- Among several dozen new fabs announced for Asia by 2029, only a minority will be built in Southeast Asia.
- Vietnam launched construction of its first chip factory in January 2026, with trial production planned for 2028: a signal of genuine ambition, but the starting point of a trajectory that took South Korea thirty years.
- Upgrading the value chain is possible—the South Korean and Taiwanese examples prove it—but it requires precise policy decisions and a twenty- to thirty-year horizon that few governments are willing to accept.
- Export controls limit access to certain advanced equipment depending on the technology and destinations, and influence the room for manoeuvre of emerging economies.
Assembly generates little value, but many countries depend on it
The semiconductor chain has a clear geography. Upstream: design, dominated by the United States and South Korea. In the middle: wafer fabrication, which Taiwan and South Korea control at more than 60% of global capacity for advanced nodes. Downstream: assembly, testing and packaging—what’s called ATMP—which Southeast Asia performs for a substantial share of global production.
This last link is what Vietnam, Malaysia, the Philippines and Thailand have built since the 1990s. Intel has been assembling in Penang since 1972; Texas Instruments has operated in the Philippines since 1979. Malaysia processes a significant share of global ATMP. These factories employ hundreds of thousands of people and generate real exports.
But the value added in assembly represents a fraction of that in fabrication. A wafer from TSMC at the 3-nanometre node is worth several thousand dollars. The cost of assembling the final chip in a Malaysian or Philippine factory represents a minor fraction of the component’s final price. AI demand has supported TSMC, whose net profit grew 36% in 2024 according to its annual results, but the margin increase results from several factors: capacity utilization, volumes, prices and exchange rates, with no automatic mechanism trickling down to downstream assemblers.
This is what SEMI figures from June 2026 confirm: significant investments in Asia and several dozen new fabs announced by 2029. Six fabs are planned for Southeast Asia. The other 58 are going where the technology is already mastered.
Thirty years of assembly have not produced a South Korean trajectory
Economist Noah Smith proposes judging industrial policies by their actual results. Evaluated on this basis, the strategy of attracting foreign direct investment in semiconductors in Southeast Asia delivers a mixed record.
Malaysia has attempted to move up the value chain. In 2023, it announced a new semiconductor policy with the ambition of developing local design capacity. Intel has invested several billion there in advanced packaging factories. But chip design requires engineers trained over ten to fifteen years, ecosystems of specialized subcontractors and productive academic partners. These assets cannot be imported: they must be built.
South Korea offers the counterexample. Samsung began manufacturing semiconductors in the 1970s with technology substantially behind global standards. The South Korean state supported the effort through public aid and continuous investment in engineer training. The trajectory required several decades before Samsung became a leading player. Economist Daron Acemoglu, in his work on institutions and technology, emphasizes that this type of value chain upgrading requires consistent policy choices over the long term, not merely tax incentives for FDI.
This is precisely what is lacking in most Southeast Asian economies. Tax incentives are generous, but industrial policy remains fragmented. Thailand offered massive tax exemptions for twenty years to Honda and Toyota in automobiles; it is now reproducing the same logic in semiconductors, without either sector ever having generated comparable endogenous technological capacity to South Korea’s. French industry faces an analogous tension: attracting foreign investment and developing sovereign industry are two objectives that do not overlap.
Six fabs, and Vietnam’s first fab: real signals, long timeline
The six facilities counted by SEMI for Southeast Asia by 2029 deserve to be read with precision. They concern front-end fabs, not advanced ATMP capacity. Advanced packaging, particularly chiplet-level packaging, represents genuine growth in complexity compared to traditional assembly. Intel and AMD are pushing in this direction, because advanced assembly partly compensates for the limits of node miniaturization.
Vietnam appears as the most ambitious economy. The announcement of a first fab in 2026, launched by Viettel, marks a symbolic break. Vietnam is mainly present in design and assembly, packaging and test activities, without established domestic chip manufacturing at that date. Moving to fabrication requires training process engineers, attracting equipment makers, building a supply chain in ultra-pure chemicals. None of that happens in three years.
Singapore occupies a different position. The city-state already houses foundries: GlobalFoundries produces chips at mature nodes there, and R&D centers for ASML, Applied Materials and Lam Research are located there. Singapore is a technological hub whose immigration and training policy has been deliberately oriented toward this sector since the 1990s. Its model is difficult to transpose to economies ten times larger.
Malaysia and the Philippines are betting on advanced ATMP. This is the most accessible niche in the short term and the most consistent with existing capabilities. Economist Isabelle Méjean, in her work on global value chains, recalls that dependence on a specific segment creates structural vulnerability: if chip makers vertically integrate assembly—a trend TSMC is accelerating—downstream subcontractors lose part of their customer base without having developed alternatives.
Export controls change the game for all latecomers
The question of EUV equipment transforms the context in which these ambitions operate. Extreme ultraviolet lithography, a central technology for producing advanced chips, particularly at the finest nodes, is produced exclusively by ASML in the Netherlands. Dutch controls on certain advanced equipment took effect in 2023; other restrictions on EUV exports to China had accumulated in preceding years. These controls apply according to national rules and case-by-case authorization, not automatically to any economy deemed risky by the United States.
For Southeast Asia, the implication is ambiguous. Southeast Asian countries are not the primary geographic targets of restrictions aimed at China, but some export controls apply to them as well depending on the equipment and destinations. Their dependence on American investors and Western technology places them in a delicate position: any rapprochement with Beijing in terms of investment or supply chains could create tensions with Western partners. The Lowy Institute, in its analysis of strategic supply chains, describes this tension as an effect of forced bipolarization: mid-sized economies lack the means to remain neutral in a technological war between great powers.
This is a reversal of the logic of the past thirty years. Integration into the global semiconductor chain had hitherto provided an argument for non-alignment: a country that assembles chips for Apple and for Huawei has no interest in choosing sides. This productive neutrality becomes harder to maintain when American CHIPS subsidies come with safeguards that restrict certain capacity expansions and technological collaborations in foreign countries of concern. Japan and the Netherlands have strengthened their national controls on certain advanced equipment in coordination with partners, rather than simply accepting American rules. American Asian partners must now integrate this evolution into their calculations.
The question of technological sovereignty arises in similar terms in other sectors: access to cutting-edge technology depends on geopolitical choices as much as investments.
Lessons from the South Korean trajectory
South Korea and Taiwan have developed advanced capabilities that Southeast Asia does not yet possess. Their experience offers lessons, though not a guaranteed recipe.
First lesson: value chain upgrading in semiconductors required a state capable of supporting an industrial effort over several decades. Samsung benefited from structured public support during its catch-up phase. It was a market-oriented bet, but protected until competitiveness became real. Economist Philippe Aghion, in his work on Schumpeterian industrial policy, distinguishes subsidies that sustain rents from those that finance skills upgrading: the former impoverish, the latter can create value if conditioned on measurable objectives.
Second lesson: Taiwan built TSMC on an explicit decision to specialize in foundry services for third parties, a model no one thought viable in 1987. Morris Chang bet that American designers would pay for a quality foundry with no competitor. The bet was not obvious. It succeeded because TSMC systematically invested in process R&D to maintain its technological edge.
These examples raise a serious objection: the technological and political context has changed since the 1970s-1990s. Technological barriers have risen. EUV equipment costs hundreds of millions of dollars per unit and does not export freely. The patents of incumbent players cover thousands of processes. China, despite massive public investments, has not yet produced a foundry capable of rivaling TSMC on advanced nodes.
This does not mean value chain upgrading is impossible. It means it requires more targeted policy decisions, technological partnerships with existing players, and specialization in segments where barriers remain surmountable. Advanced ATMP, mature chip design for specific markets, test equipment: these niches are less prestigious than 3-nanometre node fabrication, but they are accessible to an economy at an early stage.
Choices available to Southeast Asian economies by 2040
The longer question deserves to be posed without pretense. By 2040, three trajectories are conceivable for the region’s main countries, and they need not be mutually exclusive.
The first is deepening advanced ATMP. This is the least risky trajectory and the most immediately accessible. It consists of increasing complexity in assembly, chiplets, 3D packaging, heterogeneous integration, rather than entering wafer fabrication. The value added exceeds traditional assembly, technological barriers remain surmountable with partnerships, and AI-driven demand is real. Malaysia and Vietnam are the most credible candidates for this trajectory.
The signal to watch: announcements of investment by Intel, AMD and Infineon in their packaging units, and the rise of local process engineers.
The second is specialization in mature chip design. Semiconductors at 28 nanometres and above, used in automobiles, industry and consumer appliances, constitute an important share of the market by volume. These nodes do not require EUV. They can be manufactured with equipment less restricted in export. An economy that develops design capacity on these nodes and partners with an existing foundry for fabrication can build a viable industry without claiming to rival TSMC on advanced nodes.
Singapore has begun this path. The signal to follow: creation of design centers funded by local players, not merely subsidiaries of multinationals.
The third is integration into a denser regional chain, with China as pivot or counterweight depending on geopolitical choices. Beijing is investing massively in mature nodes and seeking regional partnerships in assembly. Some Southeast Asian countries could be tempted by this option if Western investment concentrates on a few hubs—Singapore, Vietnam—neglecting others. This scenario is hardest to calibrate: it offers short-term outlets but creates a geopolitical dependence that recent events have made costly for other economies. The warning signal: a growing share of semiconductor FDI from China without technological reciprocity.
The tension among these three trajectories is not resolved by a general principle. It is resolved by precise decisions: what engineer training, what technological partnerships, what access to equipment, what positioning vis-à-vis American export controls. Governments that clarify these choices before 2030 will have an advantage over those that leave decisions to foreign investors. Technological dependence builds slowly and corrects even more slowly: economies that have no semiconductor engineering training policy today will not catch up by 2035.
Vietnam’s first fab in 2026 is a starting point, not an endpoint. What matters now: the technological level of that fab, the conditions for knowledge transfer and Vietnam’s capacity to train the engineers who will put it to use in ten years.
Sources
- SEMI, Semiconductor Manufacturing in Southeast Asia, June 2026: https://www.semi.org/sea/blogs/June-2026
- Noah Smith, Democratic economic policy in the age of AI, Noahpinion: https://www.noahpinion.blog/p/democratic-economic-policy-in-the
- Lowy Institute, analyses of strategic supply chains in semiconductors (Lowy Institute for International Policy, Sydney)
- TSMC, 2024 annual results (Taiwan Semiconductor Manufacturing Company, shareholder report)
- Daron Acemoglu and Simon Johnson, Power and Progress, 2023 (PublicAffairs)
- ASML, activity reports 2023-2024 (ASML Holding N.V.)
- CSIS, Semiconductors and the US-China Tech War (Center for Strategic and International Studies, Washington)
