VIS-NXP Opens $6.7 Bn Singapore Fab as Physical AI Broadens Chip Demand

Singapore’s newest semiconductor fab is not competing to make the smallest, most advanced AI chips. It bets that the next phase of artificial intelligence will require a much broader set of semiconductors as computing moves out of data centres and into cars, factories and machines.

VisionPower Semiconductor Manufacturing Company (VSMC), a 60:40 joint venture between Taiwan’s Vanguard International Semiconductor (VIS) and Dutch chipmaker NXP Semiconductors, opened its first 300mm wafer fab in Tampines on September 28. The Singapore-incorporated company is owned by the two foreign chipmakers, not by the Singapore government. But the city-state serves as the manufacturing base and semiconductor ecosystem where the facility will operate. NXP is entitled to 40% of the fab’s capacity, while VIS operates the plant.

Singapore’s part in the venture goes beyond hosting it. VIS chairman Leuh Fang thanked the governments of Taiwan, Singapore and other countries in 2024 for the support that secured the venture’s approvals on schedule. Tan See Leng, then Singapore’s second minister for trade and industry, described the December groundbreaking as the start of a partnership between VSMC and the Singapore government.

That distinction matters to the Singapore chip story. The city-state’s economic return comes not from an equity stake in VSMC but from anchoring manufacturing capacity, jobs, suppliers and technology activity locally.

The fab is expected to create about 1,600 jobs when fully ramped up, adding another large-scale foreign semiconductor investment to a sector that already contributes close to 7% of Singapore’s GDP. Beyond the partners’ equity, $2.4 Bn from VIS and $1.6 Bn from NXP, the project is funded by customer capacity deposits, loans and subsidies, VIS has said, without naming who provides the subsidies.

Construction is complete and the plant has now entered risk production. The build took about 22 months from its December 2024 groundbreaking, a record for fab construction in Singapore, said Tan. Volume manufacturing is scheduled to begin in the first quarter of 2027 after its first sample lot achieved yields above 99%. Monthly capacity is expected to reach about 44,000 12-inch wafers by 2029.

The investment now stands at $6.7 Bn, according to VIS, down from the $7.8 Bn announced when the joint venture was set up in 2024. VIS revised the plan in May after adding silicon interposers to the product mix and reducing planned monthly capacity from 55,000 wafers to 44,000. Customer-supplied manufacturing equipment for the interposer business also changed the project’s economics.

The lower capacity does not signal weaker demand. VIS chairman Leuh Fang said in May that the revised first-phase capacity was already fully allocated and supported by a long-term customer agreement. VIS was also in preliminary discussions with customers over further capacity.

Demand has since moved faster. At the opening, VIS said it was fast-tracking the timeline for a second Singapore fab, with Fang saying the company should have built both fabs at the same time and that the first could reach full utilisation ahead of the 2029 schedule. However, a decision would not be rushed. Tan, now minister for trade and industry (energy and industry), said he looked forward to deepening the government’s partnership with VSMC as it enters its next phase of growth in Singapore.

Why Mature Nodes Still Matter to AI

VSMC will manufacture using process technologies ranging from 130nm to 40nm for mixed-signal, power-management, analogue and interposer applications. Taiwan Semiconductor Manufacturing Company (TSMC) has licensed the underlying manufacturing technologies under an agreement signed when the joint venture was established. The output will serve high-performance computing, mobile, automotive, industrial and consumer markets.

These are far from the leading-edge nodes used for the processors dominating the generative AI (GenAI) boom. But as AI moves into the physical world, it creates a different hardware problem.

Robots, vehicles and industrial systems need to sense their surroundings, move data between components, manage power and operate safely in real time. NXP has made that shift towards physical AI an increasingly important part of its strategy, developing processors, networking and control technologies for robotics and other intelligent edge systems.

NXP president and CEO Rafael Sotomayor linked VSMC directly to that opportunity at the fab’s opening, saying the additional manufacturing capacity would support technologies enabling the next generation of intelligent-edge and physical AI systems.

That does not mean every wafer coming out of Tampines will be an AI chip. VSMC will serve a much wider group of automotive, industrial, consumer and computing customers. The strategic point is that AI demand is spreading through more of the semiconductor stack, from advanced processors and memory to the power, analogue, connectivity and packaging technologies that allow those processors to function inside real-world systems.

Silicon interposers strengthen that link. Used to connect multiple semiconductor components inside advanced packages, they have become more important as chipmakers combine processors, memory and other components to increase computing performance. Adding interposer production extends that TSMC-licensed technology base into advanced packaging, linking part of the Singapore fab more directly to the architecture used in increasingly complex AI systems.

Singapore’s Return Comes from the Ecosystem

For Singapore, VSMC fits an industrial strategy that has never depended on owning the semiconductor companies it attracts. Its partners have established a strong presence in the city-state. In 2019, VIS acquired an 8-inch fabrication plant in Singapore from GlobalFoundries.

Meanwhile, NXP shares ownership of the Systems on Silicon Manufacturing Company (SSMC) with TSMC. EDB Investments, the investment arm of Singapore’s Economic Development Board (EDB), founded the fab in 1998 with Philips, whose chip business became NXP, and TSMC. However, EDB Investments sold its stake in SSMC in 2006.

Sotomayor said VSMC builds on decades of NXP’s operations and investments in Singapore.

The country drew more than S$30 Bn of semiconductor investment between 2022 and 2025 and now accounts for one in 10 chips and one in five semiconductor-equipment units worldwide, according to EDB. Singapore is also committing S$800 Mn to semiconductor R&D under its RIE2030 programme, focussing on areas including advanced packaging, heterogeneous integration and photonics.

The direct gains are also in skills and suppliers. Tan said 75% of VSMC’s 1,600 jobs would be professional, managerial, executive and technician roles, and that the fab would create work for local suppliers of speciality chemicals and gases, equipment services, facilities and logistics. VSMC has signed seven agreements with Singapore polytechnics and universities covering internships, hiring and joint research, and is taking part in government career-conversion and graduate traineeship programmes.

Recent investments are spreading across the AI hardware chain. Micron is building an HBM (high-bandwidth memory) advanced-packaging facility in Singapore, expected to expand its packaging capacity from the first half of 2027, while an additional NAND wafer fab is scheduled to provide new cleanroom space from the second half of 2028. Applied Materials opened a $500 Mn manufacturing and R&D campus in Tampines this year, more than doubling its advanced cleanroom capacity in Singapore and adding equipment-production capacity aimed partly at AI-driven semiconductor demand.

VSMC adds another layer, bringing high-volume 300mm manufacturing for speciality technologies that sit between the leading-edge AI processor and the physical systems in which computing is increasingly embedded.

The model carries an obvious trade-off. Singapore is an expensive place to manufacture, with limited land and labour compared with larger Asian economies. Its proposition, therefore, depends less on being the cheapest production location and more on combining infrastructure, suppliers, engineering talent, research capability and access to an established semiconductor cluster.

VSMC is a useful example of how that strategy works. VIS and NXP supply the equity and own the fab. Singapore captures the factory, employment and another link in a semiconductor ecosystem whose value rises as companies place more parts of the supply chain in the same market.

The physical AI thesis raises the stakes. If artificial intelligence remains concentrated largely in cloud computing, the biggest winners will continue to be companies supplying the most advanced processors, memory and data-centre infrastructure. If intelligence increasingly moves into vehicles, robots, industrial equipment and other machines, demand broadens to the less celebrated technologies that connect, power and control those systems.

VSMC is not trying to manufacture the next GPU. Singapore is betting it doesn’t need to.