Singapore’s Photonics Bet Moves Towards Industrial Scale
AMF’s integration into GlobalFoundries and a new packaging agreement reveal Singapore’s ambition to help build the optical infrastructure of AI.
Singapore’s photonics ambitions are entering a harder phase. With Advanced Micro Foundry now part of GlobalFoundries and a new agreement targeting co-packaged optics manufacturing, the city-state is strengthening its position in AI’s physical infrastructure. The test is whether decades of research can translate into reliable production, lasting customer relationships and influence over where industrial value accumulates.
The route from laboratory expertise to strategic influence runs through factories, packaging lines and customers with little tolerance for failure.
🟧 Why is Singapore moving into the connections behind AI?
As AI systems grow, moving data becomes an engineering constraint in its own right. Bandwidth, electrical losses and communication power increasingly shape what additional computing capacity can deliver.
Silicon photonics uses light to transmit data through structures manufactured on silicon. Its industrial appeal lies partly in the ability to draw on established semiconductor processes while addressing a different part of the computing system.
Singapore has been building towards this opportunity for years. A*STAR’s Institute of Microelectronics launched its silicon photonics research programme in 2007. The LUX Photonics Consortium followed in 2015, bringing research and industry together; AMF emerged as a spin-off in 2017.
That history matters. An ecosystem assembled over decades offers a stronger foundation than an investment announcement. Its value still depends on what customers can actually buy.
🟧 Does selling AMF strengthen Singapore—or move the decisions elsewhere?
GlobalFoundries announced its acquisition of AMF in November 2025, incorporating the Singapore-based foundry’s manufacturing assets, intellectual property and expertise into its international photonics business. It also announced plans for a silicon photonics research centre in Singapore, working with A*STAR.
The commercial logic is compelling. A specialised foundry can gain access to a wider customer base, investment capacity and a longer technology roadmap. Singapore-developed capabilities can travel further through a global manufacturing network.
The strategic bargain is less simple. Local expertise enters a corporate structure in which investment priorities and customer relationships are decided across several countries.
For Singapore, the acquisition can therefore be read as an opportunity to deepen its industrial role. Whether it does so depends on the responsibilities retained locally: developing processes, solving customers’ problems and shaping the next generation of products.
A factory address alone does not establish that influence.
🟧 Why could packaging determine how much of the opportunity Singapore captures?
On 25 September 2026, STATS ChipPAC announced a licensing agreement for A*STAR IME’s advanced packaging technology for co-packaged optics. The stated aim is to improve manufacturability and accelerate the transition from research into high-volume production.
Co-packaged optics brings optical components close to electronic chips within a shared package, shortening the electrical paths used to move data. In networking switches, that changes how optics are integrated into the system.
It also makes packaging a demanding engineering task. Photonic chips, electronics, fibres and light sources must work together under tight performance and thermal constraints. NVIDIA’s account of its own CPO development illustrates how much depends on process control, fibre attachment and reliable integration.
The Singapore agreement targets that industrial challenge. Licensing creates a route towards production; repeatable manufacturing, acceptable yields and customer qualification must establish whether the route works.
🟧 Will better optical connections be enough to persuade customers?
For operators, CPO must compete with an established alternative: pluggable optical transceivers that can be inserted into switch ports and replaced individually. That familiar maintenance model sets a practical benchmark for any more integrated architecture.
Lower energy consumption is valuable. Adoption also depends on whether reliability, repair procedures and the total cost of operation justify changing how a network is built and maintained.
Closer optical integration can shorten electrical connections and improve efficiency. It also places demanding optical components alongside heat-generating electronics. NVIDIA addresses part of the serviceability challenge by putting its laser sources in accessible, replaceable modules; integration does not require every component to become inaccessible.
The efficiency claim therefore needs a clear boundary. Savings in data movement do not automatically become equivalent savings across an entire AI facility.
For Singapore’s suppliers, the commercial test is sustained performance under customers’ operating conditions. The decisive evidence will come from qualified products, repeat orders and production economics.
A promising component earns attention. A dependable system earns a place in the supply chain.
🟧 Can Singapore turn a specialised manufacturing role into strategic influence?
The opportunity is to become a supplier whose expertise makes complex systems easier to build and deploy. Foundry processes, integration knowledge and packaging capability can reinforce one another when customers commit to a production platform.
The corporate geography adds another layer. STATS ChipPAC belongs to China-based JCET Group, while AMF has joined US-headquartered GlobalFoundries. Singapore’s ecosystem is embedded in international ownership structures.
That gives Singapore a potential role connecting local research with manufacturing networks and customers across markets. It also exposes a dependency: global companies can distribute investment and production across competing locations. Being a useful industrial junction requires continual renewal of the capabilities that bring those networks together.
Our reading is that Singapore’s strongest position would combine internationally connected companies with capabilities that remain difficult to reproduce elsewhere. Research depth matters, but so do experienced engineers, production discipline and the ability to resolve failures quickly.
The question is how much responsibility Singapore’s ecosystem can retain as volumes grow—and how much of the resulting value stays with it.
Signal
Singapore’s photonics bet is entering its industrial test. The decisive signal will be qualified products, repeatable yields and commercial deployment—with Singapore retaining the process expertise that makes those systems work.
Credit
Illustration: Altair Media Asia / OpenAI — AI-generated.
Caption
A conceptual illustration of silicon photonics against Singapore’s skyline, reflecting the city-state’s ambition to turn research expertise into an industrial role in AI’s optical infrastructure.
