The AI boom has made memory chips the star of the show; Singapore stakes its role in advanced packaging

Over the past two years, artificial intelligence (AI) has dominated the global tech narrative. From the explosion of large language models in 2023 to the race for chips and advanced process nodes in 2024, the spotlight has remained on delivering faster and more powerful processing capabilities.

But the tech industry evolves quickly. By the second half of 2025, memory chips – long considered a secondary component – took centre stage as reports of tight inventories emerged.

In the semiconductor industry, common memory chips include dynamic random-access memory (DRAM), NAND (short for “Not AND,” referring to the logic structure within the memory cell), and high bandwidth memory (HBM), which is developed based on DRAM technology.

According to research firm IDC, severe shortages in DRAM and NAND chips began in the second half of 2025. In 2026, supply growth for DRAM and NAND is projected at 16 per cent and 17 per cent respectively, far below historical averages. These tight supply conditions are expected to persist through 2027.

What role do memory chips play in AI?
 

Comparison of major memory technologies

Source: SPH Media

DRAM is a form of volatile memory, meaning data disappears when power is cut. It serves as a system’s working memory during operations. NAND, on the other hand, is non-volatile – meaning it retains data even without power. It is used in solid-state drives (SSD), smartphones, and data centres.

HBM is an enhanced version of DRAM, where multiple DRAM chips are stacked vertically and placed close to processors, delivering extremely high data bandwidth.

In the AI era, DRAM enables handles real-time task processing, NAND stores vast amounts of knowledge, and HBM delivers ultra-fast data throughput.

As AI models scale, the bottleneck on computing performance is no longer just compute chips, but also data transfer, and storage capacity.
 

Prices surge amid supply shortages; DRAM expected to jump 90 per cent in Q1 2026

Market observers note that this supply shortage reflects a strategic reallocation of global wafer capacity and memory chip prices are expected to climb further.

Nomura estimates that in Q1 2026, DRAM and NAND prices will rise by 90 per cent and 60 per cent, respectively, significantly higher than earlier forecasts of 56 per cent and 40 per cent. For the full year, commodity DRAM and NAND prices could reach 176 per cent and 146 per cent respectively.

The global memory market is currently dominated by Samsung, SK Hynix, and Micron Technology. According to Counterpoint Research, these three companies accounted for over 90 per cent of global DRAM market share in Q3 2025 – accounting for 33 per cent, 34 per cent, and 26 per cent respectively.

In HBM, their dominance is even more pronounced. SK Hynix holds 57 per cent, while Samsung and Micron account for 22 per cent and 21 per cent respectively.

Rising prices directly boost memory manufacturers’ profits. Nomura estimates that SK Hynix’s 2026 operating margins for DRAM and NAND could reach 76 per cent and 57 per cent respectively.

As AI customers race to secure supply, the latest generation of HBM – “HBM4” – commands even greater pricing power.

Nomura notes that while Samsung uses more advanced processes and faces higher production costs, high-speed versions of HBM4 could still command a premium of 30 per cent to 40 per cent amid the tight supply.

This has significant implications for everyday consumers.

Many research institutions believe that rising demand for high-end memory will crowd out production capacity for consumer electronics, as each wafer allocated to AI processors potentially displaces memory supply for smartphones and PCs.

As consumer electronics face rising cost pressures, consumer-oriented tech stocks have come under selling pressure – with Qualcomm, Nintendo, Logitech, and Chinese manufacturers BYD and Xiaomi among those feeling the strain.
 

Singapore: A major NAND production base

Maybank economists Chua Hak Bin and Brian Lee note that Singapore is a major global NAND production base. With Micron’s HBM chip plant and Taiwan’s United Microelectronics Corporation commencing local production, Singapore’s chip capacity and output is expected to increase, further cementing its position in the AI supply chain.

Notably, 98 per cent of Micron’s NAND flash chips are produced in Singapore. The company recently announced an investment of over S$30 billion to build Singapore’s first double-storey wafer manufacturing fab, which will create 1,600 jobs when it begins production in the second half of 2028.
 

Corporate Vice President and Singapore Country Manager for Micron Technology, Joshua Lee, said that the new wafer fab provides the necessary capacity for ongoing technological transformation.

He said: “This enables us to meet long-term market demand for advanced memory solutions and further reinforces Micron’s strong confidence in Singapore. By co-locating R&D and manufacturing, we strengthen collaboration, accelerate time-to-market, and deepen industry-academia partnerships.”

Overall, Singapore accounts for one in ten chips, and one-fifth of global semiconductor equipment production. The semiconductor industry contributes close to 6 per cent of Singapore’s GDP and employs over 35,000 people.

During the AI memory supercycle, Singapore’s electronics cluster recorded strong growth of 25.1 per cent in the fourth quarter last year and 12.7 per cent for the full year. Non-oil domestic exports (NODX) of electronics and related equipment rose 13 per cent for the year, driving overall NODX growth of 4.8 per cent.

Although current growth is impressive, Singapore is not resting on its laurels.

Deputy Prime Minister and Minister for Trade and Industry Gan Kim Yong recently noted in Parliament that the government is actively encouraging semiconductor companies and promising startups to establish and expand R&D activities in Singapore to strengthen technological capabilities in high-end chip manufacturing.

“High-end chip manufacturing is attractive because it is technologically complex and capital-intensive. It can potentially produce higher value but also carries higher risks.”

He pointed out that chip manufacturing is land and energy intensive.

“This requires Singapore to be selective in the projects we pursue. We adopt a multi-pronged strategy to grow high-end chipmaking in Singapore.”

The Singapore Economic Development Board (EDB) said in response that Singapore continues to attract semiconductor investments amid strong global demand for AI-related activities. Companies are leveraging Singapore’s strengths to expand capacity and develop technologies.

"In the long term, this will strengthen Singapore’s leadership position in the semiconductor industry and allow us to continue bringing in good jobs for our workforce,” said EDB.
 

SSIA: Singapore as a key node in advanced manufacturing and packaging

Singapore Semiconductor Industry Association (SSIA) Executive Director Ang Wee Seng highlighted that in the global memory ecosystem, Singapore’s role goes beyond a manufacturing base – it is a high-value, innovation-driven hub.

He noted that Singapore is becoming a key node for advanced manufacturing and advanced packaging.

“As a trusted, neutral, and well-connected platform, Singapore brings together multinational corporations, equipment and materials suppliers, and research institutions to promote collaboration across the memory value chain. Singapore is also emerging as a talent and innovation hub in advanced packaging, testing, materials, and system integration,” said Ang.

Bloomberg Intelligence projects that broader adoption of AI chips will drive the global advanced packaging market to grow sevenfold to US$80.5 billion (S$101.7 billion) by 2033. This means a compound annual growth rate of 26 per cent – far exceeding the semiconductor industry’s overall expected growth rate of 10 per cent.

Prime Minister and Finance Minister Lawrence Wong mentioned in his Budget 2026 speech that Singapore will actively build leadership positions in industries with strong growth potential, including advanced packaging.

Ang said Singapore’s advanced manufacturing and packaging efforts are particularly focused on HBM and AI-related chips that require complex integration.
 

While HBM’s design drastically shortens signal transmission distances and boosts bandwidth, stacking more layers increases the manufacturing complexity significantly, magnifying issues such as bonding precision.

As the 12-layer HBM4 is now entering volume production, 16-layer versions are also emerging. At the recent 2026 Consumer Electronics Show, SK Hynix showcased the world’s first 16-layer HBM4 sample.

Another memory giant also joined the race.

Despite earlier reports that it was unable to supply NVIDIA, Micron’s Chief Financial Officer Mark Murphy recently clarified that the company has achieved mass production of HBM4 and has begun supplying its customers.
 

Micron’s investment in Singapore brings spillover effects to neighbouring countries

On Micron’s recent investment in Singapore, Ang said it will generate significant and widespread spillover effects. For startups, Micron’s presence will serve as a “powerful magnet”, attracting deep-tech enterprises in advanced packaging, materials science, automation, and AI manufacturing.

He said this will encourage local suppliers to scale and enhance capabilities to integrate more deeply into the global memory supply chain. It will also create numerous high-skilled jobs in engineering, manufacturing, packaging, testing, and supply chain management.

Market Research Future notes that North America holds the largest share of market demand and is expected to reach US$75 billion by 2032, driven by technological innovation, consumer electronics, and sustained data centre demand. Europe’s market is projected to grow to US$45 billion, while Asia-Pacific is expected to expand to US$60 billion.

As for supply and demand in ASEAN, an Alibaba.com survey indicates that although the region is affected by global chip shortages, its relatively lower consumption levels and strategic position in the global semiconductor supply chain present unique opportunities.

Several major memory manufacturers are establishing or expanding production facilities in the region, especially in Singapore and Malaysia, potentially giving compliant exporters local supply advantages.
 

Ang notes that Singapore has strong capabilities in wafer fabrication, testing, and advanced packaging, while regional countries such as Malaysia, Thailand, and Vietnam contribute through assembly, testing, and back-end manufacturing, forming a complementary, regional division of labour.

However, he added that Singapore still faces key gaps in certain upstream critical materials, specialised equipment, and chip design and process development.

“These gaps present important opportunities for Singapore and regional enterprises to invest in R&D, build partnerships, and localise the value chain.”
 

Singapore plays multiple roles in the AI supercycle

Beyond memory chips, Singapore’s complete semiconductor ecosystem enables it to participate in different segments of the broader AI supercycle.
 

Singapore's semiconductor value chain layout in the AI super cycle

Source: SPH Media

Micron, Infineon, and NXP Semiconductors are typical integrated device manufacturers (IDMs), that handle chip design, wafer fabrication, and packaging and testing.

Extending upstream and downstream, Singapore also hosts semiconductor equipment companies that supply manufacturing equipment to wafer fabs and precision equipment for packaging and testing.

Further upstream, there are also critical integrated circuit (IC) design and engineering segments, including chip architecture planning, circuit integration, and engineering for manufacturability.

Last year, foundry giant GlobalFoundries announced the acquisition of local silicon photonics foundry Advanced Micro Foundry. The latter’s technology is crucial for AI-driven data centres and communications.

Silicon Photonics: Replacing electricity with light

Traditional copper interconnections face limitations in speed, latency, and energy consumption at AI-scale data traffic. Silicon photonics transmits data using light instead of electrical signals, offering higher speed, lower latency, and lower power consumption compared with electrical transmission.

Senior Vice President and General Manager of GlobalFoundries Tan Yew Kong said that Singapore has established a strong foundation in advanced semiconductor manufacturing and applied research. He added that the future challenge is less about insufficient capability but keeping pace with accelerating global demand for silicon photonics.

He added that as silicon photonics plays an increasingly important role in next-generation AI infrastructure, the immediate priority is translating research results into large-scale practical applications.

This includes strengthening manufacturing ecosystem integration, transforming silicon photonics advances into scalable and reliable systems that can be efficiently manufactured, tested, and deployed. This also means cultivating engineers with cross-disciplinary skills in chip manufacturing, optics, and system design to support this transformation.

“As global demand grows, these collective efforts will support Singapore’s continued development as a trusted hub for silicon photonics manufacturing and application innovation.”
 

Source: Lianhe Zaobao © SPH Media Limited. Permission required for reproduction.

Note: EDB sought permission from Lianhe Zaobao to independently produce an English translation of "悉看大势：AI产业爆炸式增长 存储晶片成主角被哄抢 小红点卡位封装大舞台", which was published 22 February 2026.