The advanced packaging race

WHILE much has been said about advanced packaging in the context of Malaysia’s semiconductor sector, little is actually known about what it entails.

The rhetoric is that, since Malaysia has a strong and world-renowned base in semiconductor testing, assembly and packaging dating back to the 1970s, it should be a shoe-in for these local companies to get into advanced packing.

The government is pushing to move the country’s semiconductor industry into advanced packaging, a market estimated to be worth nearly US$60bil today and more than US$100bil in the future.

In a nutshell, advanced packaging is about integrating multiple chips – such as processors, memory and specialised artificial intelligence (AI) accelerators – into a single package.

This innovation was pioneered by wafer fabrication giants, which realised that instead of trying to pack even more functions into a single chip – which has its limitations – they could stack chips or place them closer together.

The challenge then lies in integrating these multiple chips into a single package, hence the term “advanced packaging”.

This differs from traditional packaging, which mainly encloses a single chip in a protective casing and connects it to a circuit board. In advanced packaging, the package itself becomes a critical determinant of performance, power efficiency and heat management.

But can Malaysia’s outsourced semiconductor assembly and test (Osat) players easily make the leap into advanced packaging?

Osat players feature significantly among Malaysia’s listed semiconductor stocks.

Companies such as Inari Amertron Bhd, Malaysian Pacific Industries Bhd (MPI), Unisem (M) Bhd and Globetronics Technology Bhd, for example, have a combined market capitalisation of RM27.08bil.

The problem is, basic Osat work is increasingly shifting to competitors in countries such as Vietnam and China as these services become more commoditised.

The 13% share widely cited as Malaysia’s contribution to global Osat output may already be coming under pressure.

Furthermore, much of that production comes from multinational corporations (MNCs) operating in Malaysia, including Intel, Micron, Infineon and Western Digital.

According to industry veteran Tan Eng Tong, not a single Malaysian company today is capable of undertaking the more advanced levels of semiconductor packaging.

Tan, who worked in the research and product development teams at Hewlett-Packard and Seagate Technology in Silicon Valley before returning to Malaysia for stints at SilTerra Malaysia and the Penang Skills Development Centre, is now playing a key role in helping the country build advanced packaging capabilities.

He has been instrumental in the recent formation of the Malaysia Advanced Packaging Consortium (MAPC).

MAPC comprises five Malaysian companies and aims to achieve what some might consider a moonshot: developing a pilot advanced packaging line for high-bandwidth memory (HBM4) chips.

The goal is to get it up and running in 24 months before capturing 7% of the global advanced packaging market by 2035. A tall order indeed, considering Malaysia is essentially starting from a low base.

Tan says the consortium is structured as a legal entity to allow its members to jointly develop and share intellectual property, while also opening the door for more companies to join over time.

“These companies know that they cannot do it by themselves. This (structure) has never been done before in the country,” Tan tells StarBiz 7.

In fact, not that many Osats around the world have successfully ventured into advanced packaging.

Instead, foundries such as Taiwan Semiconductor Manufacturing Co Ltd, Intel and Samsung have led the way, developing their own advanced packaging technologies and offering one-stop wafer-to-package services for strategic customers.

Increasingly, however, some of this work is being outsourced to Osats to help ease foundries’ capacity constraints.

While hardly any Malaysian entities are in advanced packaging, some of this work is already being done in the country behind the closed doors of the many global semiconductor giants present here.

It is understood that companies such as STMicroelectronics and Infineon do carry out some advanced packaging work, although this has yet to spill over to local companies or create a domestic ecosystem.

Intel is expected to commence operations at its US$7bil advanced packaging facility in Penang later this year.

Notably, China has been building up its advanced packaging ecosystem. Its approach has been to incentivise MNCs to develop local supply chains and capabilities. Industry sources say MNCs in China are gauged on how they help local companies become part of their supply chains.

In comparison, while Malaysia also offers incentives to MNCs setting up operations here, these commitments are not necessarily monitored or enforced as rigorously. One reason is that the government fears stricter requirements could prompt MNCs to relocate, resulting in job losses and negative publicity.

Coming back to advanced packaging, Tan says the key lies in research and development (R&D), without which companies would struggle to know where to begin.

“Many local companies rely on being told what to do by their customers. Instead, local companies ought to kick-start some R&D or pilot advanced packaging capabilities and then start telling their global customers about that,” quips Tan.

For funding, the five companies have committed a total of RM93mil, which is being matched by RM92mil from the government through the Malaysia Science Endowment, administered by the Academy of Sciences Malaysia.

It is left to be seen how MAPC will work out for its five founding members – SkyeChip Bhd, Inari, FusionAP Sdn Bhd, Pentamaster Corp Bhd and NSW Automation Sdn Bhd.

Each company contributes a different piece of the advanced packaging puzzle.

SkyeChip designs the chip that interfaces with HBM, FusionAP develops the advanced packaging technology, Inari provides the manufacturing line, Pentamaster contributes the testing and inspection systems, while NSW Automation supplies the precision equipment used to apply adhesives and underfill materials during chip packaging.

FusionAP is a new company set up by industry veterans, focused on advanced packaging.

The idea is for the pilot advanced packaging line created by the five companies to be showcased and pitched to potential customers.

The significance of MAPC’s goals, Tan says, is to prove that the country has the R&D, production line and advanced packaging capabilities to undertake one of the industry’s most demanding packaging challenges.

According to him, HBM4 was chosen because it represents the cutting edge of advanced packaging. In other words, if Malaysia can package HBM4 chips, it can package most other semiconductor products.

“We are not focusing on HBM only. We want to expand into power semiconductors, silicon photonics and 3D packaging because they use the same mechanisms and the same principles.

“As the consortium progresses, we also want to encourage other companies to work with the five MAPC members. This is how we can develop and grow the vendor ecosystem,” Tan says.

One issue with advanced packaging is its high cost. New kid on the block, FusionAP, estimates it will need RM400mil to set up its own pilot line and another RM2bil to scale up to high-volume manufacturing.

That will not be an easy feat. SkyeChip, for example, raised only RM352mil in its IPO, despite having a long track record.

FusionAP CEO and co-founder Ooi Teng Chow says the company closed its pre-seed funding round earlier this year and is now completing its seed round, with proceeds earmarked for building its technology base and marketing strategy.

The company is targeting a Series A round next year to raise RM400mil. Ooi says he is in talks with potential investors.

Where are our local Osat players at?

Ooi argues that access to capital is not the biggest hurdle. The problem is just that local funds are “not moving fast enough” and tend to come in only after seeing traction from larger investors – by which time, “it is a bit too late”.

Instead, the challenge in advanced packaging is that its processes resemble fab operations more than traditional Osat workflows.

This entails wafer-level processing, ultra-clean fabrication environments, advanced bonding technologies, and far greater process precision than conventional Osat operations.

“A traditional assembly process takes about 14 to 20 manufacturing steps. Advanced packaging requires roughly 10 times as many.

“Not only are there more steps, but they are also much more complex, involving high-level chemicals, physics and materials. This is why it is so difficult,” Ooi says.

In today’s semiconductor ecosystem, local companies already work together across an integrated supply chain. The idea is to replicate this model for advanced packaging.

Among local Osat players, Unisem is arguably the closest proxy to advanced packaging.

The company offers wafer bumping, wafer-level chip-scale packaging (WLCSP), flip-chip assembly and advanced MEMS packaging.

Wafer bumping involves placing thousands of microscopic solder bumps onto a chip, allowing it to be flipped over and directly connected to a package.

Meanwhile, MEMS packaging is used for miniature sensors found in smartphones, cars and industrial equipment.

“Unisem’s advanced-packaging push is more clearly in “classical” Osat advanced flows. Unisem’s strategy anchors it firmly in high-volume advanced packaging for mobile and compact devices, leveraging bumping and WLCSP as key differentiators,” local fund manager Peter Lim Tze Cheng writes in his book, “What I learnt about the Semicon and EMS Industry”.

Meanwhile, Inari is prioritising higher-value packaging technologies, notes Lim.

However, he says Inari’s push remains largely developmental, with much of its current packaging business still centred on conventional radio frequency applications, and has yet to achieve the same level of commercial exposure to advanced packaging as Unisem.

MPI, on the other hand, focuses its advanced packaging efforts on power and automotive modules, especially for silicon carbide and gallium nitride applications rather than AI-centric 2.5D and 3D packaging platforms.

“Unisem and MPI can be considered broadly comparable peers in terms of services, clients, and capabilities.

“However, the key service that Unisem provides, which MPI does not, is wafer bumping.

“Unisem is the only local player that performs wafer bumping in-house. The wafers processed by MPI and Inari are bumped elsewhere before being sent to them for flip-chip packaging,” Lim tells StarBiz 7.

Sources say Infineon is looking to engage Unisem for wafer bumping and packaging work at Unisem’s Gopeng facility.

The move is reportedly part of Infineon’s broader effort to serve customers outside China, amid commercial and customer-specific restrictions affecting the sale of some chips manufactured in China.

FusionAP is targeting more sophisticated advanced packaging technologies, ranging from 2.5D and 3D to silicon photonics and co-packaged optics.

Though Ooi says the company has “surprisingly expanded quite a bit” as well, covering other advanced packaging offerings.

The capabilities of Malaysian-owned Osats today largely sit in the lower and middle rungs of the advanced packaging ladder.

They have yet to establish a meaningful presence in the more sophisticated technologies at the leading edge of the industry.

Tan argues that while Malaysian companies have gained exposure to parts of the advanced packaging process, the gap between these capabilities and the industry’s leading edge remains substantial.

“If you are doing advanced packaging at 80 nm while the industry has moved to 20 nm, (then) tomorrow 10 nm and eventually one nm, you are still behind.

“Saying you can do advanced packaging is one thing. The question is whether you can keep up with where the industry is going,” Tan says.

Hong Leong Investment Bank Research, in a recent report on advanced packaging, notes that successful entry into the segment could drive a structural re-rating of the local technology sector, underpinned by “higher average selling prices and superior gross margins” of 40% to 50% versus 15% to 20% for traditional packaging.

“Global Tier-1 Osats are already showing mix-driven margin expansion, while leading equipment suppliers have seen significant valuation re-rating as markets ascribed structural growth premiums to their critical role in advanced packaging,” the research house said.

Govt efforts limited

The rewards are indeed alluring, but some opine that MAPC may not be up to the task.

Lim points out that Malaysia has attempted similar collaborative initiatives before, like the Penang Automation Cluster Sdn Bhd (PAC), which did not fully pan out as envisioned.

A joint venture set up in 2017 between ViTrox, Pentamaster and Walta Engineering Sdn Bhd, PAC was envisioned to be a one-stop metal component supply chain hub aimed at building and managing the local precision engineering ecosystem across sheet metal fabrication, precision tooling, CNC machining and surface treatment, while also uplifting small and medium enterprise (SME) capabilities.

Although PAC was launched as a one-stop precision fabrication cluster intended to centralise orders, strengthen the local supply chain, and develop SMEs, its original business model did not fully materialise as planned.

In practice, the cluster had to adapt to policy constraints, including pioneer status and sales and service tax issues, and relied heavily on shareholder support rather than evolving into a fully self-sustaining, market-driven platform.

On MAPC, Lim says the government’s efforts in such consortiums or clusters is limited.

“The closest the US government has come to getting involved in the semiconductor business was through the Chips Act. But did the US government help Nvidia develop its chips?

“The reality is that, historically, semiconductors have never really been a business where the government can do very much.

“They can’t persuade multinational semiconductor companies to award advanced packaging contracts to local firms.

“So whenever governments say they’re going to build a semiconductor industry, it is not that easy,” he says.