Published by BusinessToday, image by BusinessToday.
On 5th March 2025, it was announced that Malaysia would be entering into a partnership with the United Kingdom (UK)-based Arm Holdings in an effort to promote our semiconductor industry, thus moving from the back-end of the semiconductor supply chain to the very front of it.
The collaboration, while still lacking some key details, is a long-term plan to ensure our future talent has the capability for complex chip design.
The government will invest USD 250 million, roughly equivalent to RM1.11 billion, over the next decade in Arm Holdings to expand Malaysia’s capacity in the semiconductor industry (The Edge, 2025). This partnership, according to Arm Holdings, will allow the Malaysian government to leverage the Arm Compute System (ACS) technology and the Arm Flexible Access programme to build its own integrated chip (IC) design environment (Arm Holdings, 2025).
Arm Holdings is also committed to training 10,000 semiconductor specialists in Malaysia in their technology.
It should be emphasised that this is not an investment that will see a return in the near future, or even within a decade. Chip design is extremely complicated. It involves decades of experience and resources; if we were to start from the ground up, we would be spending far greater amounts of money and time on it.
Malaysia has experience in the semiconductor industry, but our experience differs from what is required to design a functioning chip. We excel in the very late stage of chip production, which is packaging and testing.
Even experienced companies, when design brand-new product lines, often hire experts from other established firms in the field. Intel, when trying to enter the dedicated Graphic Processing Unit (GPU) market, hired former employees from NVIDIA and AMD to help design its own GPU (CRN, 2021; PCMag, 2021).
In the context of Malaysia, where the government is actively striving to enhance our capability in producing chips for Artificial Intelligence (AI) tasks, designing a functioning chip from scratch is a substantial undertaking.
There are some unique cases where newly established players manage to design chips that are sufficiently advanced for AI applications, but these scenarios involve additional factors.
For example, China’s Moore Threads was established in 2020, and within five years it released multiple generations of consumer-grade and server-grade products. However, Moore Threads was founded by a former global vice-president of NVIDIA and general manager of NVIDIA China, and reports suggest that Moore Threads, along with other Chinese semiconductor companies like InnoSilicon, purchased intellectual properties (IP) of GPU design from UK-based semiconductor and software design company, Imagination Technology (Tom’s Hardware, 2024).
If anything, this demonstrates that experience is the most crucial aspect of chip design.
However, instead of scouring for and poaching experienced employees from other established firms, partnering with Arm Holdings to train local semiconductor talent may produce better outcomes.
While the x86 architecture family still dominates personal computer, workstation, and cloud computing segments, the ARM architecture of Arm Holdings is actively on the rise in these segments. Moreover, ARM chips hold 99% of the mobile CPU market share and 40.8% in the automotive sector, for an overall share of 48% in Arm Holdings’ related markets (Forbes, 2024).
Virtually every smartphone within our reach is powered by chips designed using ARM architecture, irrespective of the brand. Most of Samsung’s Galaxy AI features for their newest smartphones run locally, meaning the mobile CPUs process them without resorting to cloud computing. The same applies to Apple’s Apple Intelligence on their devices.
The experience and knowledge of Arm Holdings, if successfully transferred to local specialists, will undoubtedly propel us to the forefront of chip design.
Under the current partnership arrangement, in which Arm Holdings offered to train 10,000 semiconductor specialists in Malaysia, it is being carried out under the Arm Flexible Access programme. The programme, even though it does not include the latest technology from Arm Holdings, does permit low- to no-cost access to a wide range of their IPs, tools, and training.
Obviously, it would take a long time before any return on investment is realised, and even then, our initial step into advanced chip design could still be very challenging.
This partnership with Arm Holdings undoubtedly has potential to bring many benefits to our semiconductor industry, especially those focusing exclusively on designing chips—commonly known as fabless companies.
However, if we aim for chip independence, being able to design chips is merely the starting point.
As the United States (US) reportedly imposed an export restriction on Malaysia, which only permits the import of up to 50,000 GPUs over two years and further restricts the deployment of computing capacity to a maximum of 7% of our total capacity, chip design is not the only field we should be prioritising (New Strait Times, 2025).
As crucial as chip design capability is, we must also be able to manufacture chips, which extends beyond the packaging and testing we already perform for Intel, AMD, and numerous other semiconductor giants, to include wafer fabrication.
While the global focus is now mainly on fabricating 3 nm process nodes and moving to 2 nm nodes from this year onward for advanced chips, our fabrication plants currently operate on process nodes ranging from 300 nm to 130 nm. The wafer fabrication plant with the smallest process node planned is 28 nm if the partnership with Foxconn is realised, as the latest publicly available update indicates that the fabrication plant remains on schedule as of August 2022 (The Edge, 2022; The Star, 2022).
Thus, even if our local firms eventually succeed in designing highly advanced AI chips, we would still rely on foreign foundries to produce the wafers, leaving us vulnerable to external interference. For example, the world’s largest pure-play foundry, TSMC, has repeatedly complied with US directives to restrict chip sales to China, with a recent measure involving a ban on exporting chips featuring 30 billion transistors or more, as well as chips manufactured using 16 nm process nodes or smaller, to China (Tom’s Hardware, 2025).
However, a highly advanced semiconductor fabrication plant is extremely expensive, with some estimates indicating a production line of 3 nm process nodes capable of producing 40,000 wafers per month would cost between USD 15 billion to USD 20 billion (Lin, 2023). The investment TSMC made to establish a fabrication plant focusing on 5 nm to 3 nm chip fabrication in 2016 was valued at USD 15.7 billion at the time (EE Times, 2016).
This is without accounting for the investment needed to reform and strengthen our severely compromised education system, so we can develop the requisite talent for the industry.
Furthermore, many technology giants, such as Qualcomm, Google, and NVIDIA, have begun researching and developing RISC-V, an open-standard instruction set architecture (ISA), for both end-user and server technologies (Qualcomm, 2023).
In contrast to the ARM architecture—a family of RISC ISA for which we would need to pay royalties or licensing fees to Arm Holding in order to develop and manufacture related chips—RISC-V has no such requirement.
Reports also indicate that the Chinese government is poised to introduce new policies encouraging the use of RISC-V chips to reduce its dependence on foreign technologies (Tom’s Hardware, 2025a), which only underscores brighter RISC-V’s potential.
Ultimately, the semiconductor industry is extraordinarily intricate, and to reduce our reliance on foreign technologies, we cannot neglect any segment of the supply chain by focusing solely on the front-end of chip design.
In the meantime, the partnership with Arm Holdings could well be the catalyst needed to inaugurate an entirely new chapter in Malaysia’s semiconductor industry.
Chia Chu Hang is a Research Assistant at EMIR Research, an independent think tank focused on strategic policy recommendations based on rigorous research.
