This article first appeared in Forum, The Edge Malaysia Weekly on March 31, 2025 - April 6, 2025

Earlier this month, Prime Minister Datuk Seri Anwar Ibrahim announced a bold new initiative called Malaysia’s Silicon Vision, which aims to produce the first Made in Malaysia artificial intelligence (AI) chip. As he stated, the chip will be “designed, manufactured, tested and assembled here, and sold to the rest of the world” within a 10-year time frame. The initiative seeks to empower 10 local design houses with the capability to develop high-end silicon designs for AI applications, with the goal of generating multibillion-dollar revenues.

As part of this effort, Malaysia is partnering with the UK’s Arm Holdings to gain access to seven of its high-end compute subsystem intellectual property (IP) designs, train 10,000 new integrated circuit (IC) design engineers, and facilitate the development of locally designed semiconductor products. This marks the first time that a government entity has purchased access to expensive IP designs to reduce R&D costs for local design houses and jumpstart the local silicon design ecosystem.

While investing RM1 billion to secure Arm’s IP access is a strong starting point for Malaysia’s Silicon Vision, it is however insufficient to fully realise the goal of a truly Made in Malaysia AI chip. Achieving this vision will require additional multibillion-dollar investments to empower local design houses with the necessary infrastructure and establish Malaysia’s first advanced wafer fabrication facilities.

Designing high-end AI silicon chips, with their complex architecture and massive design footprints, requires design houses to establish and maintain high-performance server farms on site. This should be a key requirement set by the Ministry of Economy to ensure the success of Malaysia’s Silicon Vision. Interested design houses must be willing to invest millions of dollars to build these facilities, as standard office PCs and laptops simply cannot provide the necessary computing power, storage and security infrastructure. High-performance server farms are essential for handling the significant computing power demands of high-end silicon design activities. These facilities must include powerful server machines, large-scale storage systems and robust security measures such as firewalls, intrusion detection and data encryption.

AI-focused silicon designs, such as the one envisioned under this initiative, typically contain billions of transistors and components. Ensuring that all aspects of the design — from planning and development to verification — are executed flawlessly is a monumental challenge. Engineers will require significant computing power and access to top-of-the-line electronic design automation (EDA) tools to efficiently carry out tasks from register transfer level (RTL) design to physical layout design. Having the right infrastructure in place is critical to ensuring a smooth first tape-out of Malaysia’s AI silicon chip, with minimal design bugs. A failure to have a robust and fully tested silicon design may lead to a non-functional chip returning from the foundry, which would severely jeopardise the entire vision.

Design houses should not be allowed to rely on public cloud providers such as Amazon AWS or Microsoft Azure for their silicon design activities. The risk of design data breaches in public cloud environments is too great. Secure, on-site server farms provide better protection for proprietary AI silicon designs and any third-party IP that design houses have access to. A single data breach could be disastrous, exposing Malaysia’s AI chip designs to competitors or malicious actors. This could lead to significant setbacks, including the need to restart the silicon design process from scratch, legal consequences from third-party IP providers, and a loss of confidence in the vision.

Setting up high-performance server farms and providing access to top-tier EDA tools will require a multimillion-ringgit capital investment. To ease this burden on design houses, the government could explore alternatives such as building centralised high-performance computing facilities that local design houses have access to. Regardless of the approach chosen, the solution must be able to provide secure, high-performance computing infrastructure to silicon design engineers.

Malaysia’s Silicon Vision outlines an ambitious goal: to manufacture the country’s first AI chip locally. This means that once the first AI silicon design is ready for fabrication, Malaysia must have the capability to manufacture high-end silicon wafers locally. In line with this vision, the silicon designs of the first Made in Malaysia AI chip will integrate Arm’s most advanced graphics processing unit (GPU)/ central processing unit (CPU) IPs. These advanced Arm IPs are designed to be manufactured with advanced technology nodes and this dictates that our first Made in Malaysia AI wafer chips be manufactured with the latest technology nodes.

Currently, a leading-edge silicon wafer fabrication facility operates at process/ technology nodes below 10nm, and only a handful of foundries worldwide possess the capability to manufacture silicon wafers at this level. One of the most prominent is Taiwan Semiconductor Manufacturing Co (TSMC), headquartered in Taiwan, which supplies major tech giants such as Nvidia, AMD and Apple. Just this month, Nvidia released its latest H200 AI chip manufactured with TSMC’s 4nm technology node.

In Malaysia, fewer than 10 companies own dedicated silicon wafer fabrication facilities, and none are equipped to manufacture high-end silicon wafers with the most advanced technology nodes. Simply put, with the current capacity of Malaysian foundries, it is impossible to manufacture the country’s first Made in Malaysia AI chip locally.

To turn this vision into reality, the government must step in to support local foundries with loans and grants, enabling them to undertake the multibillion-dollar upgrades necessary to manufacture high-end silicon wafers. Alternatively, the government could offer incentives to attract global semiconductor giants like TSMC and Samsung to establish advanced node foundries in Malaysia. Such interventions are crucial to ensure that Malaysia’s foundry ecosystem is ready in time for the tape-out of the country’s first AI silicon design.

To conclude, the government’s decision to invest RM1 billion in jumpstarting the local silicon design industry is a bold leap of faith. However, much more needs to be done to fully realise Malaysia’s Silicon Vision. I would personally love to see this vision come true knowing firsthand the power that it brings to the nation with the advanced silicon design capability. It is all over the news that the world’s greatest powers are engaged in a fierce chip war, vying for access to the latest and greatest AI and 5G chip designs. This vision would definitely elevate Malaysia’s position on the world stage should it become a reality.

Nasiruddin Nasirin is a senior silicon design engineer at Advanced Micro Devices Inc

Save by subscribing to us for your print and/or digital copy.

P/S: The Edge is also available on Apple's App Store and Android's Google Play.