A new frontier of growth

AS the world continues to make significant strides towards technological advancement, Malaysia finds itself at a critical juncture. The semiconductor industry, one of the cornerstones of Malaysia’s manufacturing sector, is witnessing a surge in demand, driven largely by generative AI (gen AI).

Demand for semiconductors increased with the rise of gen AI, which is used in applications such as ChatGPT and Sora, pushing the semiconductor industry to innovate faster and produce more capable and efficient chips.

On May 28, Malaysia announced the launch of its three-phase National Semiconductor Strategy (NSS), aimed at attracting at least RM500bil of investments in the first phase. This signals the government’s commitment to growing the industry and creating opportunities for business leaders.

At the start of an S-curve

In the latest McKinsey Global Survey on AI, 65% of respondents reported that their organisations are regularly using gen AI, nearly double the percentage from a previous survey just 10 months earlier.

McKinsey also found that gen AI could create an additional 35%-70% of value in economic impact above what other AI and analytics can unlock. With models enabling gen AI growing in complexity faster than devices’ performance can accommodate them, there is a need for for substantial increases in computing infrastructure.

To understand how to navigate this, McKinsey developed three adoption scenarios to forecast gen AI demand in 2030 — a conservative; base case and accelerated adoption. Our follow-up analyses are based on the base adoption scenario, where gen AI growth continues without severe bottlenecks in the mid- and long-term.

Based on this scenario, we discovered that the semiconductor industry is approaching a new S-curve, which means substantial capital expenditure is needed to expand data centres and semiconductor fabrication plants (fabs) while exploring advancements in chip design, material and architecture.

We undertook a three-step approach to estimate the number of new semiconductor fabs required by 2030. Using average fab sizes and utilisation benchmarks, wafer demand was converted into the required number of new fabs to fulfill gen AI demand in 2030.

Our analysis shows that gen AI demand could require 21-25 additional 12’’-equivalent fabs to be built and ramped by 2030 (beyond what has been planned and announced today), along with associated advanced packaging and testing capacity.

Navigating demand ahead

To effectively harness this opportunity, it is crucial for industry leaders in Malaysia to understand the nuanced demands of the gen AI market. In the scenarios developed for gen AI’s effect — in both the B2B and B2C markets — McKinsey’s analysis reveals a massive increase in computing, and thus wafer, demand (especially DRAM/HBM), driven by a spike in computing requirements.

This demand is concentrated in data centres (for training) and edge servers (for inference). The implications for edge devices such as smartphones are also notable, but on a smaller scale.

These scenarios, ranging from conservative to more ambitious, underscore the need for strategic planning and resource allocation to meet the anticipated surge in demand.

For Malaysia, this means prioritising investing in infrastructure that supports the production of high-performance components required by gen AI models and applications. AI data centres will form the bulk of data centre demand from 2027. These data centres need network infrastructure and electricity to function.

Data centres, in fact, could be consuming about 10% of the global electricity production by 2030.

Challenges and opportunities

Malaysia could capitalise on these advancements by building on its current Outsourced Semiconductor Assembly and Test (OSAT) and back-end expertise in the semiconductor value chain.

The first phase of the NSS will focus on advanced packaging and expanding capacity in power, analog, and other chips requiring mainstream technology, while developing local chip design champions.

This could give Malaysia a strong foundation from which to position itself at the frontier of the semiconductor value chain — that is, in front-end fabrication and advanced chip design.

Despite this promising outlook, the journey ahead is not without its challenges. The semiconductor industry in Malaysia must manage the intricate balance of demand and supply chain disruptions, while addressing the stress that increased power generation will place on the environment. Securing talent, from construction labour to data engineers, and making sufficient living options available to them is also critical.

To address these challenges, a concerted effort is required from both the public and private sectors to invest in the necessary infrastructure, foster a skilled workforce, and create a conducive regulatory environment that encourages innovation and protects intellectual property.

In embracing gen AI in the semiconductor industry, Malaysia is not just taking part in a global trend but actively shaping its future in this sector and beyond, and ensuring the country’s contribution to the world of global technology.

This article first appeared in Star Biz7 weekly edition.