Semiconductors act as one of the fundamental building blocks for technological advancement. Companies across end markets are continually demanding increasingly complex, high-performing and efficient chips, across an expanding number of applications and different system requirements – all to drive innovation within their own products.
For semiconductors to meet these increasing demands, they need to constantly innovate. At the ‘leading edge’, innovation is focused on shrinking the size of the ‘node’, and thus increasing the number of transistors per chip.
Over several decades, progress in this area has seemingly followed Moore’s Law: that the number of transistors on a chip will double every two years, leading to an exponential increase in computational power and efficiency.
Taiwan Semiconductor Manufacturing Company (TSMC) has been at the forefront of this progress and is able to produce chips at the three nanometre (nm) ‘node’, with a two nm node offering 30% more efficiency expected by 2025.
Even at the ‘trailing edge’ – older, more mature semiconductor technologies – innovation is still important, focused on improving the reliability and efficiency of chips, potentially through material science.
To facilitate this innovation, the semiconductor industry is one the most intense in research and development (R&D) as a proportion of sales, second only to pharma in 2022.
While innovation is clearly a significant cost for semiconductor companies, a high investment allocation to R&D not only creates high barriers to new entrants, but allows the industry to continue growing, not only by providing improved, more complex and powerful semiconductors, but facilitating the development of new technologies in different categories and industries – as seen with the progress made in generative artificial intelligence (AI) across sectors over the course of 2023. Consequently, the semiconductor market has grown rapidly over the past decade.
Two sectors we would highlight where there has been this significant growth in the use of semiconductors are the automotive and datacentre industries.
Growth in the automotive semiconductor end-market (14% annually through 2030), is expected to far outpace the broader sector, with long-term trends such as digitalisation, electrification and autonomous driving increasing the complexity and content of semiconductors within vehicles. Semiconductor content per vehicle averaged around $600-700 in 2021 and is expected to reach $2,000 by 2030.
Semiconductor content roughly doubles from an internal combustion engine vehicle to a battery or plug-in hybrid electric vehicle (EV), and the expectation that EVs (full or hybrid) will make up more than 50% of car production by 2027 is a significant tailwind to demand.
Datacentre semiconductor usage is expected to be the second fastest growth sector over the remainder of the decade (+13% annually). We have seen rapidly increasing demand for cloud computing capacity across industries, not only due to more businesses and industries migrating on-premise infrastructure to the cloud, but the rise of ‘Big Data’ and more data-centric analytics boosting demand for more efficient servers, and thus more complex, efficient and powerful processors.
Furthermore, advances in artificial intelligence, particularly following developments in generative AI during 2023, require specialised hardware that has the ability to handle complex and energy-intensive computations, with datacentres providing the necessary infrastructure for both the training and running of AI systems.
The increasing demands and complexity of the underlying products within these two industries are resulting in greater ‘semiconductor content per device/unit’ (measured as the dollar amount). However, this phenomenon is also present across many other use cases.
Costs associated with fitting more transistors onto a given area are one reason why the semiconductor industry is one of the most R&D-intensive industries. Advancing to smaller technologies becomes increasingly complex, meaning that at each ‘node’ (the minimum distance between transistors on a chip) the cost increase is significant.
Whilst no data is available yet, the three nm node is expected to be $1bn. The more significant cost, however, is the cost of the fabrication module (semiconductor factory), which is expected to be $5-6bn at the five nm node.
The cost of TSMC’s three nm proposed fabrication plant is estimated to be $20bn, highlighting a significant acceleration in the costs required to progress the technology further.
Geopolitical tensions have driven governments to offer significant subsidies to accelerate the onshoring of chipmaking facilities.
The Covid-19 pandemic exposed vulnerabilities in the semiconductor supply chain, with varying lockdown measures across regions driving severe shortages in many areas of the chip markets.
The supply chain disruption exacerbated existing underlying tensions and concerns over national security and supply chain stability. Strained relations between the US and Chinese governments have resulted in sanctions and restrictions over exports between the two countries since 2017, with the US ultimately aiming to limit China’s ability to acquire and manufacture chips at advanced nodes and thus slowing efforts to gain a meaningful foothold in industry and become self-sufficient.
The US’s vulnerability is clear. Whilst accounting for 25% of global semiconductor demand, the US possesses just 12% of global manufacturing capacity. Other regions have also weighed in to obtain their own slice of the rapidly growing and critical industry, with the EU, Japan, Korea and India all offering additional subsidies to incentivise chipmakers to build on their shores.
One concern with government subsidies is that they typically lead to the misallocation of capital. We do not expect this to be the case for the semiconductor industry since the subsidies are coming at a time when companies need to ramp up capacity in order to service the long-term underlying growth trends.
Either way, semiconductor equipment manufacturers such as Lam Research and KLA Corp stand to benefit from these long-term capex cycles.
Despite their superior characteristics, Guinness’ holdings have, on average, typically been in line with the MSCI World Semiconductor Index valuation on a price-to-earnings (P/E) basis.
However, recently a significant discount has emerged. With the exception of Nvidia, all of our holdings are at a discount to the MSCI World Information Technology Index and the MSCI World Semiconductor Index today.
We therefore continue to see good opportunities for holdings in the semiconductor sector and believe they continue to have good pathways for future growth, and potential outperformance.
Dr Ian Mortimer and Matthew Page are fund managers of Guinness Global Innovators. The views expressed above should not be taken as investment advice.