In 1965, based on only seven observations, Intel co-founder Gordon Moore predicted that the number of transistors on an integrated circuit board would double approximately every two years. His prophecy was dubbed Moore’s Law. Fast-forward more than 50 years, and even he would likely be astonished by the outcome of his prediction. The number of transistors on a same-sized circuit board has increased from under 1 000 to over 100 billion today, with a mere three nanometres separating them!
This has had profound consequences for our modern world, with faster computers leading to multiple technological breakthroughs that have forever changed our lives (think PCs, the internet, smartphones, better healthcare, artificial intelligence, to name just a few).
FOUNDRIES TAKE OVER
A few decades ago, most semiconductor companies made their own chips. However, as the industry moved beyond Moore’s Law, the complexity and costs associated with this process became increasingly onerous, leading to the outsourced model (in the form of foundries) slowly but surely taking over chipmaking.
The two main semiconductor types in the industry are logic and memory. Logic chips are those used for computations (such as microprocessors in CPUs and GPUs), while memory is for storing data (Samsung is the biggest player here). Within logic semiconductor foundries, there are only three players left at the cutting edge (seven nanometres and below): TSMC, Samsung and Intel.
However, considering that Samsung and Intel are integrated device manufacturers (producing chips for their own end products), their foundry businesses are small compared to TSMC, which now controls more than 80% of the leading-edge foundry market. It’s very likely that one of the devices in your home today contains a chip manufactured by TSMC.
WHY WE LIKE TSMC
TSMC was founded in 1987 by Morris Chang, who foresaw that the foundry model would become the future direction of the industry. Today, TSMC is the world’s leading manufacturer of cutting-edge chips used in smartphones, computers, data centres, autos and various other electronic devices.
The company’s biggest customers include Apple, Nvidia, Qualcomm, AMD and Broadcom. More than 80% of its revenue is derived from high-performance computing (data centres) and smartphones. At the time of writing, TSMC’s market capitalisation is nearly US$800 billion, placing it among the 10 largest companies in the world.
While the process of manufacturing semiconductors is very capital-intensive and has historically been fairly cyclical, TSMC’s dominant position means it is less cyclical than peers and boasts an attractive return on capital of more than 25%. The very high capital intensity of the sector (the newest factories cost in excess of US$20 billion to build) means that the barrier to entry is enormous, making it difficult for new players to compete. TSMC also has a strong culture of innovation and is able to attract some of the best talent in Taiwan and globally.
A GOOD TIME IN THE CYCLE
The semiconductor industry as a whole has endured a tough few years after significant chip shortages during the Covid-19 pandemic eventually led to oversupply as too many new foundries were built. We’ve seen early signs of a recovery, with demand recovering and inventory days starting to decline. We therefore expect that TSMC’s average factory utilisation, which is currently around 75%, will start to approach long-term averages of more than 90% within the next few years.
There are also certain structural drivers that excite us. Investment in AI data centres is highly topical and TSMC will certainly benefit from this. More importantly, however, being the world’s biggest factory, TSMC will most likely benefit from whoever makes or needs AI chips. One new driver is likely to be ‘on-device (or edge) AI’, which will drive a new replacement cycle for PCs and smartphones carrying larger chips. We estimate that revenue from AI is expected to be about 10% of revenue in 2024 but will grow to more than 20% of revenue by 2027.
IS THE VALUATION ATTRACTIVE?
At ~20 times forward price earnings, we do think that the market is now valuing the business appropriately compared to when we first invested last year. However, in the context of the supportive operating momentum behind it (we see earnings growth of more than 20% per year over the next three years), we don’t think it is exorbitantly priced.
In addition, compared to other options within the AI space, TSMC does still screen favourably to us, and we therefore remain comfortable to maintain a stake in the business. At Sanlam Private Wealth, the price paid for an asset remains an important tenet of our investment philosophy and we will continually reassess our position in TSMC in light of our assessment of fair value.
BIGGEST RISKS
The biggest risk to TSMC’s business is geopolitical. Even though the company is expanding globally (notably in the US state of Arizona), more than 90% of its production is still situated in Taiwan. A Chinese invasion to reunite the ‘rebel island’ with the mainland could therefore render the business worthless. In addition, US sanctions restricting Chinese imports of high-end chips have spurred significant investment in developing the Chinese semiconductor industry.
While we think it’s unlikely that China will be able to compete with TSMC at the leading edge any time soon, that country may well start to dominate the lagging edge, causing some disruption to TSMC’s more mature factory utilisation levels. The flip side is that geopolitics have increasingly led to countries wanting to bring production home, with many nation states providing generous subsidies and tax breaks to the likes of TSMC to build their factories in the US, Germany and Japan. Just this week TSMC broke ground on its first factory in Europe – in Dresden in eastern Germany.
The final risk takes us back to Moore’s Law – while the industry has repeatedly exceeded his prediction, it is becoming increasingly challenging, and there are signs that the pace of improvement is slowing. If attempting to fit more transistors onto a circuit board simply becomes too costly, the industry may yet again become commoditised at some point in the future.
However, while it might be getting harder to keep up with Moore’s Law, TSMC is finding innovative ways to add performance to its chips (like 3D packaging and better energy efficiency), and we consider it unlikely that this remarkable company will lose its competitive advantage any time soon.
