The future of society and technology requires ever-increasing connectivity, data speeds, and computational power, and those improvements depend on ever more powerful, efficient, and compact semiconductors. That is the basis for the US technology giants’ lofty valuations. Orbis, our offshore investment partner, is keen to tap into this need – but without overpaying. Alec Cutler discusses where Orbis is finding opportunity – at the right price.
At Orbis, we’ve written a lot in recent quarters about the large and growing market popularity and valuation gaps between companies perceived as higher quality and faster growing and those generalised as cyclicals and value. This runs the risk of giving the impression that we don’t like fast growth or higher quality attributes. We love investing in companies with these attributes. We just don’t like paying too much for the privilege.
There is a lot to like about the US tech giants, for instance. They are dominant businesses with deep moats, high returns on capital, piles of cash, and appealing long-term growth potential. But in all that, they are not alone. By casting a broader net, we can find similarly excellent businesses trading at much more attractive valuations.
The best example in the Orbis Global Balanced portfolio is also its biggest equity holding – Taiwan Semiconductor Manufacturing Company (TSMC). Like the US tech leaders, TSMC is a dominant business with a deep moat, high returns on capital, a pile of cash, and appealing long-term growth potential. Unlike the US goliaths, it trades just a touch above 20 times forward earnings, with a healthy dividend yield to boot.
Semiconductorification is our future, and with Taiwan Semiconductor Manufacturing Company and Samsung, we can invest in that future
We wrote about TSMC in September 2018. It is the world’s dominant manufacturer of logic semiconductors – the brains of a computer. As a foundry, TSMC makes chips designed by others, including chip designers like Nvidia, AMD and Qualcomm, device companies like Apple, and data-crunching giants like Amazon and Google. As those businesses grow, so does their demand for chips. And not just any chips.
Understanding chip technology
In small applications like phones and high-performance applications like artificial intelligence (AI), customers need the best chips possible. In semiconductors, that means chips that are built using the leading-edge manufacturing process. Generations of chip-manufacturing technology are referred to as “nodes”, measured in nanometres, referring to the lines of circuitry used to build the semiconductor. Leading-edge processes – smaller nodes – can produce chips that are faster, smaller, and more power-efficient.
Overall demand growth for chips has been robust for many years
Making semiconductors is like a contest to see who can draw the most lines on a sheet of paper. If I have a fat-tipped marker and you have a fine-tipped pen, you will win every time. Being on the leading edge is like having that fine-tipped pen. TSMC’s is five nanometres thick – about 20 atoms. Leading-edge chip manufacturing isn’t rocket science; it’s harder.
A single leading-edge fabrication plant can cost US$15bn, but money alone doesn’t do any good if you can’t get on the order book for the key equipment. And even if you can pay for a factory and secure the equipment to put in it, it still doesn’t do any good without the technical expertise to make it work. A badly run factory with the best equipment in the world will mainly produce trash in the shape of thin silicon disks.
Over time, competitors have struggled to keep up with TSMC. Ten generations ago, there were 28 companies with at least one leading-edge logic factory. Today there are only three: TSMC, Samsung, and Intel – though Intel is falling behind and seems destined to become a happy customer of TSMC and Samsung. Compare this decline in competitors to the evolution of other technology businesses, like streaming video. In 2007, Netflix had the streaming video market all to itself. Today it faces eight major competitors, including Amazon and Apple. In 2006, Amazon had cloud computing all to itself. Now it too counts two of its megacap peers as competitors. Rising competitive intensity is anathema to pricing power and shareholder returns, but falling competitive intensity, as TSMC is seeing, can be incredibly rewarding for shareholders.
Despite this attractive set-up, TSMC trades at a steep discount to both Apple, its largest customer, and ASML, its key equipment supplier. Those companies trade at 30 times forward earnings – a nearly 40% premium to TSMC. Their valuations are arguably fair, but TSMC’s discount seems unwarranted. The three firms have delivered similar levels of long-term growth, TSMC is as dominant in its field as Apple and ASML are in theirs, and TSMC’s products are more pervasive and essential to the global economy. We believe TSMC can continue to grow its earnings around 15% per annum while maintaining its very high returns on equity.
Taking a look at Samsung
The comparison to Apple is also apt for another top holding in the Orbis Global Balanced Fund – Samsung Electronics. Best known for its phones and TVs, most of Samsung’s cross-cycle profits come from semiconductors. It is the world’s market share and technology leader in memory chips – the short-term (DRAM) and long-term (NAND or flash) memory of a computer. It even supplies these chips to Apple for the iPhone.
Unlike logic chips, where Nvidia and AMD differentiate themselves through their chip designs, memory is perceived as a cyclical commodity business. Until recently, it was. Essentially all the costs are fixed – those multibillion dollar factories – so producers tend to keep the factories running even if there is an oversupply of chips. Worse, the industry suffered from repeated waves of “strategic” new entrants who overbuilt capacity and crushed prices. Extraordinary profits in good years were offset by sharp losses in the down cycle.
That has changed. The memory industry has consolidated, with just three players, led by Samsung, dominating the DRAM market. Post-consolidation, each of the three has been rational about adding capacity, which should lead to structurally higher and less cyclical margins in the memory business. As the leader, Samsung is best poised to benefit.
Yet Samsung, which has generated double-digit earnings growth and a 17% return on equity over the long term, today trades in the value stock realm – 11 times earnings and 1.5 times book value, even without adjusting for the US$80bn of net cash on its balance sheet. Memory is not the crummy old commodity it used to be, and Samsung is not the crummy old memory company it’s being valued as. In part, that is because Samsung’s vertical and horizontal integration gives it a unique edge. Making both devices and components guarantees a supply of the best chips and displays for its device businesses, while guaranteeing an anchor customer to help maintain high capacity utilisation in its components’ businesses. And as the only company globally with both a leading-edge memory business and a leading-edge logic foundry business, Samsung’s overall chip production volumes are significantly higher, giving it an edge over its competitors. Costs borne and lessons learned when adopting a new generation of logic technology often spill over to benefit memory, giving Samsung a head start in the race to perfect and ramp up the next generation of memory-manufacturing technology.
Today, Samsung’s foundry business is small compared to its memory unit, but this could change. As the only leading-edge competitor to TSMC in the globally important foundry industry, many customers – and even governments – have a stake in seeing Samsung succeed. Though TSMC wisely supports its customers’ growth rather than using its dominance to crank up margins, customers absolutely do not want TSMC to become a monopoly. If, through internal prioritisation and customer demand, Samsung’s foundry unit grows faster than the rest of the business, this could improve the company’s returns on capital over the long term, while reducing its cyclicality. Higher returns and lower cyclicality are a recipe for a higher valuation.
Understanding the risks
Of course, owning TSMC and Samsung shares comes with risks. Both are susceptible to global recessions, can be hit over short periods by customer supply chain issues, and as critical suppliers to the global economy, both are subject to geopolitical risk. TSMC is noteworthy in this regard. As the name implies, the vast majority of its research and production resides in Taiwan, making it susceptible to the tail risk that increasing US-Chinese tensions lead to issues on the island. TSMC’s high importance to all and its concerted and long-lived efforts to be fair to all mitigate this risk somewhat. We have further mitigated this risk by hedging the Taiwan dollar.
While TSMC and Samsung have risks to the downside, the world could also change in ways that would create further upside for both companies.
We have acknowledged one already – the growth of high-performance computing such as cloud and AI services. Four years ago, the overwhelming bulk of TSMC’s revenues came from chips for smartphones, which made for fairly lumpy demand, driven by industry technology leaps like 4G and 5G. Owing to the rapid growth in semiconductor-intensive cloud computing and AI applications, chip demand from high-performance computing should grow to match that of smartphones within a few years, making the logic foundry business less dependent on the smartphone product cycle.
The second source of upside comes from 5G wireless broadband technology, which will enable the long-discussed “internet of things” to emerge in earnest. The importance of leading-edge chips is already accelerating in things that have been around a long time (servers, cars, refrigerators, watches) and services that have been around a long time (weather forecasting, healthcare, education, transportation). When wireless data connections are faster than some fibre-optic connections are today, all sorts of applications that are currently impossible will become widespread. Overall demand growth for chips has been robust for many years, and should remain so, but we believe demand from “internet of things” applications could grow even faster – perhaps at 30% per annum for logic chips, with a similar tailwind for memory.
Semiconductorification is our future, and with TSMC and Samsung, we can invest in that future without having to overpay for the privilege.