The average modern car holds around 1,500 semiconductor chips – and an electric vehicle twice that. As the world goes electric, demand will soar. And this just as we are rolling out chip-hungry 5G, which makes possible the Internet of Things (IoT), where devices talk to each, collecting and sharing data to make transport, manufacturing and our lives in general more efficient and safer. If that was not enough, you are no doubt screaming at your screen: “Don’t forget AI!”
McKinsey estimates that the semiconductor industry will grow 6-8% a year till 2030, when annual revenues should hit $1trn. Around 70% of this growth will be spurred by just three industries: automotive, wireless and computation, and data storage.
Looking at AI more specifically, every 15 years sees a tectonic shift in computing. In the '60s we jumped from mainframe computers to minicomputers. The '80s brought the transition to PCs. The '90s saw the arrival of smartphones and the cloud. Today we are moving to a world of parallel processing and the graphics processing unit (GPU).
It is easy to get technical – but not necessary. In short, most of the chips we use today are central processing units (CPUs), usually consisting of four to eight cores that process and execute a series of tasks. They can be incredibly quick, but AI is throwing at them more instructions than ever.
GPUs were first developed for the high-speed, processing-heavy world of gaming and video editing. GPUs can have hundreds of cores, all working in parallel, completing several tasks at the same time and therefore faster – much faster.
That’s it – as much as you probably need to know on the technical side. Some devices will need the very fast GPUs, but the traditional CPU chips will still do for lots of tasks. And demand for them will continue to rise. This is called Jevon’s Paradox – efficiency improvements driving cheaper chips lead to greater use.
In short, then, the world wants chips of every kind with everything.
Perils and opportunities for investors
Semiconductor manufacturing has historically been a boom-and-bust industry. It is difficult to forecast demand, so we swing between having gluts and shortages of chips. Get in at the wrong time and you can be punished in the short term. For evidence of that, look at the Nvidia share price in 2022 – it halved. But zoom out to the long-term trend and you see chip production going in just one direction – upwards.
Consider geopolitics, too. Three decades ago the US manufactured 37% of the world’s chips. Today the White House estimates it is just 12% – a similar level to Europe. And yet the US requires 37% of global output. Most manufacturing today is in East Asia (China requires 29% of the semiconductors produced in the world each year).
President Biden’s CHIPS Act is designed to bring production onshore. It is estimated that c.$223bn of US-based semiconductor projects are under way or planned, helped by $54bn in grants.
Semiconductor fabrication plants are expensive to build. They also take time. From the decision to invest to constructing the semiconductor clean room, installing equipment and ramping up production takes even the most efficient manufacturers more than three years.
And this assumes they can get the staff. The US is expected to face a shortfall of 300,000 engineers and 90,000 skilled technicians across all industries over the next decade.
How to invest?
So to the question of how to invest in this sector: the short answer is ‘carefully’. This is a big secular growth theme. Historically, we have included it under ‘automation’ as one of the themes driving our global portfolio. Today it stands alone.
But how do you capture the opportunities intelligently? Looking for the strongest-positioned companies across the value chain is key. Chip designer Nvidia is a small holding in the fund currently. It recognised the market potential of parallel processing a decade ago and invested well before its peers.
It introduced the GPU to PCs in the late '90s and has used the expertise from this approach to augment the units into datacentres and more recently to facilitate AI. Its clear leadership in accelerated computing means Nvidia is capturing a greater share of the ~$250bn a year capital expenditure on data centres while also dominating other fields.
As a result, its share price has risen 240% from the absolute lows in October of last year (it was down over 60% year-on-year to that point). Even then, because earnings have gone up so fast, the stock has derated and on consensus estimates trades on a similar multiple to the depths of last year – and at a discount to its 10-year average.
Then there are the producers – just a handful, thanks to the high capital costs, extreme technological requirements and years of consolidation. We own TSMC, but also Intel and Samsung as a hedge, given the geopolitical risks of TSMC, whose main production centre is, of course, Taiwan. It is good to see TSMC building manufacturing capacity in the US and elsewhere.
We also invest in the manufacturing equipment makers – a critical part of the value chain, enabling production innovation and the constant improvement in function. Here we own ASML, which dominates the cutting-edge lithography systems that enable the most advanced chip production, as well as LAM Research, which is strong in etch and deposition, another key part of the manufacturing process.
Identifying good companies is only part of the challenge. As indicated earlier, because of the cyclical nature of this sector, you must expect volatility. While there is a clear and strong long-term trend, the cycle can provide excellent opportunities to invest.
Buy-and-hold is a nice idea, but in semiconductors it can pay to be active around long-held positions. Volatility can be your friend if you can play it to your advantage, selling high, buying low shares in companies that are critically backed by strong market positions, with a lock hold on technologies, and delivering long-term growth.
To maximise opportunities like this requires constant research and market monitoring. But play it well and the “chips with everything” theme can contribute strongly to performance within a balanced portfolio.
Alex Stanic is lead manager of the Artemis Global Select fund. The views expressed above should not be taken as investment advice.