Note: This article is a premium article that first appeared on Patron. This was written in mid 2020, but the underlying industry structure remains very relevant today.
If you enjoy articles like this, do consider supporting Financial Horse and getting access to premium articles, my personal stock watch list, as well as my personal portfolio allocation.
We’ll talk a bit about China’s semiconductor progress, TSMC/AMD/MU, and address some great questions from you guys on investing in this industry.
China’s Semiconductor Progress
So I was talking to a guy who works in semiconductors in China. This is how he described the state of China’s progress right now:
- Design of Chips – 1 to 2 generations behind US, Huawei is an exception where Hisilicon is on par with (or close to) the best like Qualcomm
- Manufacturing (Fabs) – 2 to 3 generations behind, not able to catch up short term
- Assembly, Packaging and testing – half a generation behind US
- Applications (Software design) – on par with US
I thought this was a fantastic summary.
China can (1) design chips, they can (2) assemble, package and test chips, they can (3) build software to run on the chips.
What they cannot do, is to build the chips themselves.
For that, they need to rely on TSMC, which has now evolved into the premier chip maker in the whole world for reasons we discussed in Part I.
Why is it so hard for China to catch up?
I think the simple reason is that this isn’t a technological issue like learning how to code (where the Chinese picked up quickly). Making a semiconductor wafer builds on years and years of manufacturing know how and process control, all of which are highly specialised information that is not readily accessible.
Companies like TSMC have been doing it for close to 3 decades, and their cutting edge manufacturing techniques incorporate lessons from those 3 decades.
China by contrast, is a newcomer to this industry.
For new sectors like 5G, electric vehicles or AI where there is very little existing institutional knowledge, China was able to catch up quickly.
But in these industrial heavy sectors which rely on trade secrets and manufacturing know how accumulated over decades (airplane manufacturing is another area where China is lacking), China is just way behind the rest of the world.
It was not an issue pre-2018, but once the trade war came into play and semiconductor supply became an issue of national security, China immediately needed to start building up domestic capacity, and fast.
How long for China to catch up?
There are 2 ways China can catch up: (1) Develop the technologies and know-how themselves through trial and error, (2) Acquire the technologies from other companies via M&A or industrial espionage.
(1) is out of the question, because it will take too long. If China goes with (1), we’re probably looking at a 5 year runway before they catch up with leading global players, which is clearly not acceptable to CCP leadership. It will just give the US too much power over China.
That leaves (2). Which is either M&A (buying another company), or stealing it.
It’s interesting to note that right now there is no cutting edge semiconductor Fab in China, not even those operated by MNCs. The only one that comes close is Broadcom, but it’s closing down, and it was never cutting edge nodes to begin with.
But the point is – M&A is going to be tough with all the anti-China rhetoric around the world. China tried to acquire Micron a few years back, but the deal was blocked by the US even then. In today’s climate, it’s just incredibly hard to see China being able to acquire any notable semiconductor company. No country would want to see their crown jewel being stolen away like that.
That leaves stealing, or industrial espionage. There was a big spat a year or two ago where China was basically trying to steal the secrets from Micron engineers in Taiwan.
Despite what the world may say or think, finding a way to get their hands on this technology will drastically shortcut the time to market for them.
According to the industry expert I was speaking to, if this is possible, the timeline could be cut down to 3 years optimistically – so around 2022 or 2023 for local production.
That’s still a couple of years off, so semiconductor players can rest easy for now. But the time will come eventually, and it’s definitely one to watch out for. If China can produce semiconductors on the same scale they do for concrete or steel, it’s going to revolutionize the entire industry.
Thinking behind each stock
The thinking here was that TSMC is the only Fab in the world that can hit 7nm nodes. Because of this, it has pricing power, and enjoys the lion’s share of the profits in this industry.
I’m bullish on semiconductors going forward, it’s the backbone of the entire digital economy these days.
And TSMC is the only guy who can make the most advanced semiconductors, which is just a great place to be in. You can dictate terms, and the world has to follow.
Price matters though. It made a lot of sense in the 40s and 50s range, but at 80 right now, that’s over a $400b market cap, and is starting to look a bit stretched.
Sure, the price can continue to climb, but the risk-reward has become a lot less compelling than where it was in the 50s.
The thinking behind AMD was that just a couple weeks ago (note this post was written in Aug 2020), it was at a $60b market cap, vs Intel’s market cap of $250b, and Nvidia’s market cap of $270b.
AMD’s latest line of processors have proven incredibly competitive against intel, and at a lower price point. Their GPUs under the Big Navi line also look to give Nvidia a good run for their money. And the next generation of PS5 and Xbox coming out this year run on AMD chips.
For now, AMD is a small player in both processors and GPU (dominated by Intel and Nvidia respectively). AMD is 20% of the consumer processor industry, 5% of server processors, and about 20% of GPU.
But with how competitive their new lineups look, it was reasonable to think they would be able to steal some market share away from Intel and Nvidia.
So if AMD can get a 25% share of the processors and GPU industry, and hold those gains, that would give them approximately a size 30% of Intel + Nvidia’s market cap. That’s about $170b and way higher than the $60b it was at a few weeks ago.
And this is just a 25% share of the industry, it’s conceivable it can go even higher than that to the 30% – 40% range once performance starts matching Intel and Nvidia.
Of course, a lot of things have changed since, and AMD is now at a $100b market cap. Risk-reward is less compelling at this point, but if AMD can continue to execute and steal market share away from Intel/Nvidia, there could still be big gains here as the disparity in market cap shows.
In comparison to the other 2 players here, Micron doesn’t really have pricing power.
Micron produces memory, which is a commoditised product. Much like how an oil major in the oil industry cannot set prices, Micron is not able to dictate memory prices – it has to accept the market price. This creates big boom bust cycles, just like it does in commodity markets.
So historically, the trick to making money with Micron is to buy during the down cycle when memory is dirt cheap, and sell at the top of the boom cycle when memory is crazy expensive.
In recent years, the cycles have become less pronounced, as the 3 players in the DRAM market (Samsung, SK Hynix, Micron), seem to have been more careful about increasing production. Without big increases in production, we will also avoid the bust cycle. This could create a more stable market for memory prices, and bigger profits for all players.
Of course, this is just a theory, and it could be wrong.
But personally, I had a hunch that (1) neither of the 3 wanted another big downcycle so they controlled production increases, and (2) that we may have been past the trough of the previous down cycle which was a mild one.
I originally expected COVID19 to trigger a down cycle for memory because consumers would cut spending on digital goods like new smartphones. As it turns out, that was true, but the demand shortfall was more than made up for by increased demand from institutional players like data centers, and laptops for work from home.
Throw in continued data center demand, and the coming wave of next gen gaming consoles that pack in a ton of memory, and I’m cautiously optimistic on memory demand going forward.
It’s a long article, so I’ll round up by addressing some reader questions.
1) Apple switching to ARM – Impact on the industry
Apple recently announced they’re switching their Mac series over to ARM processors.
A bit of background – All modern PCs run on the x86 architecture, which is basically chips built by AMD and Intel. They have superior performance, but poorer energy efficiency.
ARM architecture by contrast, powers the entire mobile series. So all the chips in your mobile phone or tablet run on ARM, which are built by Qualcomm, Apple, Hisilicon etc (all manufactured by TSMC Fabs). In terms of raw power they don’t match up to the x86 series, but they are way more energy efficient, which is what you want for a mobile device.
It’s early days, so it’s hard to break down the full implications of this move.
Will Apple’s move cause the entire PC industry to shift over to ARM, once the benefits of ARM PCs becomes clear? If so, it’s going to be terrible news for Intel and AMD unless they can pivot quickly.
Or will Apple’s move turn out to be a flop, and fail to hold their ground against traditional x86 processors?
It’s genuinely too early to say which way this would go.
But it has massive implications for the future of the industry, so it’s something we definitely need to keep an eye on.
2) Whether the (roughly) equally weighted ETFs make sense over stock picking.
The point being that those companies are nearly impossible to “value” (or rather traditional valuation has little relevance) given the importance of innovation. And without being an insider, your chances of being ahead of the moves are fairly limited… So I think there is a case for equal weighted diversification (doesn’t solve all problems though).
Semicons are a very technical area, to the point where it has become like pharmaceutical investing. You need to be able to evaluate the attractiveness of a new product line, keep up to date with market news, and have a good idea of which products will or will not do well going forward.
If you can, there are big profits to be made.
But if you can’t it may be easier to just go with ETFs.
Now I get the point of equally weighted ETFs, but I’m not so sure if I agree with them at the moment.
The current market is very momentum driven, very winner takes all. The winners keep on winning, the losers keep on losing.
In this kind of market, I want to be owning the winners, the big guys, so an equal weighted ETF may not make so much sense.
At some point in the future, this paradigm may switch, at which time using equally weighted ETFs may make sense. But the market can remain irrational for extended periods, so I generally prefer not to be early to this kind of paradigm shifts.
Better to ride the trend until we see signs the trend is reversing, then get out quickly.
3) Most importantly: when is a good time to enter this market? How do I know if this market is over/under-valued at this moment?
That’s a tough question. Semicons are looking very hot right now (note this article is written in Aug 2020), but when you look at the 5 to 10 year runway we have ahead of us, semicons are just going to see unbelievable growth in demand.
The entire global infrastructure runs on these stuff.
In the 20th century you invest in stuff like concrete and steel, which are used to build highways, buildings etc.
In the 21st century, the highways and buildings are all digital in nature. The world runs on data centers, server farms, 5G mobiles etc, all of which run on silicon.
Personally, for the semicon industry, I would view a pullback as a buying opportunity.
Love to hear your thoughts!
Note: This article is a premium article that first appeared on Patron. If you enjoy articles like this, do consider supporting Financial Horse and getting access to premium articles, my personal stock watch list, as well as my personal portfolio allocation.