Bull Of The Day: Taiwan Semi
It's never too frequent to talk about Taiwan Semiconductor (TSM), the preeminent "lego builder" for the AI revolution.
My colleague Ethan Feller just did a great analysis of the fundamentals two weeks ago and this week I wanted to go wide-angle on what their growth means for the global economy as they build a giant new foundry in Phoenix, Arizona.
Here was the video I made Monday morning where I showed the company growth, the huge buy opportunity I missed on August 5, and some details about the first US-based "fab" for the magicians of sub-10 nanometer chip technologies.
Why Is This Silicon Magician So Important?
Taiwan Semiconductor Manufacturing Corp (aka, TSMC) is a critical enabler of today’s technology industry, having fundamentally reshaped the semiconductor landscape in 1987. Through the vision and leadership of its founder Dr. Morris Chang, TSMC pioneered the pure-play semiconductor foundry business model.
Today, TSMC’s semiconductors are the digital engines inside more than 12,000 electronic products created by over 500 companies. TSMC’s technology can be found in devices such as internet-connected home appliances (my coffee maker), smart phones and PCs, EVs and autonomous vehicles, cloud data centers, airlines and space-stations and AI-powered applications. The world’s biggest brands and savviest startups in technology rely on TSMC to unleash their innovations to the world.
TSMC has since remained one of the world’s largest dedicated foundries with fabrication facilities (“fabs”) and subsidiaries spanning Asia, Europe and North America. Given persistent tensions between China and Taiwan, I'm sure it's not lost on my readers why bringing massive high-tech wafer fab production capacity to the US will secure the vital, global IP and technology that might otherwise be disrupted by trade wars, political wars, or even Southeast Asia "hot" wars.
Sultan of Silicon Builds Castles in US Desert
In the video linked above, I was trying to explore the TSMC website devoted to the new Arizona facility. But it was hard to navigate -- or maybe my internet connection speed was taking a nap.
But even with perfect web speed, I still get so intense about what is unfolding with NVIDIA (NVDA) GPUs and the "massively parallel architectures" of AI, that I get tongue-tied trying to cram it all in to a sub-10 minute video.
It should be simply unfathomable to any of us that TSM is building circuit boards and chipsets with less than 5 nanometers of space between the transistors -- when the coronavirus is measured as big as 50 nanometers.
The accelerated generations of chip designs in terms of increased transistor density are all about a higher degree of miniaturization.
This is what Jensen Huang of NVIDIA showed us was going to happen as gaming GPUs took over the heavy lifting from Intel desk-top CPUs. In essence, as early as 2016, he saw that what I now call "Huang's Law" was going to overtake Moore's Law.
Honey, I Shrunk the Chips!
You can see an excellent graphic of that transformation here.
But in full transparency, it's actually beyond my intellectual pay-grade to be talking about the precise engineering world of nanometers in this article. Still, my massive enthusiasm will not be bridled by my mere ignorance.
So my simplest analogy has always been this: picture CPU-driven "serial" processing as a 4-lane highway, or 8-lane expressway. Obviously, traffic jams will still always happen any day of the week, no matter where your metro.
Now, picture GPU-driven "parallel" processing as 100,000 piano players, all able to execute, process, and deliver new information and insights simultaneously. They could be a symphony, or an improv factory, with no traffic jams.
The Monolith of Phoenix
And given that my fascination is far greater than my electrical engineering knowledge, I'm always willing to listen to Sensei Huang and his key partners like TSMC.
So here I will share some key details from that website about the Sultan of Silicon's latest magic mega-complex on over 1100 acres...
In a historic announcement, in May 2020, TSMC shared its plans to invest $12B in Phoenix, Arizona – building an advanced semiconductor manufacturing fabrication. In December 2022, the company announced its commitment to build a second fab in Phoenix, increasing its total investment to $40B.
Then in April 2024, the U.S. Department of Commerce and TSMC Arizona announced up to US$6.6 billion in direct funding under the CHIPS and Science Act, fulfilling a goal to bring the most advanced chip manufacturing in the world to the United States. TSMC also announced plans to build a third fab at TSMC Arizona.
This third fab brings TSMC’s total U.S. investment to more than $65 billion, making this the largest foreign direct investment (FDI) in Arizona history, and the largest FDI in a greenfield project in U.S. history. Now with a third fab, TSMC Arizona will create approximately 6,000 jobs – and more than 20,000 accumulated unique construction jobs, as well as tens of thousands of indirect supplier jobs.
As one of TSMC’s advanced fabs, TSMC Arizona will play a vital role in the U.S. government’s goal to onshore semiconductor manufacturing and strengthen national economic competitiveness. TSMC Arizona’s first fab will operate it’s leading-edge semiconductor process technology (N4 process), starting production in the first half of 2025.
The second fab will utilize its leading edge N3 and N2 process technology and be operational in 2028. The recently announced third fab will manufacture chips using 2nm or even more advanced process technology, with production starting by the end of the decade. TSMC Arizona will be able to produce semiconductor wafers for its valued customers using the most advanced process capabilities in the country.
The Semi Cycle: Overdue for a Big Recession?
On Monday morning, when I decided I must profile TSM for the weekly Zacks Top Stock Picks video, two interesting items were already swirling through my brain:
1) I had just published my weekly Sunday newsletter on LinkedIn about the new Microsoft mega-datacenter in my backyard in southeast Wisconsin: Microsoft Stargate: "My god... it's full of GPUs!"
2) I had a dialog with a fellow investor-analyst who was wondering if the Semiconductor Cycle had peaked this year and was now due for a bigger correction/bear market.
If you've been watching since I first penned my Tech Super Cycle thesis in 2017, you know there are often many more dramatic bear markets in chip stocks than actual "fall-off-a-cliff" sales implosions.
Thoughts on the Semiconductor Cycle
So when a fellow investor-analyst asked me what I thought about a possible "classic peak" in the Semi Cycle, I quickly shared my thoughts to put the elements in perspective...
The SOX is ~140% off its 2022 bear market low.
The 17% correction in July was well within the bounds of expected volatility after such a run.
And the "panic of August" was delicious for long-term growth managers.
GPU demand isn't slowing. I profiled TSM for Top Picks today so I was just looking at rev growth for various players.
In wafer fab equipment AMAT is slowing but LRCX still looking at 16% topline into next year. ASML flat this year but projected 35% next year!
AI has reinvented the cycle. No hype. Real capex. Ever since MSFT GOOG META were lined in early 2023 to buy GH200 series, the other 997 corps in Fortune 1000 knew they had to build their own internal LLMs or go extinct. NVIDIA won't be able to make new GB200 series fast enough.
(end of my back-of-the-envelope bull case for chips and NVDA since 2022)
Minimum 30% CAGR, While NVDA Grows 50%+
Then, on Monday afternoon I was alerted to the latest research publication from Daniel Newman at Futurum Group, an independent technology research firm. Dan knows TSM and NVDA and their forward growth catalysts as well (probably better) than I do.
And he got his research report profiled on CNBC with this headline: "The AI data center chipset market is set to soar from $38B in 2024 to $138B by 2028, driven by a 30% CAGR."
I think he's being way too conservative, especially when NVIDIA already had a clear path to $1 trillion in topline (including software services) even before the earthquake of ChatGPT surfaced in early 2023.
Finally, here was my LinkedIn newsletter on Sunday profiling the virtues of Satya Nadella and Brad Smith, for Wisconsin and beyond...
Microsoft Stargate: "My god... it's full of GPUs!"
If that link doesn't work, just email me at KCook@Zacks.com
Bottom line: As I described in my video, be ready for more volatility this autumn and into the election. If stocks have another one-day "panic" like they did for the Bank of Japan carry trade, be ready to buy. Because the global-macro picture isn't going to shipwreck the megatrend of AI hyper-scalers, LLM-builders, and pure science-explorers who want to improve the world we live in.
More By This Author:
Bull Of The Day: Datadog - Wednesday, August 7Bear Of The Day: Crowdstrike
Bull Of The Day: Nvidia - Tuesday, July 23
Keith tracks to bring his readers some of the best investment recommendations on the markets today. These trends have been making fortunes for centuries, and they'll likely continue to do so ...
moreComments
![](/images/tinyMceCommentImg.png)