Where new school techies and old school dinosaur investors clash heads is on historical versus future success. No story is better represented by that than the still-smoldering battlefield between armchair investors about Intel (INTC) and AMD (AMD). I don’t need to educate most anyone about the scathing response by Wall Street to Intel’s production delay into 7nm. We don’t need a dissection of the stock chart for AMD to comprehend that they are currently priced to rob Intel of massive market share in processors.
I won’t take a stand on that hairy debate, or on other similar discussions that really illustrate the growing divide between “growth” and “value”, though the same nuts and bolts concept I provide with this analysis on TSMC’s (TSM) recent results can be applied to other hardware debates (and maybe one I can elaborate on if there’s enough interest):
- Traditional servers IBM (IBM), Hewlett Packard Enterprise (HPE) vs enterprise cloud Amazon (AMZN), Microsoft (MSFT), Google (NASDAQ:GOOG) (NASDAQ:GOOGL)
- Traditional battle in networking: Cisco (CSCO) vs Arista Networks (ANET)
- Or even infrastructure software (to a lesser extent): Oracle (ORCL) vs Adobe (ADBE), Salesforce.com (CRM), others
I’ve observed an “all or nothing” mentality between the two camps – old vs new, in many of these types of discussions. It’s either that an older technology company won’t ever recover from a technological advantage now lost, or that established tech trades at such better (cheaper) valuations and that new tech could never possibly grow into their sky-high valuations today.
It could really be something more in the middle, and going to the leader of technology hardware, TSMC, is a perfect place to illustrate this. As the dominant foundry in the world for manufacturing the semiconductors that are essential to daily life in both old and new tech, they are on the cutting edge of other semiconductor hardware and building out capacity to not just continue in 7nm, but also what they are marketing as their next future nodes:
- N7+: 7nm FinFET Plus –> “the first commercially available EUV-enabled foundry process technology in the world”
- N6: 6nm FinFET –> also uses EUV lithography, design rules compatible with N7
- N5: 5nm FinFET –> will also have EUV adoption
- N5P: 5nm FinFET Plus
- N3: 3nm –> will be “the foundry industry’s best PPA technology”
- Of course, 3D chip stacking solutions
Note that EUV is another type of lithography, essentially a type of semiconductor manufacturing (like a printer) which is of extreme relevance to WFE (Semiconductor Wafer Fab Equipment) companies like KLA Corporation (KLAC), ASML Holding (ASML), Lam Research (LRCX), Applied Materials (AMAT), etc. But more on that maybe another time.
Because TSM is, of course, headquartered in Taiwan, they are not subject to the same SEC documentation requirements such as the 10-Q. However, they have been gracious to provide investors with presentations with each of their earnings calls, and I’d like to show how the trends can be misleading and that the transition to 7nm (they call N7) hasn’t been as catastrophic to older technology nodes (like 28nm) as what seems to always be predicted.
The specific production numbers behind each technology node does not appear to be disclosed by the company in their annual reports, as they usually report revenue numbers by geographic location and reporting segment instead. Even the presentation slides where they do disclose this information can be a bit misleading without some context, which is why it needs to be presented here.
From first look, it appears that the narrative of the forceful push to new technologies like 7nm and away from old technologies like 28nm is in fact playing out. For example, since 3Q19, revenues from the 7nm process increased from 27% to 36% of total company revenues, while 28nm saw their portion of revenues decrease from 16% to 14% over the same time period, with other legacy nodes also experiencing expected declines – 16/20nm, 65nm, 0.15/0.18um (which would be 150/180 nm).
But a simple calculation on these percentages tells a much different tale:
Instead, what we see in legacy technology nodes since 3Q19 is a generally flat revenue mix, while, at the same time, 7nm revenues have fueled most of the growth in total company revenue.
That indicates that shipments are likely staying the same for older “dinosaur” tech, which could indicate that the strong cash flows that many of the old “dinosaur” value tech names have might not be going away as soon as people expected. As long as the demand is there, and the fabs have capacity and are running efficiently and fully, those cash flows become cash cows for chipmakers who still produce at much older nodes.
Moore’s law says that density doubles, but it doesn’t say that revenue potential must subsequently half. Yes, you see revenue degradation and margin compression over time if you’re only delivering in legacy technology nodes, but it doesn’t work nearly as fast as Moore’s law itself.
What’s Really Driving TSMC’s Revenue Boosts Lately
What becomes obvious after digesting this information is that, although new technology can be so disruptive, so fast and life-changing (and industry-destroying), it’s also often restricted to physical limitations like many other businesses – as I explained in my article about cloud infrastructure: How the Cloud in 2020 Changes Everything about Balance Sheet Accounting.
In the case of semiconductor manufacturing, you have a demand surge (cloud server rollouts, 5G, AI, IoT) at the same time that technology is being developed. While every component maker would love to upgrade to the latest and greatest immediately, the reality is much different. Fabs like TSMC only have so much physical capacity possible per year, and the move down to advanced nodes only creates more needs for expensive equipment, iterative design improvements – think fabless designers and their increased hiring needs as test silicon comes back: Broadcom (AVGO), Qualcomm (QCOM), even Apple (AAPL) nowadays. Manufacturing, testing, and verification problems don’t just go away as the technology improves; in fact, they multiply.
So, of course, revenue on advanced process nodes like N7 and N7+ will be necessarily higher because the actual manufacturing process is that much more expensive. As this trend continues, the revenue numbers might detract further and further from capacity and shipment realities – making deeper analysis on the actual productions of chips in these advanced nodes that much more critical to understanding the real trends of consumer supply and demand in the semiconductor industry.
Which Industries Will Drive Production for TSMC Moving Forward?
Before we can speculate on that question, we must understand what is driving this revenue now, not from the capacity side but from the demand side. Not only is this applicable for the foundries TSMC, Samsung (OTC:SSNLF), Intel (INTC), ON Semiconductor (ON), the Chinese fabs, but it’s also useful information for semiconductor companies that are all up and down the semiconductor supply chain.
The basic context shows WHYDFML (what have you done for me lately), in regards to TSMC’s latest quarter (2Q20):
- Smartphone – 47%
- HPC – 33%
- IoT – 8%
- Automotive – 4%
- DCE – 5%
- Others – 3%
The 12% QoQ increase in HPC, or High Performance Computing, is attributable partly to the new video game console refresh that’s upcoming, partly to the sales influx for new computers and tablets driven by the pandemic Work From Home and virtual schooling themes, and other factors like cloud datacenter buildouts also affected from the lockdowns.
Things like the drop in automotive (-13%) and IoT (-5%) are largely expected to rebound as the economy opens back up not just in the United States but globally, especially as manufacturing impacted by COVID-19 like the major automotives starts to fire on all cylinders.
Technology Node vs Consumer End Use
Now, it’s important to understand that the new technology nodes aren’t evenly distributed across all of these industries, and especially not all at the same time. In general, smartphones and HPC will see 7nm implementation much faster than a segment like IoT and automotive, where the improved density isn’t as applicable for the latter group as it is for the former.
After all, it’s not just density that’s important for semiconductor manufacturing, but you also have considerations like yield, performance (speed), power, and density/dimensions/size. The nature of chip design, application, and implementation simply forces different priorities depending on end use, consumer preferences, and many other factors (even supply chain and distribution, which gets so little attention in this sector).
Going back to TSMC, they’ve really outlined the application of their various technology nodes well. I’ll quote or summarize their strategy related to their major end use segments from their latest annual report, so that you know it’s not just my opinion but the opinion of management themselves.
- 5-7nm FinFET (and FinFET Plus)
- “Leading process technologies”
- “Mainstream product applications”
Segment: High Performance Computing
- 5-16nm FinFET
- Multiple advanced packaging technologies
Segment: Internet of Things
- ultra low-power (ULP) focus
- “support innovations for IoT and wearable applications”
- In general, the outlook for this industry is very bright (not just from TSMC but many other managements in related businesses):
- To support the ever-increasing demand in IoT edge computing and wireless connectivity, TSMC also offers “the most competitive and comprehensive leading-edge specialty technologies in RF, enhanced analog devices, embedded flash memory, emerging memory, sensors and display chips, as well as multiple advanced packaging technologies including leading InFO technology”
Segment: Automotive electronics
- 5-16nm FinFET
- “the industry leader in providing a robust automotive IP ecosystem”
- “broad and competitive specialty technologies”
Legacy Semiconductor Industry Takeaway
As you can see, legacy technology (above 16nm) is a big part of many of the leading expected growth industries in technology for the next several decades, and the latest revenue trends coming straight from the cutting edge leader TSMC seem to confirm that legacy technology still has very strong demand.
Though flat-line revenues in legacy nodes is not something most investors get excited about, the TSMC strategy doesn’t encompass the whole demand picture either.
I find it interesting that TSMC has not appeared to see it prudent to cut legacy production over the last 8 quarters, which has to indicate strong utilization rates for that capacity. And even if TSMC does decide to cut capacity, there are so many other fabs out there who would be willing to pick up those needs even if TSMC ceases to make that their focus – which would be a bullish signal for many chipmakers who don’t happen to be as cutting edge as AMD, for example.
There’s no doubt about it. TSMC is the #1 player in the industry, and I’m sorry, Samsung isn’t even close.
However, how TSMC and the foundry industry perform as a whole will be incumbent on the demand from the industries I highlighted above – smartphones, high performance computing, IoT, and automotive electronics – and if that demand doesn’t meet expectations, then the valuation paid for TSM could determine how an investor’s performance fares, as disappointments can cause corrections down to more “normal” fundamentals.
To understand whether investors are likely getting a good deal or not with the stock today, I’ll be performing a reverse DCF to estimate how much growth that investors are implying in their models based on how the stock is currently trading compared to free cash flows (and EPS). I’ll be using a 6% discount rate, which is aggressive but reasonable with the company’s relatively low WACC (weighted average cost of capital) and the ultra low interest rate environment we have today.
FCF estimate = $1.61 per share
I made this estimate based on FY 2017 FCF numbers (cash from operations minus capital expenditures) to try and account for the large capacity upgrades the company has recently made for the 7nm ramp-up that have greatly inflated recent capex in 2018 and 2019.
Implied growth rate = 19.10%
This seems very high and tends to reflect on the capital intensive nature of TSMC’s business, with the physical limitations of having to make massive outlays in order to increase production capacity and revenues and profits. It’s less of an indication of an extreme overvaluation from investors, per se, but much more of an indication of investors who are willing to ignore low free cash flows in favor of higher EPS and top and bottom line growth.
- TTM EPS (without NRI) = $2.90
- Implied growth rate = 10.95%
This is obviously a much more reasonable valuation and, again, simply signals investor priorities for the longs. Less about cash flows and more about earnings (and/or the company’s dominant market share and tech leadership).
All in all, TSMC has a rock-solid moat in many ways – obvious technological lead, superior scale, and capacity, a rock-solid reputation and some of the biggest customers in the world.
A deeper dive into the revenue splits between the company’s various technology nodes shows that, while their technology leadership is providing massive revenue growth, the legacy manufacturing is still also providing steady revenues at the same time. There doesn’t appear to be much cannibalization between these technologies yet, which appears to be a great sign of strong demand.
How the company performs over the next several years is largely dependent on the overall economy and the relative performance of growth versus expectations in the company’s primary business segments. The current valuation doesn’t appear to be heavily tilted to the undervalued or overvalued camp at this time.
***If you’d like more deep dives into company filings like this one, please do consider Following my profile with email alerts. To see how my picks for the Real Money Portfolio have done, check out: Post-Pandemic Performance***
Disclosure: I am/we are long CSCO, MSFT. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.