How Intel Comes to My Radar?
Following the news of TSM, the largest holding in my portfolio, I saw several articles on Intel announces delay in its 7nm CPUs, and AMD CPUs outperform Intel ones benefiting from the use of TSM’s most advance process node technology (5nm). After the announcement, Intel stock price drops more than 29%. [Note: process node between companies are not directly comparable since it is measured differently. In terms of the density of logic gates, Intel 7nm is similar to TSM 5nm. In short, Intel is one generation behind the most advance process node technology.]
The drop interests me to find out more about the company Intel and whether there exists any investing value.
|Sales per sh||16.78|
|Cash Flow per sh||7.24|
|Earnings per sh||6.5|
|Div’ds Decl’d per sh||1.26|
|Cap’l Spending per sh||3.78|
|Book Value per sh||18.1|
|Avg Ann’l P/E Ratio||10.5|
|Avg Annl’ Div’d Yiled||2.4%|
|Net Profit (Mil)||21853|
|Net Profit Margin||30.4%|
|Working Cap’l (Mil)||8929|
|Long-Term Debt (Mil)||25308|
|Shr. Equity (Mil)||77659|
|Return on Total Cap’l||21.4%|
|Return on Shr. Equity||28.1%|
The first few things I will look at are Return on Total Cap’l and Return on Shr. Equity. They measure how well the company is helping its shareholders to generate a return. Generally, I look for around 20% ROA and ROE. How to know the appropriate level? Why not 80%? Choosing the number is based on the general average (e.g. S&P500 average ROA, ROE) and industry average (e.g. Semiconductor average ROA, ROE). ROE is usually higher than ROA since the company uses leverage. If ROE is significantly higher than ROA, it means the company is using a lot of leverage. Then we may need to look at the leverage ratio as well. But, in Intel’s case, the leverage seems to be heavy.
Then, the valuation is also important. A good company will not worth an infinite price. First, free cash flow (operating cash flow minus capital expenditure) is a parameter I will look at. Discounting the free cash flow by 10%, $34.6 is the current valuation of Intel based on the assumptions that cash flow will stay flat. However, based on the historical data, cash flow grows by 11.5% annually in the past 5 years and 13.5% in the past 10 years. I would say Intel worths more than $34.6. If we assume a 3% growth rate, it worths $49.4. Another valuation metrics I will look at is PE ratio which currently at 10.5 while the broad market index, S&P500, is 22.22. To conclude, I think Intel is cheap.
Having passed the ROE, ROA and valuation test, I can spend more time to dig deeper. For companies do not pass the above test, I will just save my time.
Although historical performance may not completely represent the future performance, I still think it has reference value. If historically the company performs well, it is more likely that it will perform well in the future.
Key metrics like sales, cash flow, earnings, dividends, book value all show growth (compounded annually) in the past 10 years and 5 years. Comparing with it’s competitor AMD, AMD just started to make profit in the recent 2-3 years. So, AMD is not a good comparison. Maybe TSM is more comparable, although TSM is in the fabrication business and Intel is in the Integrated Circuit design business.
Although not as good as TSM, I would say Intel’s performance is at least comparable. And with the current price, it seems more attractive than TSM.
Furthermore, ROE and ROA are improving year on year. Net profit margin remains stable across multiple years, indicating that the company can maintain it’s power over its suppliers and customers.
Current Position and Leverage
The current position is in good shape. The working capital ratio is around 1.4. Leverage ratio is 51% which is not the most conservative among its peers. But given that the interest coverage is 52.2x, the financial position of Intel is heavy.
The following section is more technical. I have included hyperlinks for key concepts. Please have a look at the articles if needed. And I am not an expert in this field.
The above analysis and numbers are all historical. Ultimately, what’s matter is the future earnings, and cash flows. Unfortunately, no one has a crystal ball into the future. What I am trying to do is to read a lot of information on the company and try to come up with a picture on how the competition and industry will be in the next 5-10 years.
On Moore Law
Moore’s law is the observation that the number of transistors in a dense integrated circuit (IC) doubles about every two years. However, after 40 years of development, people start to talk about the technology limit to make the transistor even smaller. The most advance process node in mass production as of today is 5nm. Only TSM and Samsung have this capability. TSM announces 3nm will be ready in 2022. Both Samsung and TSM are working on 2nm process node using GAA MOSFET (basically a way to increase the surface area of the gate. 5nm and 3nm are using FinFET MOSFET). To manufacture the MOSFET used in the integrated circuit, Extreme Ultraviolet Lithography is used. In short, the shorter the wavelength, the smaller the MOSFET you can manufacture. Making even smaller MOSFET becomes exponentially more difficult. A certain kind of new lithography machine is needed. But, that may not happen within 5 years since it may involve significant change by using even smaller wavelength light, e.g. x-ray.
How to improve the MOSFET density further so that the performance of IC can improve? It is believed that the key is on the packaging. Currently, the IC is laid on a 2D surface. What if we can stack them on top of each other? The MOSFET density will increase significantly, resulting in performance improvement.
If we believe this is the development trend in the next 5 years, the question is does Intel equip with advance packaging ability? From my research, currently, Intel has EMIB (Embedded Multi-die Interconnect Bridge), Foveros, Co-EMIB, ODI and MDIO packaging technology. Although I am not a professional in the the semiconductor industry, by reading the articles online, I believe that Intel is on-par with the advance packaging technology (EMIB, Co-EMIB and Foveros are used in current higher-end products).
On Intel V.S. AMD
AMD recent CPUs are gaining market share. AMD benefits from TSM most advance process node to improve its products performance and power efficiency. In 2018, AMD started shifting the production of their CPUs and GPUs to TSMC, following GlobalFoundries’ announcement that they were halting development of their 7 nm process.
Definitely, Intel needs to catch up with the process node technology. It is unknown that whether they can do so. But, for a company as large as Intel, I am wondering if they can also outsource the fabrication to TSM, AMD’s advantage over Intel on process node technology will be gone. CPUs design is not just about process node. It is believed that Intel design capability is better than AMD (that’s why AMD is lacking behind when it had similar process node with Intel in the past). Historically, Intel is very proud of being an integrated company, having their own fabrication facilities and being a leader in the place for many years. I think they will keep this model until they really cannot.
AMD will continue to gain market share over all segments of CPUs, including PC and data center. For PC, AMD will gain market share faster than data center market since PC users are easier to switch. Having said that Intel as of today is still the market leader in terms of market share in PC and data center, especially data center.
Apart from that, Intel portfolio is much wider than AMD which includes products in data center, AI, FPGA, 5G application, IoT, memory, GPU.
On x86 V.S. ARM/RISC-V
The main difference between x86 and ARM, RISC-V is that x86 uses CISC (Complex Instruction Set Computer) and both ARM and RISC-V uses (Reduced Instruction Set Computer). I will first research on the strengths and weaknesses of CISC v.s. RISC.
CISC approach attempts to minimize the number of instructions per program, sacrificing the number of cycles per instruction. While RISC does the opposite, reducing the cycles per instruction at the cost of the number of instructions per program.
|CISC Advantages/ Disadvantages||CISC Future||RISC Advantages/ Disadvantages||RISC Future|
|little RAM is required to store instructions||RAM becomes very cheap and more cache and RAM can be put into CPU||More RAM is required||RAM becomes cheap and will not be a problem|
|Hardware centric||Efficiency is optimized in hardware level||Software centric||Need to write more efficient software and compilers|
|Software compatibility is good||Remain the same||Software compatibility is not good||Will be much better since Apple will switch to ARM.|
|Cost a lot to design, but are fairly cheap to manufacture||Cost a lot less to design, but per-chip manufacturing cost is relatively high since we need more Cache in the chip|
|High power consumption||May not relate to the Instruction Set||Low power consumption||May not relate to the Instruction Set|
After looking at many articles, it seems that both CISC and RISC have their strengths and weaknesses. In terms of performance, it seems to depend on the microarchitecture more than ISA. In terms of power consumption, we seem to have an agreement that x86 implementations consume significantly more power than ARM implementations, but this may be a design choice to focus more on performance than power. In short, both CISC and RISC can be high performance and low power consumption. It is mainly a design choice. But, the trade-off between power and performance still exists.
I think RISC (ARM) will threaten the market share of x86 (Intel) in both PC and Data Center market since software compatibility in RISC started to improve. Similarly, x86 may also threaten the market share of RISC in the mobile market if Intel designs low power consumption chips. Overall, I think the market demand is expanding, but not a zero-sum game. ARM/RISV-V competition can be good for Intel.
On AI, IoT, 5G, Cloud Computing
Intel is one of the leader in Field-programmable gate array (FPGA) (Intel acquired Altera). You may hear of GPU in AI computing. However, FPGA offers better performance in AI computing. Basically, the logic gates are programmable to tailor-made for particular AI models, for example, Microsoft’s so-termed “Project Catapult” uses FPGA to accelerate computation. I expect that more and more FPGA will be used in AI computation and onboard AI Chips, like a self-driving car.
FPGA is also heavily used in 5G. For example, the 5G networking equipment of Huawei uses a lot of FPGA. Some rumours say that since the US banned Huawei access to Xilinx FPGA chips, Huawei needs to find Taiwan companies to make ASIC chips instead, a more expensive option. From this example, we may say FPGA will be largely used in the 5G networking equipment. The growth in 5G will also benefit Intel.
For cloud computing, x86 will still be the mainstream although ARM-based CPUs may also gain market share. In the cloud computing segment, it is believed that Intel will still be the leader for at least 5 years.
The main reasons I purchase Intel are:
- Good financial performance in the past
- The shares are priced at a reasonable level which may give a good return
- The demand for chips is increasing in data centre, AI, IoT, 5G, cloud computing segment. It is expected that Intel will get its share on the market.
- Competition from AMD, ARM, RISV-V is real. But, based on my judgment, they will coexist in the market. No one will completely dominate in the next 5 years.
- ARM and RISC-V development. How likely it will replace or threaten x86 in data centre segment.
- Intel revenue-generating ability in FPGA chips
- Intel revenue-generating ability in AI, 5G and Cloud Computing segment
- Intel advance packaging capability
- Intel process node development, i.e. Intel’s own ability and whether it will outsource to TSM/Samsung