It is often said that buying stocks is based on expectations, but in reality, it is usually based on stories.
As the saying goes, “Buy the rumor, sell the news.” This means that one should buy when there is a story and sell when the news (the story’s conclusion) breaks. It is important to note that even if the story does come true, if there are no new stories, the stock will no longer rise.
The story of silicon carbide (SiC) is similar to quantum computing (old article: Lies and Truths of Quantum Computing): it has depth, barriers, prospects, and shock value. We all have a bright future.
So, are there any similarities between silicon carbide and the grand stories of graphene, nanotubes, and sodium batteries that were once popular with the public? Not really, as this material already has a practical scale of application. Today’s text may not make some people happy after reading it.
In the semiconductor industry, overtaking on a curve is a pipe dream, and all dreams must be achieved through investment of money and time.
In 1987, five students and their professor from a laboratory at North Carolina State University created Cree, as they had made a breakthrough in studying the long-crystal method of silicon carbide after years of research. Silicon carbide has the potential for disruptive innovation in LED lighting and power semiconductor.
Cree’s LED business quickly made progress. In the following 30 years, Cree has been viewed by investors as an LED company with a story of a silicon carbide chip as a bonus.
Dr. John Palmour applied for the core patent of silicon carbide MOSFET when the company was first established. However, after the 20-year protection period of the patent expired, Cree was still unable to launch any relevant products.
Two, Link to heading
At that time, the industry generally believed that although silicon carbide was naturally resistant to high pressure, its crystal defects were too numerous, making it almost impossible to produce reliable SiC MOSFET.
Especially in the field of power, reliability is the number one requirement for MOSFET. Other players in the silicon carbide semiconductor industry have shifted their focus to BJT, JFET, and SBD diodes.
Since its commercial use by Infineon in 2001, SiC SBD has gone through multiple accidents and gradually matured, resulting in low entry barriers and low profits. Combining it with solar and wind power amplifies the story of SiC SBD.
The main promoter of JFET is UnitedSiC, but with the maturity of SiC MOSFET, JFET seems no longer to be the most perfect solution. UnitedSiC was acquired by Qorvo, a RF giant in search of new directions, in 2021.
Cree walked alone on the most difficult path, facing a barrage of gunfire to attack the MOSFET from above, because only SiC MOSFET has the potential to revolutionize high-voltage power devices in the future.
Starting from 0.5-inch wafers (which should be called “grains”), Cree crystals gradually improved from being full of defects with a painfully low yield until 2011 when they barely managed to release them for commercial use, with only a 3-inch substrate. Meanwhile, silicon wafers of the same period were 12 inches and cost numerous times lower than silicon carbide. By numerous times, it means that SiC MOSFETs were completely uncompetitive at that time.
Cree perseveres to improve its long crystal technology, from 3 inches to 4 inches, then to 6 inches and 8 inches, with each step taking several years and overcoming countless technical challenges.
Nevertheless, there still exists a significant price disparity between SiC MOS and its main competitor IGBT, and its application is still niche. That is, until Musk rode in on his white horse.
Three, Link to heading
Cree’s LED business has gone through ups and downs, and by 2015, it was struggling. The company decided to cut its losses and sold its entire silicon carbide power business, Wolfspeed, to Infineon for only 850 million US dollars.
Two years later, due to the fact that silicon carbide was a material favored by the military, US regulatory authorities executed the acquisition, dramatically saving Cree.
Infineon is indisputably the leader in IGBT devices, and if Wolfspeed is also brought under its umbrella, it will be too strong in the power device field, which will indeed make the United States more wary.
After the failed acquisition, Cree’s GaN-on-SiC business experienced significant growth due to the massive leap in communication 4G/5G, prompting Infineon to outright give their RF business to Cree.
Even so, Cree is still suffering annual losses, making each day feel like a year.
Finally, we waited for Teacher Ma to arrive with the Model 3.
Four, Link to heading
Musk’s approach to car manufacturing is completely different from traditional car companies. All of his design concepts are disruptive, which allows him to boldly adopt various radical strategies, including replacing IGBT with silicon carbide MOSFETs (refer to the old article “Tesla’s Unique Character and Silicon Carbide Revolution”).
Although using expensive SiC MOSFET only extended the range by 5%, Tesla, which bears the sense of mission of defeating gasoline cars, sees the revolution of new technology. Looking at it the other way around, the 5% increase in range can reduce the required power lithium battery by 5%, which in turn saves a considerable amount of money. Tesla pursues the cost-effectiveness and high gross margin of each component to the extreme, and this is just one small example.
Without Tesla, the SiC MOSFET story would likely be delayed by 3-5 years, and it’s only a matter of time before Cree is acquired.
Due to Tesla’s rapid growth, its SiC MOSFET device supplier STMicroelectronics (ST) could not guarantee sufficient silicon carbide substrate materials, leading to their acquisition of Norstel.
ANSYS swiftly acquired GT Advanced, targeting the role of being the second supplier for Tesla. The supplier of silicon carbide substrate for the secondary line was quickly divided among several power giants, while Cree, which monopolizes about 50% of the market share, has grown rapidly.
Cree decided to sell their LED business and changed their name to Wolfspeed, focusing on silicon carbide. Today, their market value has reached a staggering $14 billion. Infineon is probably crying.
In 2009, ROHM of Japan purchased SiCrystal from its former German home, Siemens. Today, silicon carbide substrates account for approximately one-sixth of the global market.
Five, Link to heading
So, has Infineon already been completely defeated in the competition for silicon carbide? Not at all.
The story of silicon carbide goes beyond just the substrate (crystal).
This is the most misleading aspect in the entire marketing chain. We can see that in recent years, hundreds of silicon carbide companies have emerged, but they are always ambiguous about their position in the industry chain. Companies that do packaging can even brag about themselves being at the top of the food chain.
“Making silicon carbide crystals, perhaps for sandpaper, perhaps for making necklaces.”
Silicon carbide has an extremely high hardness, second only to diamond, so it can be used as a cutting tool and wear-resistant material. In addition, high-purity silicon carbide is generally indistinguishable from diamond to the average person, and can even be more dazzling than diamonds. On Taobao, a carat only costs 50 yuan, and it is known as Moissanite.
If purity, crystalline form, and unit area defects are not considered when producing semiconductor crystals, there is a suspicion of fraudulent behavior.
Six, Link to heading
Almost all stories about silicon carbide mention electric cars.
After all, new energy is one of the few bright spots in the recent economy and A-shares, and 80% of silicon carbide devices are used in cars. However, there are not many financial reports in A-shares that dare to show how many cars with SiC substrates or car MOS and design wins have been produced. This is because most of them are estimated to be zero.
If there is no MOS substrates, then can we rely on self-researched SiC MOS devices? I’m sorry, but many companies also have zero shipment in this regard.
Why is that? Why is that? Actually, it doesn’t matter. The key to A-shares is the story. If there’s a story, it can rise.
Zero has nothing to do with it. We can use WeChat to broadcast a short essay to tell everyone that we have achieved the “number 1” position, and have successfully developed our product. We are standing at the top of a snow-covered mountain, with a long and inviting snowball path below us.
“A-shares have never been a value investment, and value investing is just packaging for a story.”
So here comes the question, wouldn’t it be better if we strive to turn the story into reality?
Seven, Link to heading
Telling a story is easy, but actually doing it is difficult.
The silicon carbide MOSFET, from crystal to epitaxy to device to module, each step has been built on over ten years of technology accumulation, making it a greater challenge than the long-standing IGBT.
Companies like Wolfspeed are practicing the utmost integrity, and achieving the manufacturing of Silicon Carbide (SiC) wafers is a challenging and expensive process that requires great patience. It is a very slow and costly process that does not turn a profit for the longest time. There are over 200 types of SiC crystals, but only three can be used for MOSFET fabrication. This makes the process of creating the seed crystal a skill that is closely guarded by each company and considered a top-secret art.
II-VI Corporation, second in ranking for silicon carbide crystal, inherits its technology from the legendary GE. GE, under Welch’s leadership, was too focused on making quick profits and thus cut funding for cutting-edge technologies such as their own inventions of IGBT and SiC devices that required long-term investments but did not bring in immediate profits.
II-VI is also a company with poor profitability. Substrate business accounts for a single-digit percentage of the company, generating only about $100 million in revenue per year, most of which is for communication purposes and using semi-insulating substrates.
Translation: If silicon carbide single crystal substrate doesn’t make a profit, what about making devices? The fact is that the threshold for producing MOSFET devices is only higher, not lower. Growing crystals through melting and buying imported furnaces are enough for practice and storytelling, but producing devices requires a semiconductor factory that costs tens of billions and decades of skilled craftsmen to monitor the process, and also requires jumping over other companies’ patent walls.
Eight, Link to heading
Currently, ST is ranked number one in SiC MOSFET devices and it is recognized by Mr. Ma for its exceptional quality. In addition to its devices, ST’s packaging modules are also robust and impressive. However, ST’s stock price is rather unfortunate with a P/E ratio of only 12x.
Infineon’s SiC devices rank second, with the main electric vehicle clients being Volkswagen and Hyundai. Although the quantity is considerable, their impact on the stock price is negligible.
Third-place Wolfspeed is attempting to increase its profit margin by ramping up production of its devices. Wolf’s bet is on its new 8-inch factory in Mohawk Valley, which is expected to reach its designed production capacity by 2024.
Therefore, Wolf’s story still continues, as it is not yet the time for the lottery.
The situation we see is that Wolf is unwilling to keep all the unprofitable substrates for device production. Of course, their production capacity is also not sufficient. It is still unclear how Wolf will differentiate themselves in the future and compete with the major players in the industry. Considering their technical capabilities, product portfolio, and customer base, I personally believe that Wolf is unlikely to gain significant advantages in the device sector.
However, the possibility of Wolf ultimately gaining the courage to challenge and confront the old bosses who supply the raw materials and causing a stir is not very high.
Looking back, even if Wolf made all their substrates into components instead of selling them, would they definitely make more profit than Infineon and ST?
We have seen another version of the story: ON Semiconductor. The production of SiC MOSFETs from Infineon and ROHM, which use a multi-level trench structure, theoretically surpasses those from ST, Wolfspeed, and ON Semiconductor, which are flat structures, by at least one generation. The former not only has high output, but also has advantages in terms of performance and may have taken a significant lead in patents.
However, in terms of combining storytelling with market demand, the CEO of ON Semiconductor, Hassane El-Khoury, excels even further. Coincidentally, during the prosperous period of the automotive industry, ON Semiconductor has reported a series of impressive financial reports. Although the contribution of silicon carbide is limited, Mr. Hassane’s determination to go all in on silicon carbide has made investors extremely pleased.
As the main executor responsible for selling Cypress, Hassane may also have thoughts of taking another stab at ON Semiconductor. When it comes to carrying out such operations, timing, location and human factors are all very important. ON Semiconductor’s silicon carbide supply chain is already complete, and the external environment is also flawless. ON Semiconductor has made its move in the game.
Ten, Link to heading
According to authoritative market analysis agency Yole, even by 2025, the global total market capacity of silicon carbide MOSFET devices will only be just over 3 billion US dollars, meaning that the share that each giant can receive is limited.
If domestic players suddenly have a miraculous breakthrough in technology, production capacity can quickly climb from 0 to snatch a 20% market share, which is only 600-800 million.
Silicon carbide is not an industrial necessity, but rather a beautiful spray of water under the giant wheel of new energy.
In a situation where market liquidity is still abundant and money is unwilling to enter the real economy, being beautiful is enough.
Reaching consensus in the market doesn’t necessarily require the so-called truth or meticulous calculation, as it seems that looking good is the key.