Since changing its name, has the silicon carbide leader been struggling?

Last night, Teacher Zhu Jing posted a paragraph on her WeChat Moments.

She said: I couldn't find a few articles analyzing why Wolfspeed fell from grace today on the entire internet... I really have to think about metaphysics. Indeed, since the end of 2021 when Cree changed its name to Wolfspeed, its market value has been dropping continuously, down to only 20% of its original value...

It sounds harsh, but Teacher Zhu is speaking the truth. I took a look, and now Wolfspeed's market value is only a little over $1.7 billion, compared to over $12 billion at its peak three years ago. It's almost really down to just 20%...

Today, I also came across an article titled "Playing a Good Hand Badly: Silicon Carbide Wafer Giant Wolfspeed on the Brink of Bankruptcy."

Of course, the title is a bit exaggerated, bankruptcy is not imminent in the short term, but in the long term, it's really hard to say!

However, the original author only listed Wolfspeed's financial data, such as Q4 revenue of $200 million in 2024, slightly below expectations, with a loss of up to $174 million, a loss per share of $1.39, far exceeding the expected 85 cents, with a total annual loss of $860 million. It is the worst-performing company in the entire Philadelphia Semiconductor Index, with a cumulative decline of 69%.

Everything said is correct, but there is a lack of a summary of history, a judgment of the future, and unique insights. The biggest difference between me and others, Aito and others, is that I have my own thinking, and I can present my thought process and conclusions to you. Today, let's talk about Wolfspeed, why is it not doing well?

Uncovering Wolfspeed's History

Three years ago, it was still a hot star company, the leading company in silicon carbide globally. How did it fall so drastically in just two to three years? If you didn't know, you would think it was a company in the A-share market. Of course, the stock price in the secondary market cannot explain everything, nor can it represent the future. AMD also had its darkest moments back then, but it eventually rose from the ashes and returned as a king.

However, a continuous decline with no end in sight in the short term can still indicate some issues. After all, the capital market votes with money, so such a decline definitely indicates that something is wrong. So is it that the silicon carbide track is not doing well, or is it just Wolfspeed that is not doing well? Or are there other reasons?

First, let's understand Wolfspeed. Wolfspeed, formerly known as Cree, is a U.S. company founded in 1987 and listed on Nasdaq in 1993. In the past, it was also one of the earliest companies in the world to work on LEDs.

Back when the technology for growing gallium nitride on sapphire was not mature, early LEDs, especially high-power LEDs, used silicon carbide as the substrate material because the lattice matching between silicon carbide and gallium nitride was about 95%, so the quality of the grown gallium nitride epitaxial layer was good, making it suitable for LED production Cree's main business is to produce silicon carbide substrates, grow gallium nitride epitaxial layers, and sell LEDs, eventually becoming the global leader in silicon carbide.

In 2017, the global power semiconductor leader Infineon once offered $850 million to acquire Cree's Wolfspeed Power&RF division, but the deal was rejected by the US due to antitrust scrutiny. Interestingly, Infineon then sold its RF-related business to Wolfspeed, making Wolfspeed to some extent the strongest company in the global compound semiconductor application field.

In 2021, to demonstrate its transformation determination, Cree not only divested its LED business, which accounted for two-thirds of its revenue, but also changed its parent company's name from Cree to Wolfspeed.

As a result, Cree transformed from an LED company to a leading company in the compound semiconductor industry, with businesses including compound material manufacturing, as well as power, RF, and other device production, design, and applications.

It must be said that Cree had a deep foundation in the silicon carbide industry for over 30 years, but the company used to be lukewarm. Every 5 or 6 years, industry giants would come out and proclaim that the era of silicon carbide was coming soon, similar to the current hype around the silicon photonics era, with a surge followed by silence.

However, everything changed in 2018.

That year, Tesla first tried using silicon carbide power modules in its electric vehicles to replace traditional silicon power modules, instantly igniting the silicon carbide industry along with the electric vehicle industry, and also boosting Cree at the time.

Since then, silicon carbide has moved from being just talked about to being actively used. Electric vehicles provided an excellent application scenario for silicon carbide, opening up its practical application.

Previously, silicon carbide was a niche product, with very few high-power RF, power, military-related applications. It was rare to see practical applications of silicon carbide products. However, Tesla's move completely changed the industry's perception.

People discovered that silicon carbide not only had high power density but also excellent high-voltage resistance, large current capability, and high-temperature performance. It also reduced weight, volume, and peripheral components, making it a perfect fit for the limited space in electric vehicles. Although silicon carbide is expensive, the high value of electric vehicles can justify the high cost of silicon carbide, making the business logic clear!

Therefore, in electric vehicles, silicon carbide finally found its best application scenario and quickly gained popularity. At that time, silicon carbide production capacity was extremely limited, and Cree, as the world's largest supplier of silicon carbide substrates and epitaxial wafers, took off directly!

Pressure from Chinese counterparts

It cannot be said that Wolfspeed's current plight is completely unrelated to Chinese counterparts, but there is still some connection. Chinese counterparts, in their ruthless bargaining tactics, have pushed Wolfspeed to the brink Ten years ago, in 2014, the Nobel Prize in Physics was awarded to three giants who made outstanding contributions to the blue LED industry: Isamu Akasaki, Hiroshi Amano, and Shuji Nakamura.

Back in those years, with the concept of the LED industry gaining popularity, it sparked a wave of LED company stock prices, such as Sanan Optoelectronics. Despite experiencing the stock market crash in 2015, by 2017, the stock price continued to hit new highs, with an increase of over 200%.

Everyone was immersed in the capital frenzy, but unfortunately, everyone forgot that saying.

"In the eyes of the Chinese, once technology is mastered, it will definitely be sold at cabbage prices. Only Chinese counterparts are considered counterparts, not those from abroad."

The Moore's Law of the LED industry - Haitz's Law

Similar to the famous "Moore's Law," LEDs also have their own industry development rule, known as "Haitz's Law," which states that the price of LEDs decreases to 1/10 of the original every 10 years, while the luminous output increases 20 times.

Therefore, the two key elements for LED development are price and performance, ultimately achieving an improvement in the overall cost-effectiveness of LEDs. However, disappointingly, China irrationally slashed prices, causing LEDs to lose direction.

In 2015, with the optimization of technologies such as large-size sapphire crystal growth, patterned sapphire substrate (PSS) technology, sputtering transparent conductive layer technology, optimized reflective electrode structures, semiconductor automation equipment, and the introduction of MOCVD equipment, the focus of the LED industry gradually shifted to the mainland.

After Cree announced the development of 303 lumens per watt technology in 2014, companies from Europe, America, Japan, South Korea, and Taiwan basically fell silent, leaving only mainland Chinese companies. And China's contribution to Haitz's Law was mainly sacrificing profits by continuously slashing prices to extend, rather than letting technology lead the industry development forward as in the integrated circuit industry governed by Moore's Law.

Foreigners used to describe China as the "butcher of prices." Anything that Chinese people are capable of will eventually be sold at low prices, from steel to glass, electrolytic aluminum, polysilicon, photovoltaics, and other industries, all of which have become industries with excess capacity. This time, it's LED's turn.

Although mainland China has not invested much in technology, it has put a lot of effort into reducing costs. The localization of silicone gel has led to a rapid decrease in glue costs; to make LEDs brighter, bold bosses drive the LED chips with high current overload regardless of whether the LED lifespan is good or not, forcing the device's current density to be three times higher; the use of alloy wires and the substitution of palladium-coated copper wires have further reduced costs.

Several well-known companies, in order to break through the limits of formal wear, actually test the material inspection standards based on the ratio of current to area. For a chip with an area of 10mil*30mil, if the test current is 300 milliamperes and the voltage is greater than 3.4 volts, it is considered unqualified. Such requirements are difficult to meet for formal wear chips, as they need to sacrifice epitaxial structures to meet them, leading to more serious defects. As a result, the quality of these companies' LED chips deteriorates, and the prices continue to decline, relying only on output to maintain their position

Why are there still 6-inch wafers in China while 8-inch wafers are more common overseas?

This question was actually explained very clearly by Professor Pengxin during the live broadcast.

The reason why China is still using 6-inch silicon carbide FAB is because there are still many 6-inch silicon wafer resources in China. At least there are still many devices that are compatible with both silicon and silicon carbide, such as Nikon's i11 and i14 lithography machines. These devices are widely available. However, if you were to look for an 8-inch lithography machine, it would be very difficult. The same goes for cleaning, testing, deposition, and other equipment. In short, resources for 6-inch silicon wafer equipment are extremely abundant, while 8-inch equipment is much scarcer.

Although silicon carbide has its own dedicated equipment such as epitaxy, etching, implantation, annealing, PVD, etc., which require new equipment, major foreign manufacturers directly use 8-inch wafers. 6-inch wafers are transitional and experimental equipment. However, some things can still be cost-effectively modified from silicon equipment. Regardless of how much less efficient or lower yield it is to modify 6-inch equipment compared to brand new 8-inch equipment, as long as it can be used and produce chips, you can modify it first and worry about the rest later! Use it while modifying it, there are subsidies anyway. Using all 8-inch wafers, not to mention being expensive, doesn't seem significantly stronger at the moment. To be honest, many 8-inch equipment are not considered mature yet, blindly switching to 8-inch is like being a guinea pig.

In conclusion, 8-inch equipment is very expensive, and the cost of building a factory is much higher than that of 6-inch equipment.

This situation is very similar to the dispute between Topcon and HJT heterojunction lines in the photovoltaic industry.

You see, traditional photovoltaic companies tend to use Topcon, while newcomers tend to use HJT. Why? Isn't it because Topcon lines have good compatibility with prec lines and low modification costs...

So the fact is that Topcon accounts for more than 70% of the market!

Why don't foreign countries use 6-inch wafers? It's because they have phased out their 6-inch wafers and no longer have resources for them. In this case, why bother with 6-inch wafers when you can directly use 8-inch wafers? After all, the area ratio between 6-inch and 8-inch wafers is about 1:1.77, the advantages are quite obvious. The downside is that the cost of building an 8-inch wafer factory is much higher than that of a 6-inch wafer factory, one requires heavy investment in new equipment, while the other is about saving money and making do, can they be compared?

Therefore, it is destined that making MOSFET on 6-inch wafers will definitely be lower in cost than making MOSFET on 8-inch wafers. Of course, the future will definitely belong to 8-inch wafers, but personally, I believe that from a commercial logic and cost perspective, 6-inch wafers cannot be eliminated by 8-inch wafers in the short term and will continue to exist in the long term, simply because of cost considerations.

The only way for 8-inch wafers is to make high-end MOSFETs, and in addition, they need to develop structures similar to Trench structures, try every means to increase power density, reduce chip area, and increase the output of dies on a single wafer in order to make money.

This leads to the second topic, the debate between trench and planar structures on silicon carbide MOS.

Currently, worldwide, there are three factions on this issue, the planar MOS faction such as Cree (Wolfspeed), and the majority of domestic silicon carbide manufacturers in China;
The trench camp, mainly represented by Rohm, has a well-known reputation for its advanced technology. However, the cost is also high. The main issue lies in the anisotropy of silicon carbide, making etching much more difficult than silicon due to its hardness. Therefore, high-energy CCP etching is required, and ICP is not sufficient.

The third group falls between the two, such as Infineon. Its trench structure is different from Rohm's, with angled trench gates, known as 4C. This group seems to have no accurate name. There is a big brother in China researching silicon carbide who calls them semi-packaged structures.

Currently, in terms of device performance alone, trench structures are definitely superior. However, when compared to planar structures, the complexity of the process increases significantly. Trench processes are too complicated to implement easily.

Of course, in terms of gate reliability, trench structures still have a significant advantage. The SiC-SiO2 interface is much better than the Si-SiO2 interface.

In other words, to maintain the same gate oxide reliability, both planar and trench structures need to use the same thickness of gate oxide layer. The area of planar MOS devices must be significantly larger than trench gate devices. If the same chip area is to be maintained, in order to maintain low on-resistance, planar MOS devices require a thinner gate oxide layer. However, this leads to higher gate oxide stress and reduced reliability.

Therefore, each has its own strengths at present. It's hard to say for sure whether trench structures will definitely outperform planar structures. As the saying goes, it depends on the application, the customer, the scenario, and ultimately, the cost.

Professor Ji also mentioned during the live broadcast that although planar structures seem to have many disadvantages compared to trench structures, they have mature processes, high yields, and the technological roadmap has not reached its end, with great potential to be tapped. If this potential is realized, could it give a new lease of life to 6-inch fabs?

At this point, I bring up another non-issue, which is the battle between 6-inch and 8-inch substrates and epitaxial wafers.

There is no need to say much about 6-inch wafers now. In China, simple 6-inch silicon carbide SBDs, the kind of lower-quality 6-inch substrates, are already priced at less than 4000 yuan. For slightly better quality, they are used for epitaxial growth to make MOS devices, costing less than 5000 yuan. However, currently, 8-inch silicon carbide substrates generally cost over 9000 yuan.

In other words, the area ratio of 6/8 inches is 1:1.77, but the cost ratio is over 1:2. Therefore, using 8-inch wafers clearly incurs a loss. Of course, the final usable area of 8-inch wafers is larger than 6-inch wafers, slightly offsetting the disadvantage in terms of area cost ratio.

Most importantly, 6-inch wafers currently have sufficient volume, while 8-inch wafers are just starting to be built in China, with limited substrate production. Of course, there will be sufficient substrate capacity in the future, but the fabs have not been built yet. The question is, who will be the hero? Or will the hero not succeed and become a martyr? Brothers, this is not about scientific research, it's about doing business. Making money is always the top priority. You may not make money in the short term, but if you don't make money in the long term, you definitely won't be able to continue. Just like the current struggles of Wolfspeed.

Wolfspeed is currently stuck at this critical juncture. Although it still has over 2 billion US dollars in cash, it is losing up to 600 million every quarter. This money won't last long
After talking about the substrate, let's talk about epitaxy. A 6-inch wafer can be used for low-end SBD without a long epi layer, but for MOS, a long epi layer is necessary. Now, with the 6-inch epitaxy technology, since the domestic introduction of equipment from an Italian company called LPE (note: LPE has been acquired by ASMi), and thorough digestion, it has been localized and can be done very cheaply. Domestic production can now be done for less than 5 million RMB per unit, with the technical direction called horizontal airflow.

An 8-inch Epi can also be done using horizontal technology, but the uniformity, consistency, and quality are poor. These wafers are not suitable for making MOS, especially for high-end ones like 3300V1200A.

Therefore, high-end MOS chips require high-end Epi, with a technical route called vertical airflow, represented by Japan's Nippon Electric Glass. The downside is that this equipment is extremely expensive, costing over 30 million RMB per unit, while domestically produced ones can be done for around 25 million RMB.

In summary, from the 6-inch substrate, to epitaxy, to fab production, to the product line, it is cheap, with a high degree of industrial maturity. The downside is fierce competition, price wars, turning a high-margin industry into one where no one can make money, similar to the photovoltaic industry. It is worth mentioning that Zhuhai is not doing well, and all competitors are hoping for its downfall to leave breathing space for others.

The cost of 8-inch substrates, epitaxy, fab production, and process difficulty are all challenges. The domestic industry is mature but not high, just 1-2 years ahead of foreign counterparts. Therefore, the challenge lies in the transition from 6 to 8 inches, how much it will cost, how many years it will take, who will be the martyr, who will be the hero, who will benefit, and who will suffer.

From the current perspective, in my opinion, 6-inch silicon carbide should remain mainstream for a considerable period of time. While 8-inch seems promising, there are still many issues to be resolved. If these issues are not addressed and costs do not come down, it is destined to be marginalized. It sounds poetic to talk about it positively, but in reality, it's like dying before seeing the dawn.

Returning to the initial question posed by Mr. Zhu, how did Wolfspeed fall from grace in just two to three years?

Wolfspeed's embarrassment lies in the fact that they haven't made much money yet, haven't fully monopolized the technological position (or have a technological gap that Chinese counterparts are catching up with), and haven't monopolized the production capacity. They are already feeling the deathly gaze from domestic Chinese competitors.

From LEDs to silicon carbide power semiconductors, Wolfspeed's boasting has never been fully realized, and their path has never been smooth.

Author: Chen Qi, Source: What's the plan, Qi-ge?, Original Title: "Since the name change, is the silicon carbide leader not doing well anymore?"