Learn about the secrets of Cree, the leading SiC company

SiC was discovered as a material in the 19th century. In the following century, SiC was widely used in mechanical processing and smelting due to its high hardness. It was mainly used in the fields of functional ceramics, advanced refractory materials, abrasives and metallurgical raw materials. It was also promoted as a kind of jewelry - moissanite.

Although the first SiC diode was born as early as 1907, due to the variability of SiC's own structure, the exploration of SiC's industrial crystal growth lasted throughout the 20th century and is still being explored and improved. As a wide-bandgap semiconductor, compared with traditional silicon-based devices, SiC's breakdown field strength is 10 times that of traditional silicon-based devices, and its thermal conductivity is 3 times that of traditional silicon-based devices, making it very suitable for high-voltage applications. SiC components can achieve lower on-resistance, higher switching speeds, and adapt to higher temperature working environments.

Comparison of the core indicators of several semiconductor materials (of course, the table lacks a core element - industrialization cost):

Source: CCID Think Tank

The industrial chain of SiC power devices includes: substrate -> epitaxy -> devices (design, manufacturing, packaging and testing) -> applications and solutions -> terminal product applications.

Source: Drip Research

Substrate and epitaxy belong to the upstream links of the SiC industry. As the material link supporting the development of the entire industry, its importance cannot be overemphasized, whether from the perspective of the performance of the final device or the cost factor.

As a pioneer and leader in the development of the entire industry, Cree has an absolute leading advantage in the field of substrates and epitaxy - with a market share of about 50%. In the past two years, it has been "hard to find a piece". For downstream device giants, the best way to ensure supply is to sign a long-term supply agreement with Cree. The amount of Cree's current long-term supply contract for SiC materials has exceeded 500 million US dollars, and this number is expected to continue to increase rapidly - this model will play a great role in promoting the development of this industry, which will be elaborated later when explaining the driving force of industrial development.

In addition to Cree, II-VI, Dow Corning, Rohm, and Showa Denko are the main players. Together, they form the basic ecology of the SiC material industry, and their trends will also have a significant impact on the entire SiC industry.

As the home team of the world's largest semiconductor device, domestic SiC material companies have also developed rapidly in the past two years. Thanks to the mutual support and promotion of all links in the industrial chain, domestic suppliers on the substrate side have been able to provide relatively good 4-inch substrates in large quantities, and the evolution of 6-inch wafers is also accelerating. In terms of epitaxy, after years of process accumulation and optimization, some domestic companies are basically in the same group as foreign giants, and they are not inferior in head-on competition. Good progress on the material side is crucial to the development of the domestic SiC industry.

In terms of materials and equipment, due to the uniqueness of SiC substrate material growth, several substrate giants, including some domestic substrate manufacturers, have developed and produced their own substrate furnaces, which has further increased the difficulty of entering the substrate industry - foreign substrate giants basically do not sell furnaces to others, and the quality control of substrate growth, a lot of know-how is concentrated on the substrate furnace.

Epitaxial furnaces are relatively mature, mainly using MOCVD equipment, and the equipment players are very concentrated, including AiXtron, LPE and VEECO. With the continued popularity and high expectations of this industry, the capacity expansion of various companies has made epitaxial furnaces in short supply, and the delivery time is generally more than half a year.

Source: Drip Research

In terms of devices, they are divided into IDM manufacturers and Fabless manufacturers. Here we mainly explain SiC power devices. As for the application of SiC substrates in LED and GaN devices, LED will not be explained, and GaN devices will be described in detail in the next article. IDM manufacturers are currently mainly foreign giants. Except for Cree, most of them are traditional power semiconductor giants. Their relative competition and power in the traditional power semiconductor market also affect the development rhythm of the SiC industry. These will be described in detail later. Domestic IDM manufacturers have gradually made efforts to catch up in recent years (according to the prospectus disclosed by China Resources Microelectronics, it will invest 600 million to focus on supporting SiC and GaN, and China Resources Microelectronics’ layout in the third and a half generations should still be an IDM approach), but the gap is still relatively large.

In the Fabless-Foundry model, independent third-party fabs are currently mainly X Fab and Taiwan Epistar. In the future, new domestic players will probably join in (whether it is the capacity conversion of the four- and six-inch lines of traditional Si Fabs or new investments). The Fabless-Foundry ecosystem built around these fabs will gradually become a very important force in SiC power devices.

In terms of downstream applications, SiC has applications or potential applications in power electronics and many traditional high-voltage and high-power power semiconductor applications. As shown in the following figure:

Source: Yole

EV is a large downstream market for SiC power devices/modules, which have been applied in OBC, DC/DC and Traction Inverter. Tesla is the first automaker to integrate full SiC power modules in its Model 3, and its engineering design department works directly with ST - thanks to ST's accumulation in power semiconductor packaging, this is currently the source of most of ST's SiC business. The main inverter of Tesla M3 consists of 24 1-in-1 power modules, which are assembled on pin-fin heat sinks. As shown in the figure below:

Image source: Automotive Electronics Design

Professional research organization Yole Development has made predictions on the application space, capacity and respective share of SiC power devices (as shown below). According to Yole's data, the CAGR of SiC power devices will exceed 30% in the next few years, with new energy vehicles and charging facilities being the two fastest growing application scenarios.

Source: yole

At present, SiC diodes are very mature, and both domestic and foreign players are able to mass produce them. However, there is a big gap between domestic and foreign players in other devices such as MOS. From the application perspective, although most application customers have (or claim to have) research and solutions for SiC devices, frankly speaking, a deep understanding of SiC devices, exploration and research of their stress and boundaries, from the perspective of the application end, still requires a longer process.

Difficulties in the industrialization of major links

Precise control during the production of substrates has always been a core difficulty. The growth temperature of SiC single crystals is as high as 2,300°C, and silicon carbide has only two phases, "solid-gas". Compared with the three phases of "solid-liquid-gas" of the first and second generation semiconductors, it is much more difficult to control, and there is no relevant technology for reference. In addition, there are almost 200 isomers of SiC single crystal structure, and the free energy differences between many crystal forms are very small, which poses a great challenge to the industrial growth and preparation of single crystals. As a direct result, defects in SiC single crystals have always been the core problem to be solved.

In terms of epitaxy, the challenge of defect control in the substrate continues - similar to the substrate, precise control of its growth process is also difficult. The design of the growth process needs to comprehensively consider multiple factors such as device requirements, defects and process control. The challenges on the device side are mainly on MOS tubes and above (in terms of structural complexity), such as the preparation of the device gate oxide layer. Even the industry giants of power devices still need to continue to break through and improve. Due to the high hardness and high temperature processing environment of SiC materials, the challenges in the doping process are also relatively large.

The matching of applications and solutions directly determines the promotion and application of SiC devices. Currently, most of the application environments of power semiconductors are still based on silicon-based products. It is necessary to design the application ecology and supporting peripheral circuits according to the characteristics and requirements of SiC devices. Therefore, the lack of application and solution providers is a focus that urgently needs to be broken through for the development of the entire industry.

Of course, the core issue cannot be avoided: cost.

Some opinions that are often heard are that silicon carbide devices are still too expensive. According to the law of industrial development, only when the cost and price fall below certain critical points will large-scale batch applications take off. This is a dynamic and complex process.

Here we will only analyze the "cost" from some logical levels - those who support the application of SiC power devices will mention that the system cost of SiC power devices is close to and has great potential to be lower than the system cost of Si-based devices, because SiC power devices will make the supporting circuits simpler and have fewer units, thereby reducing costs at the system level. As shown in the following figure:

Source: CCID Think Tank

Infineon also explains this logic in its public information, as follows:

Source: Infineon

Source: Infineon

If we extend this logic a step further: Chips are actually circuit systems, and SiC power device chips themselves benefit from the more efficient characteristics of smaller size. For products with the same indicators and performance, the number of dies that can be produced on a SiC wafer is roughly equivalent to the output of 6 Si wafers.

Therefore, we believe that the cost potential and logical support of SiC power devices are tenable. However, this logic must also withstand the challenge and test of actual application conditions. One is the wafer cost and the cost of each die. A more appropriate comparison is to compare with the situation of 12-inch lines. How many times will this critical value be compared with the situation of 12-inch Si wafers? One is that the positive and negative benefits of smaller size must be taken into account, especially in combination with subsequent packaging and applications. Perhaps many of the apparent benefits actually need to be discounted.

The difficulties in SiC industrialization are summarized as follows:

• Materials – high cost, defect density (yield), wafer size and wafer supply

•  

• Devices – high cost, production lines, long-term reliability, packaging, and reliable supply chain relationships

•  

• System/Solution——Peripheral Support and Application Ecosystem

•  

Source: Yole

Of course, the promotion of any new thing is influenced by the inherent development logic and driving force of the industry, and the same is true for SiC power devices. Next, let's try to analyze the current industrial momentum of SiC power devices.

The driving force of the SiC power semiconductor industry

First, let’s look at a recent explosive piece of news:

On May 7, Cree announced that it would invest USD 1 billion to expand SiC production capacity, including integrating an 8-inch wafer fab (USD 450 million will be invested in North Fab to expand the factory and production capacity) and a SiC material factory (USD 450 million will be used for mega factory and the remaining USD 100 million will be used for related investment in other SiC businesses), expanding its SiC material and wafer manufacturing capabilities by 30 times (compared to Q1 2017).

We don't know how much confidence Cree has in SiC's future (by 2024), but playing such a card is consistent with the signals Cree has been sending to the outside world in the past two years. For example, in the Earnings Call Transcript for Q2 of fiscal year 2019, Cree expects capital investment of about US$220 million in fiscal year 2019, most of which will be used to expand Wolfspeed's production capacity. Even if "...that may reduce our near-term Wolfspeed gross margin." Second, Cree believes that the turning point of adopting electric vehicles and silicon carbide has been reached - automakers have announced plans to spend at least US$300 billion on electrification projects, and their interest in silicon carbide is very high. Therefore, the company needs to ensure its long-term supply capacity in the field of SiC. These trends are being verified by the long-term wafer supply agreements they have signed with downstream companies, which currently total more than US$450 million, including a contract worth more than US$250 million signed with STMicro this year.

As Cree's strong competitors in SiC materials, II-VI, Dow Corning, Rohm and Showa Denko have also continued to expand their production capacity in recent years. II-VI stated in its Q3 2019 conference call: SiC-based power electronic devices (including those for electric vehicle applications) grew by 70% in 2018. The entire SiC application market is strengthening, especially in China. In terms of capital investment, a large part of its capital investment in fiscal 2019 was used for SiC. II-VI has a mass production improvement plan for each of its SiC material factories, the core of which is to maximize continuous production improvements, and plans to double its production capacity in the next 18 to 24 months. Combined with the feedback from communication and close contact with downstream customers, II-VI believes that the market has clearly underestimated the demand for SiC substrates. In addition, in terms of locking in downstream demand, similar to Cree, most of II-VI's SiC substrate demand is long-term contracts.

Showa Denko, a leading SiC epitaxial manufacturer in Japan, has also been active in the past two years. After announcing the expansion of its SiC epitaxial production capacity for four consecutive times, its SiC epitaxial production capacity has increased from the initial 1,500 wafers/month to 9,000 wafers/month in Q1 2019. Taiwan-based epitaxial manufacturer Chip-Chips had a 4-inch SiC epitaxial production capacity of 1,500 wafers/month in 2018, and plans to add a 6-inch production capacity of 1,000 wafers/month in 2019.

The driving mechanism of an industry's development is very complex, but the underlying factors can be summarized as: cost (capacity), price (demand), performance, application and final export. Here is a rough simplification of SiC's industrial driving force (the logic of performance has been explained above, so I won't add it here), as shown in the following figure:

Source: Drip Research

Judging from the actions of upstream material manufacturers, they have already or intend to turn the key to the starting point of the next round of industrial development - regardless of whether much of the production capacity has been realized, a statement can also have a similar effect: for players in the industry, do you believe the statement of a company with an industry status like Cree, or not?

Of course, there will be many very detailed issues in the subsequent power transmission, which we will continue to track and study.

Material manufacturers are not blindly expanding their production capacity - this can be seen from the above summary of their public statements. One is that they have many long-term agreements, and the other is that they feel that the industry momentum is stronger than many people think (although this may also be a means for them). Next, we will also sort out some public information of several SiC device giants and sort out their views on the SiC industry momentum.

First, here is STMicroelectronics' Earnings Call Transcript for Q1 of fiscal year 2019:

“Then about silicon carbide, so our revenue target for 2019 is to be about US $200 million so to double the revenue addressing MOSFET, but also in certain extent diodes. Mainly, this revenue will come from our engagement in our main customer. But important to say, again, and to share with you that we are already engaged with 30 important programs with various customers across the various region in the world to address both electrification of the car and industrial end market. But early in 2019, the main part of our revenue will be linked to our already engaged main customer.”

ST's sales revenue target for silicon carbide devices in fiscal 2019 is about $200 million, which means that the revenue from MOSFET/diodes will double compared to 2018. This part of the revenue is currently supported by the main customer (Tesla). However, ST already has about 30 major projects underway with downstream customers.

“For Silicon Carbide, we announced the agreement to acquire the wafer manufacturer Norstel, on top of the multi-year supply agreement with Cree we already announced......In Q1, we also started to provide global carmakers with samples of our ACEPACK Drive module with SiC embedded.”

ST acquired wafer manufacturer Norstel and reached a multi-year supply agreement with Cree. It also began to provide samples of SiC-embedded ACEPACK driver modules to global automakers.

Then let's take a look at On Semi's Earnings Call Transcript for Q3 of fiscal year 2018:

“We recorded our first silicon carbide revenue from automotive end markets in the third quarter. We are actively engaged with leading global automotive OEMs on many silicon carbide projects. We expect silicon carbide will be a significant driver of our automotive content increase, driven by electrification of the drive train. We expect to see strong acceleration in our automotive silicon carbide revenue for foreseeable future.”

On Semi reaped its first silicon carbide revenue in the automotive end market in Q3 of fiscal year 2018, and is actively working with leading global automotive OEMs on many silicon carbide projects. It is expected that silicon carbide will become an important driving force for On Semi's automotive business, and silicon carbide revenue will grow significantly.

“Capacity utilization in third quarter was in the mid to high 80s. Expect that to be similar, maybe coming down slightly in the fourth quarter. Then silicon carbide, we are outsourcing the raw wafers, we have long term agreements on that front. We do internally our own raw wafers for regular silicon, and we talked about that several times that we are increasing our capacity to serve more and be less dependent on these input costs, but not on internal--not on the silicon carbide right now....The production based on the capital investments we made this year will get us to about 50% internal supply, and I don't see that shifting significantly in 2019.”

On Semi has long-term agreements with partners for silicon carbide wafers, which is different from silicon wafers. Currently, more considerations for SiC products are to increase service capabilities and reduce input costs, and outsourcing is a suitable option. The internal supply is about 50%, and this has not changed much in fiscal 2019.

“We are not giving specifics on that yet, but as I mentioned, in total silicon carbide would be in the tens of millions this year, ramping multiples each year.”

On Semi's total silicon carbide production in fiscal 2018 will reach tens of millions, increasing several times every year.

Next, let's sort out Infineon's Earnings Call Transcript slide for Q1 of fiscal year 2019:

It is expected that the sales revenue share of Infineon SiC devices will increase from 1% in 2018 to 20% in 2025. From a single device supply to a multi-module supply:

Source: Infineon

Infineon's progress and plans for SiC solutions in the automotive industry (2018->2020), from charging modules to main transformers:

Source: Infineon

Application and penetration forecast of SiC power devices in EV charging systems:

Source: Infineon

"So Silicon Carbide itself is still young and ramping for us. We're talking as I mentioned in the last call tens of millions last year, ramping to hundreds of millions in the next couple of years. So, it's still pretty early in that. We did see our first usage in EV automobiles for Silicon Carbide in both third and fourth quarters. I'm expecting EV to ramp quite nicely with all of our related products this year. I don't have an exact growth number but it's certainly going to be very high-double-digit numbers."

For Infineon, the overall positioning of SiC is still in its early stages - the market is changing from tens of millions of dollars per year to hundreds of millions of dollars per year. In the second half of fiscal year 2019, it is expected that Infineon's SiC products will begin to break through in the application of electric vehicles. The overall judgment is that silicon carbide devices will grow rapidly.

Finally, let’s sort out the layout of Toyota, a “super export port”, in the field of SiC:

As a veteran driver who established a SiC-dedicated cleanroom in 2013 and released the Camry Hybrid SiC prototype in February 2015, Toyota stopped most of its promotion in the SiC field in 2017. Many friends in the industry believe that this is not Toyota's abandonment of SiC, but evidence that it has entered the "latent mode" of mass production. Toyota's current strategy is to further expand the electrification technology used in the Prius, which is shipped in large quantities, to other models. If this policy is followed, the next generation Prius, which will be launched around 2020, should use new SiC power components.

Source: Toyota

Judging from some of the information released by these giants, the application scenarios and trends of SiC are relatively clear, but each company has its own plans for the SiC pie: Infineon, as the leader of power semiconductors, began to promote its CoolSiC as early as 2016, but did not follow up vigorously in terms of its actions - it may be that the maturity of its SiC devices is not good enough and its reliability needs to be tested for a longer time, or it may be that it is weighing the impact of SiC devices on its traditional strong Si-based power semiconductor business. ST and On Semi have much smaller burdens and take bigger steps. ST is more concerned about the possibility of using SiC to increase its share, while On Semi's strategy is probably to continue its past second-place strategy in various markets - follow up in a timely manner, but not to be reckless.

No matter how the players in the industry compete, and even though SiC power devices still have their own various problems, the node for the explosive application of SiC devices is likely to be within the next one or two years.

For Cree, it has made arrangements at the most important stages of the development of the third-generation semiconductors. In terms of SiC materials, it is the undisputed overlord. In terms of SiC power devices, it is also a pioneer. However, I think it should give priority to or even fully consolidate its leading position and moat in the material side. Devices will be a big market, but compared with power semiconductor giants, Cree has no advantage, and with the development of the SiC field in the future, the division of labor in the industry will be more detailed. The long-term competitive advantages and disadvantages will also be more obvious.

Of course, in the GaN business, Cree may face different paths. We will analyze this in the last analysis article of Cree. I wish you all the best.