Abstract: From the evolution of the operating model of a silicon carbide and gallium nitride original factory, this paper discusses the logic behind it and the reference significance for Chinese chip companies to go overseas
Preface
In the article at the end of 2021, "Can Navitas Semiconductor Go to the Peak and Win the Third-Generation Semiconductor Crown?", the author put forward the following speculation: "... In the future, Navitas may adopt a semi-fabless model similar to the Monolithic Power System, in which the company's self-developed process is produced in multiple foundries in order to gain foundry pricing power and flexible production layout, while accelerating product development and quality control through its own test equipment and production capacity."
A year and a half later, in mid-2023, that prediction came true when Navitas announced that it would invest $20 million to build a silicon carbide epitaxial plant at its headquarters in Torrence, Southern California. The company will purchase three Aixtron G10-SiC silicon carbide epitaxial growth equipment. This model is Aixtron's new epitaxial system, introduced in September last year, and supports two silicon carbide epitaxial wafer production configurations: 9x150mm and 6x200mm. It is reported that the same model of equipment was also purchased by Wolfspeed for the production of 8-inch epitaxial wafers.
According to Navitas' plan, the first equipment is expected to be qualified and ready for production in 2024, which will generate an additional $200 million in production value for the company. The investment also marks Navitas' transition from fabless to semi-fabless.
Semi-fabless模式
As the name suggests, the semi-fabless model combines elements of both fabless and IDM models. In addition to focusing on core competencies such as chip design, technological innovation and product market strategy, this model also has its own processes and corresponding equipment and production capacity in some specific manufacturing steps. At the same time, the rest of the manufacturing process is outsourced to wafer foundries and packaging and testing foundries.
The semi-fabless model provides semiconductor companies with the flexibility to operate their business, allowing them to devote most of their resources to design and marketing without having to invest too much capital expenditure in manufacturing. At the same time, this model enables the company to master the technology and capacity of some key manufacturing links, promote the entire production supply chain from point to point, and take into account the protection of core intellectual property rights.
The origins of this model are no longer known. However, according to the author's search, Mr. Xing Zhengren, the founder and CEO of Xinyuan Systems, was the first to propose this concept. In a 2009 interview, he told reporters: "...... Our business model is not a pure Fabless model. We just don't have our own production lines and assembly line workers, but we have our own proprietary manufacturing process technology, and at the same time, we do everything except have production lines. MPS is the first analog chip company to adopt this semi-fabless model."
Three and a half generations of companies under the Semi-fabless model
Navitas' choice of silicon carbide epitaxial wafers as its first step into semiconductor manufacturing has a certain internal logic.
First of all, epitaxial growth accounts for a significant proportion of the cost of silicon carbide devices. Taking 1200V silicon carbide MOSFETs as an example, epitaxy can account for about 20% of the total cost. In addition, if the combination of front-end and back-end processes and application practices in this process ignites the skills of distinguishing between killer and non-killer defects (fatal and general defects), double cost reduction can be achieved in substrate procurement and yield control of self-produced epitaxial wafers. That's why Navitas got into this space in the first place.
Cost model of 1200V SiC MOSFET (Source: 01 Xinwen, PGC Consultancy)
Secondly, the silicon carbide epitaxial layer directly affects the performance and reliability of silicon carbide power devices, especially MOSFETs. For example, the epitaxial layer not only determines the withstand voltage rating of the device, but the process control during the production process also affects the macroscopic defect density of the resulting epitaxial wafer.
At the same time, considering that Navitas, like many of its competitors, has chosen X-FAB as the foundry for its SiC MOSFETs, its self-developed independent epitaxial growth IP can be more suitable for Navitas' Trench-assisted Planar MOSFET technology to a certain extent, and the technology is already differentiated from the competitor's products from the epitaxial layer.
Finally, in-house production of epitaxial wafers can help Navitas improve its control over the supply chain, not only increasing the diversity of silicon carbide material sources, but also shortening the production cycle from substrate to device, enabling a flexible capacity layout.
In addition to the above factors, Navitas dared to take the first step in its own production because it found the right leader. In early 2023, Navitas appointed David Greenock as Vice President of the company, responsible for silicon carbide materials and epitaxial operations. Prior to that, Greenock served as Director of Silicon Carbide Epitaxial Products and Business at two related companies, Globitech and X-Fab. In addition, during his tenure at Globitech, Greenock was responsible for the introduction and implementation of the Aixtron G10-SiC epitaxial growth equipment. Therefore, it was only fitting that Navitas chose him as the head of extended operations.
Aixtron G10-SiC silicon carbide epitaxy system (Source: Aixtron)
Of course, Navitas is not the first third-generation semiconductor company to practice the semi-fabless model. Swedish silicon carbide start-up Ascatron has also adopted this business model, designing its own silicon carbide power devices while producing key epitaxial materials in-house, while outsourcing chip manufacturing and packaging. In 2020, Ascatron was acquired by II-VI.
Semi-fabless到IDM？
At Navitas' recent 2023 investor conference, CEO Gene Shelden revealed to the media that the company plans to shift from the past model of relying entirely on X-FAB and TSMC for front-end wafer manufacturing to a wider range of foundry partners, such as some third-generation semiconductor fabs that have come by transforming old silicon fabs. Navitas is even considering making small, strategic investments in these new partners. This signifies that Navitas is not only moving from fabless to semi-fabless, but also to move further towards a fab-lite (light fab) model in the future.
The reasons for this strategic adjustment are also related to de-globalization. In my work in recent years, I have met many customers who are reluctant to accept products from the Chinese supply chain. For example, Dell has made it clear that the chips it chooses cannot have Chinese elements in the production process. As a result, even though a South China supplier had a more advantageous price and development cycle for embedded PCBs, one of the author's power module products was forced to choose a European supplier with a higher price, a longer development cycle, and more complex communication.
Similarly, Navitas' plan to invest in wafer manufacturing is also intended to diversify risks geopolitically. According to CEO Gene Sheridan, "We are building a Chinese supply chain to serve the Chinese market. Other markets are met by Western supply chains (west-for-west), which may include the United States, Europe, or even Asia."
Navitas' current GaN and SiC supply chains (Source: 01 Xinwen, Navitas)
In addition, Navitas' choice to participate in the refurbishment of the fab rather than directly participate in the retrofit of the old fab at this stage is also more suitable in terms of investment. Compared with the greenfield investment of investing hundreds of millions or billions of dollars to build a new fab, there are still many old silicon-based fabs available in the United States, and the investment required is only a fraction of the former. At the same time, many facilities and equipment of the old production line have long been depreciated, and the production cost is relatively low.
A similar example is Bosch's acquisition of TSI Semiconductors in California this year. The company plans to transform the fab, located near Sacramento, California, into one of Bosch's three global chip manufacturing pillars, with silicon carbide wafer production expected to begin in 2026. The main reason for this acquisition is not only the relatively small total investment, but more importantly, it will help Bosch to establish production capacity in North America more quickly.
Enlightenment for domestic production
There are also a number of domestic third-generation semiconductor manufacturers in China that have taken similar actions. For example, Zhanxin, Basic and Yangjie were all fabless models in the early stage of their startups, but now they are all building or have built their own wafer production lines. In addition to the fact that these new production lines are greenfield investments, which are different from Navitas' brownfield investment plans, the biggest difference is that Navitas' production layout goal is risk diversification, while the practice of domestic silicon carbide manufacturers brings risk concentration.
Not only are the wheels constantly being reinvented, but the existing and under construction capacity may be far beyond what can be absorbed locally (Shortage or Excess, Survive or Destroy?). At the same time, there are only a few overseas production capacities that can be relied on, such as X-FAB and Hanlei, and it is difficult for Chinese manufacturers to exert direct influence. This means that in the future, when domestic original factories go overseas, they may have no choice but to face the requirements of foreign customers for non-Chinese supply chains and the need to provide capacity guarantees.
To a certain extent, whether Chinese chip manufacturers can establish a controllable overseas wafer manufacturing layout may become a key part of whether Chinese chip companies can smoothly expand overseas markets and eventually develop into international companies.
brief summary
Taking Navitas, one of the most successful small companies in recent years, as an example, this paper analyzes the observations and reasons for the development of third-generation semiconductor companies from fabless to semi-fabless and then to fab-lite, and further discusses its significance for domestic manufacturers.