Boss Semiconductor is one of the AI fabless companies targeting the artificial intelligence (AI) market, similar to the newly emerged companies like Furiosa AI, DeepX, and Mobilint. The difference is that while other fabless companies were founded by engineers in their 30s and 40s, Boss Semiconductor was established by veteran engineers with over 15 to 20 years of experience, including CEO Jea-Hong Park, who previously served as vice president of the System LSI and Foundry division at Samsung Electronics after working at IBM.
In particular, CEO Park has worked for 23 years in the semiconductor (DS) division of Samsung Electronics, where he contributed as an elite designer. He has previously developed mobile application processors (AP) for the initial iPhone and chips for Tesla's autonomous driving. He also has extensive collaboration experience with global companies like Audi. Recognizing Boss Semiconductor's potential, Hyundai Motor has made investments on two occasions.
The trust and partnership with Jim Keller, the renowned 'semiconductor legend' and CEO of TensTorent, also demonstrates Boss Semiconductor's strength. Park noted in a meeting with ChosunBiz at the company's office in Pangyo, Seongnam, on the 28th of last month that "there is a trust relationship built through various collaborations with CEO Keller, who led chip designs at Tesla and Intel," and explained, "Boss Semiconductor signed a cooperation agreement with TensTorent early on and decided to mount TensTorent's neural processing unit (NPU) on the vehicle AI accelerator.
Among the more than 200 research and development (R&D) personnel at Boss Semiconductor, many are veterans who have worked for years in Samsung semiconductors. Park stated, "While there are cases where a good idea is enough to start AI semiconductor design, understanding of the semiconductor industry is required in the process of realizing it," and added, "Boss Semiconductor is composed of personnel with long careers and practical experience gained at Samsung semiconductors.
This month, Boss Semiconductor, marking its third year since establishment, has already achieved significant results. The main product, the vehicle AI accelerator 'N1', has completed development and is expected to secure samples in Samsung's foundry 5-nanometer process within this year, aiming to begin mass production in the second half of next year. Recently, they signed a joint technology development contract with Intel, which is competing against Qualcomm in the automotive semiconductor market. The plan is to integrate Boss Semiconductor and Intel chips in the advanced driver assistance systems (ADAS) of major global automotive companies. Below is a Q&A with CEO Park.
─You decided to start a business after a successful career at Samsung Electronics.
"The semiconductor industry in our country has not developed relative to its environment. Although there is a second-largest foundry company in the world and several electronic and IT corporations like Samsung and LG that could be customers for fabless companies, there were almost no 'proper' fabless companies. I wanted to use the design capabilities I accumulated at Samsung Electronics to advance the semiconductor industry in our country. I wanted to leave some legacy so that my juniors could continue and establish a proper fabless company that can compete on the global stage.
Recently, not only the United States but also countries like China and Taiwan have seen significant development in the fabless industry. This is due to having many talented individuals and active government investment and incentives. Since the fabless industry ultimately provides key components for the system industry, if fabless companies are not properly formed, the system industry becomes reliant on foreign countries.
─What is important in creating a proper fabless company?
"Talent, know-how, and investment are all important. This applies not only to fabless companies but to other businesses as well, but it starts with having customers. There should be customers who will use the products we create, and the cycle of us producing the products and receiving market feedback should run about three times to establish a proper market position. In the initial stages, there may be negative feedback, but having a sense of mission and incentives for domestic products among customers who can support us is essential.
The government has been investing to nurture the fabless industry because it has not performed well so far. There are national projects and direct support as well. The important thing is that corporations tend to choose stability over risk and select proven products like Qualcomm instead of domestic products. This creates a vicious cycle. Someone must break this vicious cycle. Government investment in the first phase would be helpful. Support measures are needed for the fabless industry to take root, such as providing subsidies when using domestic fabless chips like in China.
─The preference of global IT companies for large fabless companies is driven by the convenience of providing the software and hardware platforms themselves.
"The current strategy is to diversify the business like Broadcom. Broadcom operates as a fabless company while also providing ASIC (custom semiconductors) designs. Entering the automotive semiconductor market with our products alone will take time. It will take more than five years to generate revenue. It is not easy to endure. Therefore, I want to create a virtuous cycle by initially generating revenue through custom semiconductor business and growing the company's capabilities while launching our automotive semiconductor products.
─Is there a specific reason for targeting automotive semiconductors as your first product?
"While the influence of establishing a cooperative relationship with Hyundai Motor at the time of starting the business played a role, there was also a shortage of semiconductor supplies during the COVID-19 pandemic, leading to a situation where vehicles could not be assembled. The automotive industry may not understand semiconductors well, but there was a strong demand for some measures to be taken. In contrast, there was not as much demand in the mobile sector. Ultimately, I started with automotive semiconductors depending on market demand. However, I am not solely targeting automotive semiconductors; I want to create a foundation to expand into various fields, including automotive, data centers, and mobile, like Qualcomm and MediaTek.
─I heard that Boss Semiconductor's research and development staff are veterans.
"Most of them are core staff who have worked in Samsung Electronics' system semiconductor sector for about 15 to 20 years. Although Samsung Electronics' system semiconductors are evaluated to have lost competitiveness for various reasons, the technical capability of human resources is not behind anywhere in the world. People who have gained extensive experience in such large semiconductor companies are at the center of Boss Semiconductor. The qualitative level of technical personnel can be seen as a differentiating factor for Boss Semiconductor.
─Are there important principles for survival in the global competition?
"Many who have done business in semiconductors often start AI semiconductor businesses with good ideas. However, they may lack some understanding of the semiconductor industry. The semiconductor business is not just about having a good idea; it must lead to manufacturing, mass production, and sales, and it's not easy with just an idea. There are many issues related to product viability and quality. To prevent these issues, understanding of the entire industry is crucial.
If small semiconductor companies face quality issues and recalls, they could go bankrupt. They need to manage quality thoroughly while ensuring product viability and consider ways to lower costs. Proper planning of products to match customer demand is also essential. Since Boss Semiconductor is supported by people who have experienced the semiconductor industry cycle, it has the trust of investors. Especially in the high-end semiconductor field, Boss Semiconductor is the only domestic fabless company composed of personnel with know-how.
─Qualcomm holds a solid position in the automotive semiconductor market.
"Major automobile companies are concerned about being dependent on Qualcomm. Qualcomm's main customers are not all of the global automotive industry but a few large corporations. The demand for multi-vendor solutions against Qualcomm's monopoly will continue. Since Qualcomm's resources and engineers are limited, it is difficult to meet the demand of all automobile companies. In order to avoid the situation where smartphone companies like Samsung Electronics and other competitors use Qualcomm while bearing high price burdens in the mobile AP market, automotive companies are also making various attempts. The collaboration of Boss Semiconductor with major automotive OEMs in Europe is already a sign that many companies are trying to avoid dependence on Qualcomm.
─If there is a crucial strategic goal for Boss Semiconductor this year, what is it?
"This year is extremely important for Boss Semiconductor. The first chip sample has been released, and software development is currently underway based on this. It must be successfully developed so that automotive companies can demo and test Boss Semiconductor products. The second chip sample, reflecting modifications made on the first product, is expected to be tape-out (prototype mass production) at Samsung Foundry. This product needs to perform well. In addition to our own chip this year, we must also secure orders for custom semiconductor projects, which I believe will be an essential step in securing the company's future competitiveness.