Samsung packaging, too difficult.

In early 2022, the ambitious Samsung planned to make a big impact in the contract manufacturing market. At that time, Samsung, who was favoured by three major clients - Qualcomm, Nvidia, and Tesla, won new orders for Qualcomm's Snapdragon 888 series and Nvidia's Ampere chips, taking away several deals from the leading contract manufacturer TSMC. Even after TSMC announced a slight price increase, Samsung also raised its wafer prices by 20%. However, Samsung seemed to be aware of its weaknesses compared to TSMC in contract manufacturing. Apart from the advanced front-end processes, there was also a significant gap in advanced packaging in the back-end, especially in fan-out packaging. While TSMC held over half of the market share in fan-out packaging, Samsung only had 0.7%, not to mention the popular CoWoS 2.5D packaging. Samsung's approach was to reform and recruit talent. At the end of 2021, under the leadership of Samsung's DS division president Kyung Kye-hyun, a special task force for advanced packaging commercialization (TF) was established. The advanced packaging business team was upgraded, and Samsung hired Lee Sang-hoon, former Vice President responsible for EUV technology research at Intel, and Kim Woo-pyung, who was appointed head of the US Packaging Solutions Center from Apple. However, the most eye-catching move was in March 2022, when Samsung hired former TSMC packaging executive Lin Jun-cheng as the vice president of the semiconductor department's advanced packaging business team, to lead the development of advanced packaging technology within the team. Lin Jun-cheng has an impressive background in the packaging industry. During his time as Deputy Director of R&D at TSMC from 1999 to 2017, he applied for over 450 US patents and made significant contributions to the development of TSMC's CoWoS and InFO-PoP packaging technologies. He also successfully secured a major deal with Apple for TSMC. After leaving TSMC, he joined Micron and helped establish a 3D IC advanced packaging development pipeline, and then worked for the equipment company Tenjeong Technology, contributing significantly to the company's transformation and development in the packaging equipment industry. For Samsung, Lin Jun-cheng's joining was just one part of a large plan. The South Korean company, known for its memory products, aims to compete with TSMC in the packaging market and achieve a "corner overtaking" strategy. Samsung has been a formidable competitor to TSMC and founder Chang Zhengmou, boasting the advantage of being the only company in the world that owns all memory, logic foundry, and packaging businesses under one roof. Although Samsung has never truly surpassed TSMC in wafer manufacturing, earlier on, Samsung demonstrated its advantage in packaging due to its accumulation of relevant technologies in the memory field. As early as 2005, Samsung announced it would introduce eight-layer stacked wafer packaging technology, and in 2010, the iPhone 4 confirmed the usage of this technology. When examining the internals of the 32GB version of the iPhone 4, a flash memory chip from Samsung, model K9PFG08U5M, was found. This chip, decoded from the product number, is a 256Gb MLC (multi-level cell) NAND flash memory device. Upon disassembling the chip from the circuit board and conducting X-ray inspections, it revealed an eight-layer stacked wafer. The total thickness of the package (including the substrate) was about 0.93mm, with the wafer stacking part approximately 670m thick. The thickness of each wafer ranged from 55 to 70m, with the thickest wafer at the bottom. Compared to the 1.4mm thickness announced in 2005, although it did not reach the thinnest, it was still impressive. A closer examination of the cross-section revealed that the distance between the top wire bond ring and the top surface was very close, with a wire diameter of 25m and a distance of less than 10m from the packaging surface. This meant that the packaging material had been compressed to the utmost at that time. From today's perspective, the success of smartphones, including the iPhone, not only benefited from process development but also relied on advancements in packaging technology, with the example above demonstrating Samsung's packaging technology in the NAND product. What truly brought Samsung's packaging into the spotlight was the processors used in the iPhone, which both used the same Samsung packaging technology - PoP (Package-on-Package) packaging. PoP is now commonplace, but in the confines of a smartphone, the ability to integrate DRAM and SoC with PoP is an essential technology. Compared to traditional packaging, PoP packaging requires less substrate, reducing the circuit board's size and weight. Additionally, the shorter interconnection of the DRAM achieves faster data transfer rates, balancing cost savings with improved performance. This technology allowed Samsung to easily win orders for the A6 and A7 processors, causing a moment of doubt for TSMC and Chang Zhengmou. In 2011, TSMC lost the competition for the A6 processor order, with the reasons for this failure being debated. The prevailing explanation at that time was due to TSMC's unstable process technology. A Samsung executive told South Korean media that TSMC's process was unstable, and an industry insider relayed information from Apple suggesting that the instability was not in the front-end logic circuit process as commonly believed, but in the back-end 3D packaging, which was the main reason for TSMC's defeat. "TSMC tried for two months without success; this is a big blow to Chang Zhengmou," the industry insider said. At that time, TSMC and Samsung were already at odds. During a conversation with reporters, when asked about Samsung being considered a respectable competitor, Chang Zhengmou responded...Before the conversation was finished, Zhang Zhongmou immediately retorted, "I didn't say Samsung is a 'respectable' opponent, I said, Samsung is a 'formidable' opponent!"Next, the journalist mentioned again, Samsung is a formidable opponent, the usually steady Zhang Zhongmou, at this moment somewhat displeased, said, "I said they are fearsome opponents, not respectable opponents, fearsome in English is formidable. Please do not get it wrong." Later on, the journalist mentioned a third time, and Zhang Zhongmou said very unhappily, "I emphasize again and again, Samsung is fearsome, not respectable, are you a spy sent by Samsung?" A small scenario fully showed the tense relationship between TSMC and Samsung. In order to combat Samsung's challenge and secure the crucial alliance and customer Apple, TSMC began to heavily invest in advanced packaging. On October 26, 2011, during TSMC's third-quarter financial conference, Zhang Zhongmou unexpectedly announced a commercial model, the so-called CoWoS (Chip On Wafer On Substrate), which places logic chips and DRAM on an interposer layer, and then packaging them onto a substrate. CoWoS is essentially a simplified version of 3D packaging technology, commonly known as 2.5D. Zhang Zhongmou mentioned, "With this technology, our business model will provide a full range of services, we intend to make the whole chip!" This announcement caused a stir in the Taiwanese semiconductor industry. Former partners suddenly turned into competitors, as the leading foundry not only took over most of the wafer manufacturing market but now wanted to enter the packaging industry. In contrast, in South Korea, this news did not attract much attention because Korean companies focused on dominating the storage chip market that comprised more than 10% of exports, leading to less emphasis on downstream departments, including packaging. Samsung and Hynix focused more on reducing the semiconductor's circuit linewidth rather than seeing packaging and other backend processes as low-technology. In 2012 when TSMC entered advanced packaging, Samsung decided to halt new investments in backend facilities outside of core technology, and outsourced the growing backend facilities to other companies. According to Korean media data, Korea's sales share in the packaging market was only 3% in 2012, demonstrating the Korean semiconductor companies' disregard for packaging. After five years of preparation, TSMC finally launched the wafer-level packaging CoWoS and inFO-PoP, with the latter becoming the key to TSMC securing all orders for the A10 processor. TSMC's advanced packaging technology helped Apple's self-developed processors reach a new stage. System Plus Consulting found that the A10 processor's thickness was less than 0.3mm and had better memory integration capabilities. "The 3D components developed by TSMC and Apple significantly reduce thickness by 30%, compared to traditional PoP packaging systems like Samsung's Exynos 8 processor or Qualcomm Snapdragon 820 processor using MCeP and copper ball technology from Shinko." Stphane Elisabeth, an RF and advanced packaging cost engineer at System Plus Consulting, commented. Yole pointed out that TSMC fully utilized inFO technology as an advantage on the WLP platform, eliminating expensive manufacturing steps, leading to a significant cost reduction for Apple's A10 processor. With a significant investment in FOWLP and inFO technology, TSMC's WLP landscape changed, allowing complex PoP integration to be achieved through mass production on the fan-out packaging platform. While TSMC proudly showcased its inFO and CoWoS packaging technologies, Samsung struggled because it suddenly realized it could not produce a corresponding advanced packaging technology like TSMC. Just like advanced processes, packaging technology also requires careful planning, with the need for patent processes on one hand and the reintroduction of production lines on the other. It is not something that can be built in a day. In early 2017, although Samsung won over a big client, Qualcomm, and secured a large order for the Snapdragon 835, it faced challenges meeting Qualcomm's FOWLP packaging demands and had to resort to outsourcing. The backend packaging for the Snapdragon 835 was eventually handled by STATS ChipPAC and Amkor Technology. At the time, Korean media reported that Samsung Electronics and Samsung Motors were developing Fan-Out Panel Level Packaging (FOPLP) technology, which was more efficient than FOWLP, but the production yield was not yet mature. It was not until the end of 2018 with the Exynos 9110 that Samsung introduced its first generation of FOPLP. However, this more advanced technology did not gain much traction in the smartphone market. Although in 2019 Samsung Electronics acquired PLP business from Samsung Motors for 785 billion Korean won (approximately $5.81 billion), it mainly applied in smart wearables represented by the Galaxy Watch, making much less profit compared to the popular FOWLP. By extensively outsourcing packaging and neglecting the backend processes, Samsung lost Apple's orders and was defeated in the competition with TSMC. A lost Samsung Many might wonder, setting aside the higher threshold CoWoS, why did Samsung choose to abandon the relatively simpler FOWLP and tackle the tougher FOPLP? Samsung also had its hardships. Korean media pointed out that since companies like Infineon, Qualcomm (including NXP and Freescale), the inventor of eWLB technology Epic Technologies, and TSMC held core patents in this field, Samsung needed to seek R&D and government support to avoid these patents. "In the field of FOWLP,...""Evading the patents of IP technology is not easy, considering our relative lag in technological development, government support will be necessary." Dr. Choi Kwang-seong of the Electronics and Telecommunications Research Institute (ETRI) said in an interview at the "2016 Next Generation Semiconductor Intellectual Property Forum" hosted by the Patent Office and the Next Generation Semiconductor Research Association.He pointed out, "It is not easy to implement FOWLP in active devices, but TSMC has realized the importance of backend processes and has actively invested in it. To avoid competition from patents, they have adopted a face-up approach rather than a face-down approach. It is expected that Samsung will also adopt a similar approach, the key is how to achieve differentiation." The patent technology mentioned by Choi Kwang-sung is one aspect, but beyond patents, the market is also a crucial factor. Looking back, why is TSMC boldly investing in advanced packaging, and why is Samsung outsourcing advanced packaging? The core reason is customers. As long as customers are willing to pay, TSMC's advanced packaging production line will always have high utilization rates. In contrast, Samsung has faced challenges in its wafer foundry department and hastily invested in production lines, which may be difficult to recover costs from. In conclusion, it is not that Samsung does not see the importance of FOWLP, but rather that the customer base is limited, making it difficult to establish a foothold in the market. As early as 2017, Samsung poached packaging expert Oh Kyung-seok from Intel to accelerate the development of WLP technology, and in 2018, they planned to convert Samsung's panel factory in TIAN AN, Korea into an encapsulation factory to mass-produce 2.5D and fan-out packaging. However, it wasn't until the Exynos 2400 processor in 2024 that we saw Samsung's own FOWLP technology. The underlying reason is not hard to guess - internal resistance has prevented advanced packaging from being properly recognized. In 2022, Samsung planned to invest approximately 200 billion Korean won (about $1.65 billion) in a FOWLP production line at the TIAN AN semiconductor wafer plant, but the decision was doubted by many executives who argued that there were no "key customers" and demand could not be ensured. Even if the FOWLP production line was established, it might not be fully utilized. However, Samsung cannot avoid making changes because their once proud turnkey service, which allowed them to produce chips under one roof, has been overtaken by TSMC. Without a more significant response to advanced packaging, relying solely on advanced processes is no longer enough to attract customers from TSMC. Although progress in WLP has been slow, Samsung has established a relatively complete solution for packaging outsourcing since 2018, developing three advanced packaging technologies in horizontal integration and vertical integration directions: I-Cube, H-Cube, and X-Cube. According to Samsung, I-Cube is a 2.5D packaging solution in which chips are placed side by side on an intermediate layer. To enhance computing performance, customers of I-Cube typically request an increase in the intermediate layer area. In response, Samsung has introduced two I-Cube solutions: I-CubeS and I-CubeE. H-Cube, short for Hybrid Cube, is another 2.5D packaging solution introduced by Samsung. This solution aims to address the severe shortage of printed circuit boards (PCBs) in the semiconductor industry. X-Cube is a full 3D packaging solution that uses chip vertical stacking technology. It connects two vertically stacked bare dies using microbumps or more advanced copper bonding technology. Samsung began mass producing microbump type X-Cube products in 2024 and copper bonding type X-Cube products in 2026. Unfortunately, the window of opportunity for Samsung to develop advanced packaging between 2022 and 2024 is limited. According to TSMC's financial reports, their total revenue from advanced packaging in 2023 exceeded $6 billion, ranking second in the global outsourced packaging market. If we only look at advanced packaging, TSMC is undoubtedly the No.1. On the other hand, although Samsung has not disclosed their revenue from advanced packaging, during Samsung's annual shareholder meeting in March 2024, Samsung Electronics CEO and semiconductor division head Kye Hyun Kyung stated that they expect advanced packaging products to generate over $100 million in revenue in 2024. He mentioned that Samsung's investment will yield results as early as the second half of the year. The $100 million figure, compared to $6 billion, is a stark reminder even to the most confident Samsung employees. What's more unsettling is that the $100 million prediction may have been overestimated, as Kye Hyun Kyung also made a bold assertion to regain the top position in the semiconductor world within 2-3 years. When shareholders pointed out the lag in HBM (High Bandwidth Memory) development, Kye Hyun Kyung responded by saying they were preparing to prevent such incidents from happening again and results would soon be visible. With both of these objectives unmet, one can only wonder how much room for growth there is in the $100 million projected revenue from advanced packaging. For present-day Samsung, their initial investment in advanced packaging was intended to gain a share of the outsourcing market. However, when the investment failed to yield returns, they had no choice but to reevaluate their strategy. At the end of last year, news from the Korean industry revealed that Samsung had begun reorganizing its advanced packaging supply chain to enhance its competitiveness. Aside from reevaluating the existing supply chain, Samsung planned to establish a new supply chain system. Reports suggested that Samsung first focused on equipment, prioritizing performance over existing business relationships or collaborations. They even considered returning purchased equipment that was previously bought for setting up packaging production lines to reassess their performance and suitability. Several sources revealed that Samsung would conduct a comprehensive review from scratch, ultimately aiming to diversify the supply chain, including replacing the existing one. This means that Samsung is going back to its old ways in 2012, focusing less on heavy R&D and building production lines and more on outsourcing more packaging businesses. Just a few days ago, Vice President Lin Jung-cheng mentioned at the beginning of this article left on December 31st due to the expiration of his contract with Samsung Electronics' semiconductor department. His departure was...Perhaps heralding the end of a showdown between competitors.In 2022, the ambitious ambitions that Samsung once made during the industry boom have turned into a dreamlike bubble by 2024. Samsung, where to go The current Samsung gives a feeling of neglecting the fundamentals, with advanced packaging lagging behind and even being unable to enter the top three. Firstly, in terms of HBM, it is still in the process of catching up with Micron until now. HBM3E has not yet received recognition from Nvidia, and then in terms of DRAM, Samsung is currently behind Micron in the progress of the sixth generation 10nm production. Lastly, in terms of NAND, Micron was the first to launch 300-layer NAND, bringing more crisis to Samsung. In the outsourcing market, Samsung is also on a downward trend. Currently, Samsung has secured outsourcing orders from Tesla and Google at low prices, but Tesla's next generation of fully autonomous driving chips will shift to TSMC, and Google has similar plans. For these two customers, Samsung may only be a low-cost experimental ground for verifying chip design feasibility, rather than a long-term outsourcing partner. Under the trend of outsourcing packaging, Samsung's outsourcing business will only become less attractive. The next department that may be restructured is probably the wafer manufacturing department, reminding people of the bend overtaking at 14nm back then, leaving people with a sense of sorrow. This article is reproduced from Semiconductor Industry Observation, edited by GMTEight: Chen Yufeng.