The packaging and testing plant has been unable to keep up with the footsteps of the foundry, and Moore’s Law has begun to be in a hurry. Instead of working in a hurry, we should go outside.” In 2011, TSMC’s Yu Zhenhua said to the media.
Yu Zhenhua is a veteran who joined TSMC in 1994, a former subordinate of Jiang Shangyi, and a hero of TSMC’s post-Moore’s Law era.
TSMC to do advanced packaging
In 2011, Zhang Zhongmou, who returned to the arena, announced at the third-quarter legal conference that TSMC would do advanced packaging. For this reason, Zhang Zhongmou invited Jiang Shangyi, who had retired, to helm the R&D again, and the specific task fell on Yu Zhenhua’s shoulders.
The news that the foundry boss is going to do packaging spread like wildfire, and the entire semiconductor industry was caught in a noisy discussion. Among them, ASE was pushed to the forefront, because it is not only the world’s largest packaging and testing factory, but also TSMC’s supply business.
On the third day after TSMC’s legal seminar, it happened to be ASE’s legal seminar. Dong Hongsi, the chief financial officer of ASE, was repeatedly asked how he thought about TSMC’s entry into the packaging field. Dong Hongsi reluctantly confirmed the truth, and then changed his words. Indicates that this technology will only be used in a very small number of specific high-end products, with limited impact.
This remark angered Yu Zhenhua, who was in charge of advanced packaging research and development. The latter immediately countered: “All high-end products will be used in the future, and the market is huge.” This seems to be a consensus now, but it caused an uproar at the time because it was the first At one time, TSMC and the packaging and testing factory were clearly opposed.
Later, in his speech at SEMICON Taiwan, Yu Zhenhua talked about TSMC’s advanced packaging. A practitioner in the packaging and testing industry interpreted Yu Zhenhua’s speech as: “What he means is: you are all finished, only me is left. TSMC To conquer the world.”
Ma Guanghua, the vice president of R&D of Silicon Products, even asked Yu Zhenhua on the spot during the speech: “You said that, do we all have no jobs?” The atmosphere at the scene froze to freezing point in an instant.
No matter how criticized by the outside world, TSMC’s determination to do advanced packaging remains unchanged. In fact, a few months before the legal meeting, TSMC had already made 3D IC packaging and silicon interposers available to customers in the latest version of the design guidelines. Only when the technology matures in the future, it will be fully replaced. In order to develop CoWoS technology, Zhang Zhongmou specially allocated 400 R&D engineers to Yu Zhenhua.
Sometimes wishes are beautiful, but reality is cruel.Advanced packaging is a big mountain, and conquering technology is only the first step to shake it. Only if customers are willing to use it can be considered a success.
Winning the Tianwang Mountain of Advanced Packaging
After more than two years of hard work by Yu Zhenhua, TSMC finally developed the CoWoS technology, but when it was implemented into the product, only one company was willing to place an order, and this company was Xilinx, and the rest of the customers felt that it was too expensive. Xilinx’s high-end FPGA chips have high unit prices and pursue performance, so they are willing to adopt this technology, while other customers need cost-effective products, especially potential customers Apple.
Image source: Frontiers of Technology
At that time, TSMC wanted to win the order from Apple. When competing for the A6 order, Samsung was superior in 3D IC packaging technology, and finally the A6 order went to Samsung, which may be one of the incentives for TSMC to win the advanced packaging.
Seeing that TSMC is done, butCoWoS package priced 5 times higher than customer expectations, which had to make Zhang Zhongmou feel melancholy. Suddenly one day, Jiang Shangyi rushed into Zhang Zhongmou’s office and told Zhang Zhongmou that Yu Zhenhua had dug a big gold mine.
The reason why Jiang Shangyi, who is nearly seventy years old, is so excited is that Yu Zhenhua has come up with a simplified design, which can simplify the CoWoS structure as much as possible, and reduce the price to one-fifth of the original. This technology is the later InFO technology.
Since then, TSMC’s advanced packaging has been divided into two parts. The more economical InFO packaging technology has become the first choice for mobile phone customers. This packaging technology has also become one of the keys for TSMC to take orders from Apple. The CoWoS packaging technology, which focuses on the high-end customer market, has ushered in its own dawn due to the development of artificial intelligence.
Image credit: Deepmind
In 2017, AlphaGo, a high-tech product that defeated the world’s chess king, used the artificial intelligence chip TPU 2.0, which is TSMC’s CoWoS packaging technology; later, Intel and Facebook wanted to challenge NVIDIA’s monopoly on deep learning, and jointly launched the Nervana-like neural network. The network processor also uses the CoWoS package.
Unconsciously,CoWoS encapsulation seems to have become the norm for AI.
Post Moore’s Law
The time has come to 2020, TSMC’s 5nm has achieved mass production, and the limit of Moore’s Law mentioned by Yu Zhenhua ten years ago is becoming a reality. According to a survey by the Center for Security and Emerging Technologies (CSET) of Georgetown University, in the past three generations of chips, the average price of TSMC’s die has fallen sharply. $5 more in a generation, Moore’s Law failure is just around the corner.
Advanced packaging has become a powerful tool to improve chip computing power and reduce the average price of chips. Liu Deyin, chairman of TSMC, publicly stated last year (2020) that microsizing and 3D IC technology will allow the computing density per unit to double every two years in the future.
After nearly a decade of development, TSMC is one of the major leaders in high-end high-density fan-out wafer-level packaging. In order to maintain its leading position in the market, TSMC announced that in 2020, capital expenditure will be US$17 billion, of which 10% will be used for advanced packaging. On January 14, TSMC announced that its capital expenditure in 2021 will reach 25 billion to 28 billion US dollars, of which about 10% will be used for advanced packaging technology mass production requirements.
Under the huge capital investment, a lot of packaging and testing factories have been allocated.According to media reports, TSMC started to build a new high-end IC packaging and testing factory in Miaoli County, Taiwan Province, China last year, and is expected to be completed in May this year. In addition, there has been news recently that TSMC plans to build an advanced packaging and testing plant in Japan, which is expected to be completed in 2025.
The packaging and testing plant is surging
TSMC has unique conditions for packaging. If TSMC fails to make advanced packaging, it only needs to produce another wafer for the customer. It has a large chance for trial and error and a high fault tolerance rate, but the packaging and testing factory is different. If a wafer is damaged by the packaging and testing factory, it must compensate according to the market price. A slice would cost tens of thousands of dollars.
But in any case, traditional packaging has reached an inflection point, and advanced packaging represents the future. In order to attack TSMC and to occupy a place in 3D IC packaging, starting in 2012, the two heroes of packaging and testing, ASE and Sipin, have deployed advanced packaging and testing.
Image source: ASE official website
Sipin applied for 5 hectares of land in Zhongke, which is planned to be the first manufacturing land for 3D IC packaging and advanced packaging such as copper pillar bumps. ASE launched its investment plan to return to Taiwan. It invested 700 million US dollars to expand the second phase of the new factory in Kaohsiung Nanzi Processing Zone, focusing on the advanced flip chip packaging, chip planting bumps and 3D IC required for high-end mobile phone chips. Packaging and testing capacity.
In terms of technology, ASE is more optimistic about the SiP (system-level wind installation) business opportunities brought about by the trend of heterogeneous integration of chips, and has achieved fruitful results through continuous overweight. In 2019, ASE’s total revenue was 2.5 billion US dollars, of which SiP contributed At $230 million, revenue from SiP exceeded expectations by $130 million.
The silicon products are more tortuous.
In 2015, ASE publicly announced that it would acquire 25% of Sipin at a premium of 34%, and the acquisition amount was as high as NT$35 billion, which meant that the two major packaging and testing giants entered the integration period. Agreed on a wholly-owned acquisition. Later, in 2020, the acquisition was still being reviewed and approved, and Sipin sold its mainland branch to a mysterious big factory.
SMIC forms alliance with Changjiang Electric Co., Ltd.
TSMC’s packaging road began at the end of 2008, when TSMC established a wire and packaging technology integration department. It is said that Zhang Zhongmou initially selected Liang Mengsong to study advanced packaging in order to surpass the limits of Moore’s Law. However, compared with advanced manufacturing process, packaging is a cold bench, and it will not make money in the short term, so Liang Mengsong left, and this is the combination of Jiang Shangyi and Yu Zhenhua.
Different from Liang Mengsong, Jiang Shangyi believes that when the integrated circuit is at its extreme, the whole system should be studied in turn. As the progress of advanced manufacturing processes becomes slower and slower in the future, packaging and circuit boards will become the bottleneck restricting the entire system, so Advanced packaging is imperative. With advanced packaging, the architecture of the entire system can be completely changed. Engineers no longer pursue making chips smaller and smaller, but divide large chips into small chips and reassemble them.
In 2017, Liang Mengsong joined SMIC. In the following two years, he brought advanced manufacturing processes to SMIC. However, in 2020, the international situation is turbulent. Can the advanced manufacturing process continue to move forward and become one question.
At this time, Jiang Shangyi returned to SMIC, and he could give SMIC technical support in the field of advanced packaging technology and small chips in addition to advanced manufacturing processes, which would also help SMIC get out of the predicament.
In fact, SMIC also sees the prospect of advanced packaging. In 2016, SMIC, through its wholly-owned subsidiary Chip Power Shanghai, invested in Changdian Technology at a price of RMB 2.655 billion, plus the US$100 million equity in the previous acquisition of Xingke Jinpeng was converted into Changdian Technology. Equity, SMIC became the single largest shareholder of Changjiang Electronics Technology. Later, with the large fund’s shareholding in 2018, SMIC made it possible.
Image source: Founder Securities
The strategic cooperation between SMIC and Changdian Technology, from the perspective of the industrial chain, can allow domestic chips to avoid supply chain risks as much as possible and accelerate domestic substitution. In terms of technology, SMIC and Changdian Technology are the industry leaders of Foundry and OSAT. It is possible to better cooperate, so that advanced manufacturing and advanced packaging and testing can cooperate more closely with each other, so as to break through the technical limit of domestic cores.
In a hurry, time has entered the third decade of the 21st century. In the past ten years, TSMC led the global semiconductor industry to the limit of Moore’s Law, but it also realized the limitations of Moore’s Law, so it has sharpened its sword in ten years, starting with advanced packaging, and wants to challenge the limit of computing power again.
And the packaging and testing factories are not to be outdone, they are constantly innovating in technology, and the market is constantly conquering cities. Mainland companies also dare not slack off. It remains to be seen what the next decade of the semiconductor industry will be.
The Links: G190EG01 V0 LTM190E4-L02