Before 2017, the total capital expenditure of the whole industry did not exceed 70 billion US dollars. In 2017, Samsung gambled on memory, and its semiconductor spending soared to 24.2 billion US dollars, driving memory manufacturers to set off another round of investment frenzy, and finally driving global semiconductor spending in 2017. It hit a new high of $95.6 billion in 2016, $27.8 billion more than in 2016, and semiconductor equipment manufacturers made a lot of money.
On January 4, Taiwanese media reported that TSMC raised its capital expenditure in 2021 to more than $20 billion – possibly to $22 billion. On January 9, according to the Nikkei Asian Review, Samsung Electronics will invest more than $30 billion in the semiconductor business in 2021. If the plan is finally realized, it will be the first time that TSMC’s capital expenditure exceeds 20 billion US dollars, and it is also the first time that Samsung Electronics has invested more than 30 billion US dollars in semiconductors, marking the gold dollar bombing of the semiconductor industry leader to a new level.
What is the concept of 30 billion US dollars? Since the beginning of the 21st century, there have been two years of industry-wide capital expenditures of less than 30 billion US dollars, in 2002 and 2009, respectively. In 2003, the industry-wide expenditures were only 31.3 billion US dollars. Maybe you would say that these figures are too old, but in fact, before 2017, the total capital expenditure of the whole industry did not exceed 70 billion US dollars. In 2017, Samsung bet heavily on memory, and its semiconductor spending soared to 24.2 billion US dollars, driving memory manufacturers Set off another round of investment frenzy, and finally drove global semiconductor spending to a new high of 95.6 billion US dollars in 2017, 27.8 billion US dollars more than in 2016, and semiconductor equipment manufacturers made a lot of money.
Another set of data is also interesting. From 2017 to 2019, the capital expenditure of all semiconductor manufacturers in China was US$30.8 billion, and the total value of chips produced by semiconductor companies in China (including foreign companies) in 2019 was US$22.7 billion. That is to say, Samsung’s one-year investment is the sum of the three-year investment of Chinese manufacturers, and the sum of all the chip sales produced by China in one year is not enough for Samsung’s one-year capital expenditure. The total amount of the second phase of the big fund is 29 billion US dollars, less than 30 billion US dollars.
The crazy investment of Samsung and TSMC may also allow them to open the gap with Intel in terms of manufacturing process. Let’s first look at the capital expenditures of the three companies in 2019. Samsung was the highest with $19.2 billion, followed by Intel with $16.2 billion in capital expenditures, and TSMC third with $14.9 billion. In 2020, Samsung spent $26.5 billion on semiconductors, TSMC also spent $17 billion, and Intel’s data has not yet come out, but the first three quarters were down year-on-year compared to 2019, and the total spending in the three quarters was $10.392 billion. Unsurprisingly, full-year 2020 capital expenditures should be within $15 billion.
Before 2017, these three companies were generally close in semiconductor spending. In 2017, Samsung began to sprint, but Samsung’s investment included a large amount of memory equipment investment. In terms of pure logic process investment, Samsung and TSMC did not lead Intel by much. However, in the past two years, Intel’s process development has encountered difficulties. From time to time, there has been news that Intel will increase production outsourcing. If the outsourcing ratio is really increased, it indicates that at least 2021, Intel’s capital expenditure may be less than 15 billion US dollars, thus For the first time, TSMC’s capital expenditure was pulled above $5 billion.
This is a very dangerous signal for Intel. The competition of advanced technology has always been a knockout, and once the investment cannot keep up, it will be eliminated from the market. There are nearly 30 players in the 130nm process node worldwide, and only 10 players remain at 28nm. The rest of the manufacturers give up advanced process research because they cannot keep up with such a high-intensity investment rhythm. Announced to abandon advanced technology research and development, and now only four manufacturers are willing to invest in advanced technology, namely TSMC, Samsung, Intel and SMIC.
At present, Intel, which is mainly based on the IDM model, has encountered difficulties in advanced technology investment. TSMC and SMIC are pure foundry models, and the cost can be amortized to multiple customers. In addition to Samsung’s huge memory production line, its foundry scale is second only to TSMC. Intel can only undertake advanced process research and development by itself. Now mass production is blocked, and there is no rice in the pot. The initial investment cannot be recovered, and it does not dare to increase investment only to expand 10nm production capacity. This is extremely embarrassing, and it may be pulled if not invested. There is a gap, but such a high investment has never been effective, even Intel can’t bear it.
From the perspective of Intel’s capital expenditure in 2020, it is not impossible to find TSMC or Samsung for large-scale outsourcing in the future – of course, Intel will retain a considerable degree of self-production ratio, such as more than 50%, once the self-production ratio is too low, then Intel will It will be difficult to maintain investment in advanced technology, and in the end, there is only a way to go to waferless in advanced technology.
Advanced technology will eventually become a game for a few people.