TSMC's new 28nm plant has made significant progress-EEWORLD

Collect

The Public Works Bureau of the Kaohsiung Municipal Government in Taiwan recently approved the construction permit for TSMC's Kaohsiung plant. TSMC stated that the Kaohsiung plant will start construction as scheduled this year and begin mass production in 2024.

Taiwan's Public Works Bureau issued a press release today, stating that TSMC obtained the miscellaneous licenses for earth retaining safety measures and pile foundation works on June 10 this year after obtaining approval for land use rights, urban plan changes, environmental impact assessments, and geological sensitivity assessments, and applied to start construction on September 19. As for the factory building application, after obtaining approval for the traffic impact assessment, the factory building construction license was approved today.

TSMC pointed out that the Kaohsiung plant will proceed according to the plan and is expected to start construction this year. It is currently preparing for land preparation and thanks the relevant government agencies for their assistance. TSMC's Kaohsiung plant is expected to start mass production in 2024, mainly using 7nm and 28nm processes.

TSMC's Kaohsiung plant is located in the Nanzih Industrial Park, the former site of CPC Kaohsiung Refinery. The Kaohsiung City Government hopes to combine the Qiaotou Science Park, Lukut Science Park and Nankut Science Park in the future to create a semiconductor industry cluster.

TSMC's two factories in Kaohsiung are said to be tailor-made for Apple

TSMC (2330), the world's leading foundry, has confirmed that it will set up a plant in Kaohsiung. There have been many reports in the market that the plant may delay mass production, but TSMC officials have repeatedly emphasized that the timing and plan for mass production at the Kaohsiung plant remain unchanged. In addition, it is widely rumored in the industry that the plant was established mainly for Apple, a major American mobile phone manufacturer. Recently, it was reported that TSMC's Kaohsiung plant will break ground on August 7, and TSMC officials clarified that it was invited to participate in the "Nanzi Industrial Park Groundbreaking Ceremony" on that day, not the groundbreaking ceremony of its own Kaohsiung plant. As for the groundbreaking schedule of its own Kaohsiung plant, it will be later, but the progress and plan of breaking ground in 2022 and mass production in 2024 remain unchanged.

TSMC officials also revealed that the Kaohsiung plant will be engaged in 7nm and 28nm process foundry: According to the equipment circle, TSMC's 7nm and 28nm plants in Kaohsiung are mainly established for Apple.

Apple has been developing its own 5G modem chip for some time. It is expected that the chip will be officially mass-produced from 2025 and built into Apple's mobile phones and other products. The RF chip that goes with it will be produced by TSMC's 7nm process in Kaohsiung. As for peripheral chips such as amplifiers in the front-end RF module, they will be produced by another 28nm plant in TSMC's Kaohsiung.

As for the recent unsatisfactory market conditions, there have been many reports that TSMC's production expansion progress will be delayed, including the Kaohsiung plant. However, industry sources say that the factory buildings will be completed first. As for the time for internal equipment to be shipped, whether it will be delayed depends on the equipment vendors' delivery status and market conditions. But for TSMC, as long as it can cooperate with customers' mass production time, it will be fine, especially since the two Kaohsiung plants are dedicated factories set up for Apple.

Apple and Qualcomm engaged in a legal battle over patent licensing fees in 2016. The two sides reached a settlement in 2019 and signed a six-year licensing agreement. Although Apple still purchases 5G modem chips from Qualcomm, it has already invested in its own research and development of 5G modem chips. As for the development and mass production of this chip, although the progress has been repeatedly delayed, it will be produced by 2025 at the latest and built into its own new mobile phones of the year. TSMC, as the foundry of this chip and peripheral chips, will benefit the most.

TSMC's 28nm chips still have considerable profits

According to a report by Strategy Analytics, after ten years, TSMC's 28nm chips still have considerable profits, and TSMC, UMC and SMIC are further expanding their 28nm production capacity.

In 2021, the total revenue of 28nm wafer foundry exceeded US$7.2 billion, with TSMC accounting for approximately three-quarters of the revenue share.

The report pointed out that TSMC is still investing in 28nm and will set up a 7nm and 28nm process wafer fab in Kaohsiung, which is expected to start construction this year and start mass production in 2024. In addition, TSMC will establish a subsidiary, Japan Advanced Semiconductor Manufacturing Co., Ltd. (JASM), to provide foundry services in Kumamoto City, Japan, initially providing initial technology for 22/28nm process; JASM's wafer fab in Japan is scheduled to start construction this year and start production at the end of the year.

Recently, TSMC released its financial report and reiterated that the company's 3-nanometer (N3) chip will be put into production in the second half of this year and contribute revenue in the first half of next year; TSMC reiterated that the 2-nanometer chip (N2) will be mass-produced in 2025.

In the second quarter, TSMC's 5nm process wafer shipments accounted for 21% of the company's revenue, and 7nm accounted for 30%; overall, TSMC's advanced process (7nm and more advanced process) revenue accounted for 51%, which continued to expand compared with 50% in the previous quarter.

TSMC calls on customers to abandon 40nm and switch to 28nm

When it comes to chip manufacturing, we talk more about leading-edge nodes and the most advanced chips made using them, but the fact is that there are thousands of chip designs that were developed many years ago and are made using mature process technologies that are still widely used by the industry. In terms of execution, these chips are still as perfect as when the first chip was made, which is why product manufacturers are increasingly using them for manufacturing.

But on the manufacturing side, there’s a hard bottleneck for further growth of these chips: All the capacity at older nodes has already been built — and they’re not going to be built anymore. As a result, TSMC recently began strongly encouraging customers on its oldest (and least dense) nodes to migrate some of their mature designs to its 28nm-class process technology.

Today, about 25% of TSMC’s revenue comes from manufacturing hundreds of millions of chips using nodes 40nm and older. For other foundries, the revenue share of mature process technologies is higher: 80% of UMC’s revenue comes from nodes 40nm and higher, while 81.4% of SMIC’s revenue comes from obsolete processes. Mature nodes are cheap, have high yields, and provide adequate performance for simple devices such as power management ICs (PMICs). But the cheap wafer prices for these nodes come from the fact that they were leading nodes themselves long ago, and their construction costs were paid for by the high prices that leading-edge processes can fetch. That is, there is no profit driver (not even the equipment) to build new capacity for these older nodes.

That’s why TSMC plans to expand production capacity at mature and specialty nodes by 50%, with a focus on fabs with 28nm capabilities. As the last (viable) generation of TSMC’s classic pre-FinFET manufacturing process, 28nm is positioned as the new sweet spot for producing simple, low-cost chips. And, in order to consolidate the production of these chips into fewer and more widely available/scalable production lines, TSMC wants to keep customers using older nodes on 28nm.

“We don’t have the capacity to [expand] the 40nm node right now,” said Kevin Zhang, senior vice president of business development at TSMC. “You build a fab and the fab isn’t going to be online two or three years from now. So, you really need to think about where the products of the future are going, not what the products of today are.”

While TSMC’s 28nm nodes are still subject to the same general cost trends as fabs in general — because they are more complex and more expensive on a per-wafer basis than older nodes — TSMC hopes to convert customers to 28nm in a way that balances this with the greater number of chips per wafer that the smaller node offers. So while companies will have to pay more, they will also get more total chips. And none of this takes into account potential ancillary benefits of the new node, such as reduced power consumption and potentially greater clock speed (performance) headroom.

“So, a lot of customers’ products today are at 40 nm or even earlier, 65 nm,” Zhang said. “They’re moving to more advanced advanced nodes. 20/28 nm is going to be a very important node to support future majors. We’re working with customers to accelerate [their transition]. [...] I think customers will get benefits, economic benefits, scaling benefits, you’ll get better power. But they already have a chip that works. Why do that? Because going to the next node, you get better performance and better power, and overall you get system-level benefits.”

In addition to multiple 28nm nodes designed for various customer applications, TSMC is also expanding its lineup of specialized 28nm and 22nm (22ULP, 22ULL) process technologies to address a variety of chip types that currently rely on various obsolete technologies. As with the overall move to 28nm, TSMC is looking to attract customers to newer, higher-density process nodes. And if it's not 28nm/22nm, then customers can also choose to transition to more powerful FinFET-based nodes that are part of TSMC's N16/N12 series (for example, N12e for IoT).

TSMC to expand production capacity of mature and specialty nodes by 50%

TSMC revealed at a technology summit held earlier that by 2025, its mature and professional node production capacity will expand by about 50%. The plan includes building a large number of new wafer fabs in Taiwan, Japan and mainland China. This move will further intensify the competition between TSMC and chip foundries such as GlobalFoundries, UMC and SMIC.