TSMC 2nm, important information

According to the latest news reports, TSMC will start producing 2nm GAA-based wafers at its Fab 20 plant in Hsinchu in the fourth quarter of 2025. The plant has a production capacity of 30k wpm. Subsequently, Fab 22 in Kaohsiung will also start production with a production capacity of 30k wpm, and production is expected to start in the first quarter of 2026.

N2P will start production at the end of 2026, but will not have the previously announced back-side power delivery. TSMC CEO Wei Zhejia said that the planned capacity of the 2nm process has exceeded the 3nm process. It is reported that the cost of 2nm wafers will reach more than $30,000. Both N2 and N2P will use TSMC's NanoFlex technology, which allows chip designers to mix and match units from different libraries (high performance, low power consumption, area efficiency) in the same design.

N2 is 15% denser than N3E, and has a 10% to 15% higher performance than N3E at the same power, or a 25% to 30% lower power consumption at the same frequency and complexity.

Following N2E, N2P and voltage-enhanced N2X will be launched in 2026.

TSMC's 2nm process will be put into mass production in the second half of next year, and trial production is currently in full swing. The equipment installation process for Fab20 in Baoshan, Hsinchu has come to an end, and trial production will begin in July. The first P1 plant of Fab22 in Kaohsiung is also about to be completed, and equipment installation is expected to begin in December, with trial production as early as the second quarter of next year.

Foreign media reported that TSMC has already conducted risk trial production of 2nm to ensure a stable yield before mass production. As early as December last year, TSMC showed Apple and Nvidia the test results of the 2nm process technology prototype.

Regarding the progress of 2nm trial production? TSMC said that the 2nm process technology is proceeding as planned. The construction of wafer fabs in Kaohsiung and Arizona, USA are proceeding as planned and progressing well.

In an exclusive interview with Taibao, Liu Peizhen, director of the Industrial Economic Database of the Taiwan Institute of Economics, said that TSMC had originally planned to conduct trial production of 2nm in the second half of this year, and had previously heard that the installation time was earlier than originally expected. It seems that 2nm is expected to enter mass production smoothly in the second half of next year.

As for where the 2-nanometer production base is located, she mentioned that Baoshan, Hsinchu is the first 2-nanometer production base. In addition, after the completion of the three factories in Kaohsiung, 2-nanometer chips will also be produced simultaneously.

According to the media, citing the supply chain, the first 2nm plant in Kaohsiung is about to be completed, and equipment installation is expected to start in December, with trial production as early as the second quarter of next year. As for the second plant in Kaohsiung, the installation schedule will be in the second half of next year or even the beginning of the year after next. The monthly production capacity of both plants is about 20,000 pieces.

As for overseas plants, the first plant of Fab21 in Arizona, USA, is in trial production and will start mass production of 4 nanometers next year. Liu Peizhen believes that TSMC's most advanced process will be run in Taiwan first, and it is also necessary to confirm whether the "yield" can reach a certain level and measure the cost performance. It will be moved to the United States later for mass production of related production lines, which is a more reasonable situation.

According to the latest report from Morgan Stanley, TSMC's 2nm monthly production capacity will increase from 10,000 pieces in trial production this year to about 50,000 pieces next year. By 2026, the A20 chip built into Apple's iPhone 18 will be mass-produced using the 2nm process, and the monthly production capacity will reach 80,000 pieces. The 3nm production capacity will be expanded to 140,000 pieces at the same time, of which the Arizona plant in the United States will have a production capacity of 20,000 pieces.

Apple is undoubtedly TSMC's largest customer, contributing 25% of TSMC's revenue last year. Apple will also become TSMC's first 2nm customer, and it is reported that it has reserved all of TSMC's initial 2nm production capacity for the production of M5 chips. The MacBook Pro laptop with built-in M5 chips is expected to become the first batch of new products using the 2nm process.

In addition to the MacBook Pro laptop, well-known analyst Ming-Chi Kuo recently stated that the processor of the iPhone 17 in 2025 (code-named A19) will still use TSMC's N3P (second-generation 3nm) process; the processor of the iPhone 18 in 2026 will adopt the 2nm process, but based on cost considerations, by then, not all series of iPhone 18 processors may use 2nm.

Liu Peizhen mentioned that in addition to Apple, IC design companies such as high-performance computing, AI chips and even mobile phone processors will be TSMC's 2nm customers. "AMD, Nvidia, MediaTek and Qualcomm are all potential 2nm customers." There are also opportunities in the automotive electronics field. For example, NXP may adopt 2nm in the later stage.

She pointed out that in terms of 2nm orders, TSMC is clearly better than Intel and Samsung, and many major manufacturers have expressed their intention to place orders with TSMC for 2nm. Therefore, there will not be too much problem in filling the 2nm capacity utilization rate in the future. The strong level of 3nm and N3P orders will continue to the 2nm process, continuing to establish TSMC's international leading position in the world's advanced process.

How eager are IC design giants for 2nm? TSMC Chairman and President Wei Zhejia once said that customers have shown great interest and participation in 2nm. They all hope to enter N2 (2nm), N2P and even A16 processes as soon as possible. At the same time, they also hope that the subsequent advanced processes can further reduce power consumption. Wei Zhejia hopes to establish sufficient production capacity in the next two years to meet customer needs.

"TSMC's 2nm technology uses a nanosheet transistor structure, which will be the industry's most advanced semiconductor technology in terms of density and energy efficiency." He also said domineeringly, "After TSMC enters the 2nm process, there will be only one manufacturer in the world that is not TSMC's customer, and the rest are all customers." TSMC will start mass production of 2nm in the fourth quarter of 2025; as for the actual contribution of 2nm to revenue, it will start from the end of the first quarter or the beginning of the second quarter of 2026.

Wei Zhejia emphasized that N2 is a very complex and large technology node. Customers took longer to prepare product design tape outs. So they all contacted TSMC at an early stage, hoping to ensure product blueprints and increase production capacity. From the perspective of product cycle time, N2 will contribute more to revenue than N3, just as N3 contributes more to revenue than N5.

As for TSMC's competitor Samsung, according to Korean media reports, Samsung's 2nm (GAA surround gate) yield rate is less than 20%, and it has decided to withdraw from its Tyler, Texas, USA plant, leaving only a small number of staff. Liu Peizhen said, "Samsung's investment in its US plant has slowed down, reflecting the company's difficulties in the foundry business, and it is indeed unable to snatch more customers from TSMC."

Talking about Intel, she said frankly, "They are currently facing layoffs and capital expenditure cuts, as well as many operational issues. They are unable to focus on the advancement of advanced processes, and it is indeed difficult for them to overtake TSMC."

TSMC's senior vice president of business development and global business and deputy co-chief operating officer Zhang Xiaoqiang once pointed out that N2 (2 nanometer) technology is a revolutionary node, which changes the transistor architecture from FinFET (fin field-effect transistor) to Nanosheet (nanosheet transistor). Through continuous enhancement and collaborative optimization of design technology with innovative structure, nanosheet transistors can provide excellent energy efficiency. Currently, N2 is progressing smoothly, with the conversion target reaching 90% and the yield exceeding 80%.

Through the adjustment of component width, N2 with NanoFlex innovative technology can maximize the advantages of performance, power consumption, and area (PPA). It enables lower-height components to effectively save area and have higher energy efficiency, while higher-height components maximize performance. Customers can optimize the combination of high and low components in the same design block, increase the speed by 15%, and achieve the best balance between area and energy efficiency.

TSMC said that the 2nm process technology provides full-generation performance and power consumption advantages. Compared with N3E, the speed increases by 10-15% at the same power consumption; at the same speed, the power consumption is reduced by 25-30%, and the chip density increases by more than 15%. The number of product design finalizations (tape outs) of 2nm technology in the first two years will be higher than that of 3nm and 5nm in the same period.

TSMC emphasized that the development of 2nm process technology is progressing smoothly, and the device performance and yield are in line with or better than expected. It will enter mass production in 2025 as scheduled, and its mass production curve is expected to be similar to that of 3nm. N2P is an extension of the 2nm family, with a 5% increase in performance over N2 and a 5-10% reduction in power consumption. It is planned to be mass-produced in the second half of 2026; N2X is also planned to be launched in 2026.

Trump was elected as the US President. The academic community is worried that Trump 2.0, who emphasizes "Made in the USA", may require TSMC to set up a factory in the United States in advance for the 2-nanometer process. Although Taiwan's senior official Kuo Chih-hui stated that according to the Taiwan Technology Protection Act, it is not possible to go overseas for production. However, experts believe that this is probably an issue that cannot be stopped. In addition, the United States is fully committed to leading the development of global AI. In the future, it will usher in an era of AI technology-led semiconductor manufacturing and the search for new materials. Due to the situation, TSMC may start 2-nanometer production in the United States in advance.

TSMC will hold a ceremony to celebrate the official production of the first batch of 4-nanometer process wafers at its first plant in Arizona next month. The outside world is worried that Trump may not be satisfied with this and may ask to move the latest advanced process there, and Taiwan's chip industry will lose its leading edge. However, it still depends on whether Trump really threatens and whether TSMC will gain or lose.

When Kuo Chih-hui attended the Economic Affairs Council yesterday, he said that this is a topic worth studying. Taiwan has a technology protection law. As for whether 2-nanometer production can be carried out overseas, "according to relevant laws, I think it is not possible at present." The Ministry of Economic Affairs said that it will still need to be reviewed by the Investment Review Commission.

As for Trump and his Republican Party calling for the abolition of the chip law, Guo Zhihui said that the chip bill will not only benefit TSMC, but also Intel and Samsung in the United States. This is a legal requirement and cannot be easily changed.

Yang Ruilin, research director of the Institute of International Strategic Development of Industrial Technology, analyzed that the global development of AI chips almost all rely on TSMC's production, but the United States has taken a different approach and dominated the core technology of AI development. It will become the leader of the global semiconductor industry if it wants to optimize semiconductor process technology with AI. In the future, these key technologies and manufacturing will definitely be kept in the United States to avoid outflow. Among them, key manufacturing will still have to be undertaken by TSMC, but TSMC is only responsible for manufacturing. To grasp the trend, it is bound to increase investment in the United States. The United States must also rely on TSMC to realize the new generation of semiconductor manufacturing dominated by AI technology. TSMC is absolutely impossible without it.

Under such a structure, it can be observed that in the future the United States is likely to require TSMC to bring advanced processes below 2 nanometers to the United States in advance, or even establish a research and development center in the United States. TSMC's technology will not be hollowed out, and the United States will continue to deepen cooperation with TSMC and will not suppress TSMC.

Recommended Reading