3nm, in high demand

Yesterday, Taiwan Semiconductor delivered an excellent report card, driving the company's stock price to hit a new all-time high.

As shown in the figure below, in TSMC's third-quarter revenue, 5nm has become the company's main process, contributing 32% of this quarter's revenue. As seen from the figure on the right, this situation has been maintained since Q1 of 23. In the current quarter, 5nm contributed 31% of TSMC's revenue, while 7nm only accounted for 20%. In the previous quarter (Q4 of 22), the contributions of 7nm and 5nm were neck and neck.

In the following period, 5nm has been the main revenue driver for the company, but 3nm is quietly rising. Since TSMC successfully mass-produced 3nm FinFET (N3) process technology in 2022, this advanced process has contributed a considerable share to its revenue in just two years.

With the explosive growth of AI applications, major chip manufacturers are rushing into 3nm, and a fierce market competition has fully begun.

Seven Chip Giants Squeeze into 3nm

In many segmented fields, mobile chips, such as Apple's A series, have always been the "bellwether" of semiconductor technology, leading the adoption of the most advanced process nodes. However, with the rapid development of artificial intelligence, AI chips have also begun to emerge in this field and have even become the new "leader." Currently, many AI chip companies have almost all joined the battle for TSMC's new 3nm node.

3nm is the most advanced process node currently in mass production. In 2022, TSMC led the industry in successfully mass-producing 3nm (N3) process technology. N3 is another all-generation process technology following the 5nm (N5) process technology. Following the N3 process technology, TSMC has introduced enhanced versions N3E and N3P processes that support better power consumption, efficiency, and density. In addition, TSMC further provides a wide range of technology combinations to meet diverse customer needs, including the N3X process tailored for high-efficiency computing applications, and the N3AE solution for automotive customers to adopt the most advanced process technology early.

Currently, TSMC's 3nm project list includes Apple A18, Apple M4, A19 chips for the iPhone, Qualcomm's upcoming Snapdragon 8 Gen4 chip, Intel's new Lunar Lake and Arrow Lake CPUs, AMD's new Zen 5 CPUs and the upcoming Instinct MI350 series AI accelerators, as well as NVIDIA's next-generation Rubin R100 AI GPU (to be equipped with future HBM4 memory), and MediaTek's new Dimensity C-X1 (automotive chip). ![] (https://wpimg-wscn.awtmt.com/0f2784b4-69f5-47d3-8249-0108b6eab654.png)

Since the A17 Pro chip last year, Apple has been the first to use the 3nm process. This year, the A18 and A18 Pro further adopt Taiwan Semiconductor's N3E and N3P processes. According to Tianfeng International analyst Ming-Chi Kuo's tweet, Apple's A19 Pro chip for the iPhone 17 in 2025 will continue to use Taiwan Semiconductor's N3P process/3nm technology. The processor for the iPhone 18 model in 2026 is expected to use Taiwan Semiconductor's 2nm technology. However, not all new iPhone 18 models will be equipped with 2nm processors for cost reasons.

![] (https://wpimg-wscn.awtmt.com/8a3ef455-0625-433e-bb8d-13998678ec10.png)

Since Apple's self-developed M-series computer chips, M chips, like A-series chips, have quickly adopted advanced processes. Last year, Apple's M3 series began using the 3nm process. The next generation M4 chip will use Taiwan Semiconductor's second-generation 3nm process, with a transistor count of 28 billion.

![] (https://wpimg-wscn.awtmt.com/0ca973dd-5e83-4fe4-971e-7c04fa379ef6.png)

Qualcomm, the second largest mobile chip supplier after Apple, also started using the 3nm process for the Snapdragon 8 Gen 4 this year. With the Qualcomm Snapdragon Summit approaching, more details about the Snapdragon 8 Gen 4 have been released. It is reported that this is Qualcomm's first smartphone SoC with a self-developed Oryon core, which will be mass-produced using Taiwan Semiconductor's 3nm N3E process. Qualcomm has named it Snapdragon 8 Elite this time, no longer following the naming convention of "4th generation Snapdragon 8" used before. This chip will also become the flagship benchmark for the Android camp.

In addition, Qualcomm's major competitor MediaTek has already released the Dimensity 9400, its first 3nm chip, using Taiwan Semiconductor's second-generation 3nm N3E process, with energy efficiency improved by up to 40%. MediaTek claims that this is the only flagship 5G smartphone chip with the latest Armv9.2 CPU advantage.

![] (https://wpimg-wscn.awtmt.com/b33e732f-1a29-49af-b389-42bb026bae19.png)

Google's Tensor series of mobile chips has been manufactured by Samsung for the past four generations. However, starting from Tensor G5, Google will switch to Taiwan Semiconductor, using the 3nm process and combining it with Taiwan Semiconductor's InFO-POP packaging Tensor G5 is Google's first completely self-designed mobile chip, with the previous four generations of Tensor chips all based on the Samsung Exynos platform for modification and customization. This means that Samsung has lost another important customer, due to issues with the output and power efficiency of Samsung's foundries.

However, it is now an industry consensus that the smartphone market has entered a period of stagnation, while the artificial intelligence market represented by ChatGPT is showing great vitality. Chip giants NVIDIA, AMD, and Intel are all fully investing in the research and production of AI chips, striving to gain a share in the AI chip market.

NVIDIA's Rubin AI GPU is rumored to adopt a 3nm process, but the GPU based on the Rubin architecture is expected to be launched in 2026, making NVIDIA the latest among the major players to release a 3nm chip. Currently, NVIDIA is fully developing the Blackwell GPU, and reports suggest that the supply of this GPU for the next year has already been sold out. It must be said that NVIDIA's GPUs are indeed in high demand, and they are not overly aggressive in terms of process technology.

The AMD Instinct series GPU is an important weapon for the AI market. The Instinct series has previously used a 5nm/6nm dual-node Chiplet mode, but according to AMD's roadmap, starting from the MI350 series, it will be upgraded to 3nm. The next generation Instinct MI350 series is set to be launched in 2025. AMD has stated that the Instinct MI350 series will be based on the 3nm process node, offering up to 288 GB of HBM3E memory and supporting FP4/FP6 data types.

On the Intel side, the Intel AI PC processor Lunar Lake SoC has already adopted the 3nm process, using Intel's Foveros 3D packaging technology, with chip production outsourced to Taiwan Semiconductor Manufacturing Company (TSMC). The Lunar Lake architecture mainly consists of two components: the compute chip and the platform controller chip. The entire compute block, including the P cores and E cores, is built on TSMC's N3B node, while the SoC block is manufactured using TSMC's N6 node.

Another Intel chip that will adopt the 3nm process is the standalone GPU codenamed Falcon Shores. However, with the significant challenges Intel has been facing recently, it seems that Intel is making some adjustments, including layoffs, cost reductions, and shifting focus from the training market to x86 architecture inference. These changes may affect Falcon Shores, which was originally expected to be released next year.

In addition to making mobile processors, MediaTek is also targeting the vast AI market. On April 26th this year, MediaTek unveiled the Tianji Automotive Platform at the Beijing Auto Show. The CT-X1 cockpit platform adopts a 3nm process, while the CT-Y1 and CT-YO adopt a 4nm process, bringing a significant breakthrough in computing power to intelligent cockpits.

MediaTek is also looking to enter the AI PC market. Earlier reports stated that MediaTek is collaborating with NVIDIA to develop an AI PC chip using a 3nm process. The chip is expected to enter mass production in the second half of 2025. This chip will incorporate NVIDIA's GPU IP to provide powerful graphics processing and artificial intelligence computing capabilities for AI PCs, thereby entering the high-end laptop market.

It can be seen that the next generation of AI chips are all embracing 3nm. Why are AI chips favoring advanced processes?

• Surge in computing power demand: The complexity of AI models continues to rise, increasing the requirements for chip computing power. Advanced processes can provide higher transistor density and faster processing speeds to meet the massive computing demands of AI.
• Energy efficiency is crucial: AI chips need to be fast and energy-efficient. Advanced processes bring lower power consumption, extending device battery life, which is crucial for AI applications on mobile devices.
• Fierce market competition: The AI chip market is highly competitive, with major manufacturers launching their own products. Adopting advanced processes not only enhances product performance but also provides a competitive advantage in the market.

TSMC is Making a Fortune

In the field of advanced 3nm chips, TSMC is undoubtedly the undisputed leader. Due to Samsung's yield issues and Intel's lack of mass production, TSMC is leading the way in this process. In the fourth quarter of 2023, despite only Apple adopting TSMC's 3nm process, its revenue still accounted for 15% of the company's total revenue, reaching as high as $29.43 billion, fully demonstrating the huge potential of this advanced process.

In the first and second quarters of this year, the 3nm process node accounted for 9% and 15% of TSMC's total revenue, respectively. The first quarter's share was relatively low, mainly due to seasonal fluctuations in the smartphone market. As shown in the chart above, entering the third quarter, 3nm accounted for 20% of the company's revenue. TSMC CFO Wendell Huang stated that the performance in the third quarter of 2024 benefited from strong demand for 3nm and 5nm technologies in smartphones and AI-related applications, a trend expected to continue into the fourth quarter. Consolidated revenue for the fourth quarter is expected to range between $26.1 billion and $26.9 billion.

However, according to the Electronic Times, with the two high-end processes of 3nm and 5nm, TSMC is expected to generate approximately NT$1 trillion (US$31 billion) in revenue in the first three quarters. TSMC's revenue for the first three quarters of 2024 totaled NT$2.02585 trillion, a 31.9% increase from the same period in 2023 With more and more customers adopting this process, the 3nm process node will account for a considerable share of TSMC's revenue. According to ICSmart.cn, TSMC's N3 series nodes (including N3B and N3E) are expected to account for over 20% of the foundry's total revenue in 2024. It is projected that by 2025, 3nm will represent over 30% of TSMC's revenue in 2025.

Considering the strong demand for AI, coupled with consumer product orders from IC design companies such as Apple and Qualcomm, TSMC's production capacity remains tight.

In a situation of supply shortage, prices are on the rise. It is reported that TSMC plans to raise the prices of advanced processes such as 3nm and 5nm by 8% to ensure a stable long-term profit margin.

Previously, DigiTimes pointed out that TSMC will charge higher prices for its cutting-edge 3nm chips, with 3nm chips expected to exceed $20,000. This also means that the next generation of CPUs and GPUs will be even more expensive than they are now. With a significant increase in production costs, the chip industry is bound to pass on these costs to downstream customers and consumers. It is understood that while the previous generation Snapdragon 8 Gen3 processor chip from Qualcomm had a selling price of around $200, the new Snapdragon 8 Gen4 processor chip may cost over $250, undoubtedly leading to higher prices for new flagship smartphones.

Apart from advanced processes, there are rumors that TSMC is also facing capacity issues with CoWoS due to the demand for AI chips from major players such as AMD and NVIDIA. The Commercial Times cited industry sources indicating that TSMC plans to add new CoWoS-related equipment before the 3rd quarter and has requested equipment manufacturers to deploy more engineers to enhance the capacity of the Longtan AP3 plant, Zhunan AP6 plant, and the Central Taiwan Science Park AP5 plant.

According to sources quoted in a MoneyDJ report, it was previously estimated that TSMC's CoWoS capacity is still in short supply, with a monthly capacity of 35,000 to 40,000 pieces this year. With additional outsourced capacity, next year's output could reach over 65,000 pieces per month, or even higher.

Conclusion

In summary, the seven major chip giants are all rushing into 3nm, which is not only a technological competition but also a competition for production capacity. With the continuous development of technologies such as AI and 5G, the demand for chip performance will continue to grow. The mature application of the 3nm process will provide a solid foundation for these emerging technologies, propelling the entire electronics industry to new heights. However, technological advancements also come with challenges such as high manufacturing costs and yield issues In the future, balancing performance, power consumption, and cost will be an important issue for chip manufacturers. However, for Taiwan Semiconductor Manufacturing Company (TSMC), the good days are still beckoning to them.

According to Nikkei Asia, TSMC's Chairman and CEO C.C. Wei stated that a major customer revealed that the demand for AI is "crazy". He believes this is just the beginning, and this trend is expected to continue for several years. Wei mentioned that almost all AI companies collaborate with TSMC, so TSMC's view on the industry's prospects can be described as both broad and deep.

Furthermore, as Wei mentioned, the customer inquiries for the 2nm process are higher than for the 3nm process, and there is also strong demand for the A16 chip.

Wei stated that high-performance computing (HPC) is shifting towards small chiplet designs, but this will not affect the adoption of the 2nm process. Currently, customers have higher demand for 2nm than for 3nm, and the expected capacity will also be higher. As for the advanced A16 chip, there is strong demand, especially for AI server applications. TSMC is actively preparing relevant capacity to meet customer demands.

"Investors are concerned about future PC and smartphone demand, expecting single-digit growth in PC and smartphone shipments in the future, but silicon content growth will exceed shipment growth," Wei said.

Author: Du Qin DQ, Source: Semiconductor Industry Observation, Original Title: "3nm, Being Snapped Up"