Source: Semiconductor Industry Watch. At yesterday's Conputex conference, Dr. Lisa Su released the latest roadmap. Afterwards, foreign media morethanmoore released the content of Lisa Su's post-conference interview, which we have translated and summarized as follows: Q: How does AI help you personally in your work? A: AI affects everyone's life. Personally, I am a loyal user of GPT and Co-Pilot. I am very interested in the AI used internally by AMD. We often talk about customer AI, but we also prioritize AI because it can make our company better. For example, making better and faster chips, we hope to integrate AI into the development process, as well as marketing, sales, human resources and all other fields. AI will be ubiquitous. Q: NVIDIA has explicitly stated to investors that it plans to shorten the development cycle to once a year, and now AMD also plans to do so. How and why do you do this? A: This is what we see in the market. AI is our company's top priority. We fully utilize the development capabilities of the entire company and increase investment. There are new changes every year, as the market needs updated products and more features. The product portfolio can solve various workloads. Not all customers will use all products, but there will be a new trend every year, and it will be the most competitive. This involves investment, ensuring that hardware/software systems are part of it, and we are committed to making it (AI) our biggest strategic opportunity. Q: The number of TOPs in PC World - Strix Point (Ryzen AI 300) has increased significantly. TOPs cost money. How do you compare TOPs to CPU/GPU? A: Nothing is free! Especially in designs where power and cost are limited. What we see is that AI will be ubiquitous. Currently, CoPilot+ PC and Strix have more than 50 TOPs and will start at the top of the stack. But it (AI) will run through our entire product stack. At the high-end, we will expand TOPs because we believe that the more local TOPs, the stronger the AIPC function, and putting it on the chip will increase its value and help unload part of the computing from the cloud. Q: Last week, you said that AMD will produce 3nm chips using GAA. Samsung foundry is the only one that produces 3nm GAA. Will AMD choose Samsung foundry for this? A: Refer to last week's keynote address at imec. What we talked about is that AMD will always use the most advanced technology. We will use 3nm. We will use 2nm. We did not mention the supplier of 3nm or GAA. Our cooperation with TSMC is currently very strong-we talked about the 3nm products we are currently developing. Q: Regarding sustainability issues. AI means more power consumption. As a chip supplier, is it possible to optimize the power consumption of devices that use AI? A: For everything we do, especially for AI, energy efficiency is as important as performance. We are studying how to improve energy efficiency in every generation of products in the future-we have said that we will improve energy efficiency by 30 times between 2020 and 2025, and we are expected to exceed this goal. Our current goal is to increase energy efficiency by 100 times in the next 4-5 years. So yes, we can focus on energy efficiency, and we must focus on energy efficiency because it will become a limiting factor for future computing. Q: We had CPUs before, then GPUs, now we have NPUs. First, how do you see the scalability of NPUs? Second, what is the next big chip? Neuromorphic chip? A: You need the right engine for each workload. CPUs are very suitable for traditional workloads. GPUs are very suitable for gaming and graphics tasks. NPUs help achieve AI-specific acceleration. As we move forward and research specific new acceleration technologies, we will see some of these technologies evolve-but ultimately it is driven by applications. Q: You initially broke Intel's status quo by increasing the number of cores. But the number of cores of your generations of products (in the consumer aspect) has reached its peak. Is this enough for consumers and the gaming market? Or should we expect an increase in the number of cores in the future? A: I think our strategy is to continuously improve performance. Especially for games, game software developers do not always use all cores. We have no reason not to adopt more than 16 cores. The key is that our development speed allows software developers to and can actually utilize these cores. Q: Regarding desktops, do you think more efficient NPU accelerators are needed? A: We see that NPUs have an impact on desktops. We have been evaluating product segments that can use this function. You will see desktop products with NPUs in the future to expand our product portfolio.

Yesterday,$Taiwan Semiconductor (TSM.US)$Turned in an excellent report card, driving the company's stock price to a new all-time high.

As shown in the following picture, in taiwan semiconductor's revenue for the third quarter, 5nm has become the main process of the company, contributing 32% of the revenue this quarter. It can be seen from the figure below that this situation has been maintained since Q1 of 23. In the current quarter, 5nm contributed 31% of taiwan semiconductor's revenue, while 7nm accounted for only 20%. In the previous quarter (Q4 of 22), the contributions of 7nm and 5nm were equally important.

Subsequently, 5nm has always been the main contributor to the company's revenue, but 3 nanometers are quietly rising. Since 2022, when taiwan semiconductor successfully mass-produced 3nm FinFET (N3) process technology, in just two years, this advanced process has made a significant contribution to its revenue.

With the explosive growth of AI applications, major chip manufacturers are rushing to the 3nm process, and an intense market competition has fully begun.

Seven major chip giants are squeezing into 3nm.

In many specific areas, mobile chips, such as $Apple (AAPL.US)$ A-series, has always been a "barometer" of semiconductor technology, adopting the most advanced process nodes. However, with the rapid development of artificial intelligence, ai chips have also begun to emerge in this field, even becoming the new "leader." Currently, many AI chip companies have joined the battle for Taiwan Semiconductor's 3nm new node.

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

Currently, the TSMC 3nm project list includes Apple A18, Apple M4, and A19 chips for iPhone. $Qualcomm (QCOM.US)$The upcoming Snapdragon 8 Gen4 chip, $Intel (INTC.US)$ the new Lunar Lake and Arrow Lake CPUs, $Advanced Micro Devices (AMD.US)$ the new Zen 5 CPU and upcoming Instinct MI350 series AI accelerators, as well as $NVIDIA (NVDA.US)$the next generation of Rubin R100 AI GPU (featuring future HBM4 memory), MediaTek's new Dimensity C-X1 (automotive chip).

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

Since Apple's self-developed computer chip M series, the M chip, like the A series chip, has quickly adopted advanced technology. Last year, Apple's M3 series started using 3nm technology. The next generation M4 chip will use taiwan semiconductor's second generation 3nm process, with a transistor count of 28 billion.

Qualcomm is the second largest mobile chip supplier after Apple. Qualcomm's Snapdragon 8 Gen 4 also started using 3nm process this year. With the upcoming Qualcomm Snapdragon Summit, more details about Qualcomm Snapdragon 8 Gen 4 have been revealed. It is reported that this is Qualcomm's first smartphone SoC equipped with self-developed Oryon kernel, and will be mass-produced using taiwan semiconductor's 3nm N3E process. Qualcomm has named it Snapdragon 8 Elite this time, no longer using the previous naming convention of "fourth generation Snapdragon 8". This chip will also become the flagship benchmark for the Android camp.

However, in addition to this, one of Qualcomm's major competitors, MediaTek, has already released the Dimensity 9400, which is MediaTek's first 3nm chipset, using taiwan semiconductor's second generation 3nm process N3E, with energy efficiency increased by up to 40%. MediaTek claims that this is the only flagship 5G smart phone chip with the latest Armv9.2 CPU advantage.

$Alphabet-A (GOOGL.US)$ / $Alphabet-C (GOOG.US)$ The Tensor series mobile chip from the past four generations has always used Samsung's process outsourcing. However, starting from the Tensor G5, Google will switch to Taiwan Semiconductor, using Taiwan Semiconductor's 3nm process and paired with Taiwan Semiconductor's InFO-POP packaging. The Tensor G5 is Google's first completely self-designed mobile chip, while the previous four generations of Tensor chips were all based on the Samsung Exynos platform for modification and customization. This means Samsung has lost another important customer, due to issues with the yield and power efficiency of Samsung's manufacturing plant.

However, the smartphone market has entered a sluggish period, which has become an industry consensus, while the artificial intelligence market, represented by ChatGPT, has shown great vitality. The three chip giants NVIDIA, AMD, and Intel are fully investing in the research and production of AI chips, striving to carve out a share of the AI chip market.

There are rumors in the industry that NVIDIA's Rubin AI GPU will adopt the 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 is already sold out for the next year. It must be said that NVIDIA's GPUs are indeed in high demand, and they do not need to be overly aggressive in terms of technology.

The AMD Instinct series GPU is an important weapon in the AI market. Instinct has previously used a 5nm/6nm dual-node Chiplet mode, but from multiple AMD roadmap releases, by the time of the MI350 series, it will be upgraded to 3nm. The next generation Instinct MI350 series will be launched in 2025. AMD stated that 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.

When it comes to intel, the AI PC processor Lunar Lake SoC from intel has already used 3 nanometers, adopted intel's Foveros 3D packaging technology, and outsourced chip production to taiwan semiconductor. The Lunar Lake architecture consists mainly of two components: a computing chip and a platform controller chip. The entire computing block, including the P-core and E-core, are built on taiwan semiconductor's N3B node, while the SoC block is manufactured using taiwan semiconductor's N6 node.

Another 3nm chip that intel is planning to use is the standalone GPU codenamed Falcon Shores. However, with the significant challenges that intel has been facing recently, intel seems to be making some adjustments, including layoffs, cost reductions, and shifting focus to x86 architecture inference instead of training markets, which could affect Falcon Shores. The product was originally expected to be released next year.

In addition to making phone processors, mediatek is also eyeing the broad AI market. On April 26th of this year, mediatek unveiled the new Tianji automobile platform at the Beijing Auto Show. The Tianji auto cockpit platform CT-X1 uses a 3nm process, while CT-Y1 and CT-YO use a 4nm process, bringing significant computational breakthroughs to smart cockpits.

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

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

There is a surge in computational demand: The complexity of AI models continues to increase, leading to a growing demand for chip computing power. Advanced processes can provide higher transistor density and faster computation speed, meeting the immense computational requirements of AI.

Energy efficiency is crucial: AI chips not only need to be fast but also energy efficient. Advanced processes can offer lower power consumption, extending device battery life, which is particularly critical for AI applications on mobile devices.

Intense market competition: The competition in the AI chip market is exceptionally fierce, with major manufacturers introducing their own products one after another. Adopting advanced processes can not only enhance product performance but also give a competitive edge in the market competition.

Taiwan Semiconductor's profits are booming.

In the advanced 3-nanometer chip field, Taiwan Semiconductor is undoubtedly the worthy leader. Due to Samsung's yield issues and Intel's lack of mass production, Taiwan Semiconductor has taken the lead in this process. In the fourth quarter of 2023, despite only Apple as a customer adopting Taiwan Semiconductor's 3-nanometer technology, its revenue still accounts for 15% of the company's total income, reaching as high as $2.943 billion, fully demonstrating the enormous potential of this advanced process.

In the first and second quarters of this year, the 3-nanometer process node accounted for 9% and 15% of Taiwan Semiconductor's total revenue, respectively. In the first quarter, the proportion was relatively low, mainly due to seasonal fluctuations in the smart phone market. Entering the third quarter, as shown in the chart, 3-nanometer technology contributed 20% to the company's revenue. Taiwan Semiconductor CFO Huang Renzhao stated that the performance in the third quarter of 2024 will benefit from strong demand for 3-nanometer and 5-nanometer technologies in smartphones and AI-related applications, which is expected to continue into the fourth quarter. Consolidated revenue in the fourth quarter is expected to range between $26.1 billion and $26.9 billion.

According to Electronic Times, with the help of the high-end 3-nanometer and 5-nanometer processes, Taiwan Semiconductor is expected to generate about NT$1 trillion ($31 billion) in revenue in the first three quarters. In the first three quarters of 2024, Taiwan Semiconductor's total revenue reached NT$2025.85 billion, a 31.9% increase from the same period in 2023.

As more customers adopt this technology, the 3nm process node will account for a significant share of Taiwan Semiconductor's revenue. According to ICSmart.cn, Taiwan Semiconductor's N3 series nodes (including N3B and N3E) will account for over 20% of the total revenue of the foundry in 2024. It is estimated that in 2025, 3nm will represent over 30% of Taiwan Semiconductor's revenue in 2025.

Considering the strong demand for AI, along with consumer product orders from companies like Apple and Qualcomm, Taiwan Semiconductor's production capacity remains tight.

In a situation of supply shortage, the trend of price increase has emerged. It is reported that Taiwan Semiconductor plans to raise the prices of advanced processes such as 3nm and 5nm by 8% to ensure long-term stable profit margins.

Earlier, DigiTimes pointed out that Taiwan Semiconductor will charge higher prices for its cutting-edge 3nm chips, with 3nm expected to exceed 0.02 million US dollars. This also means that the next generation of CPUs and GPUs will be more expensive than now. With a significant increase in production costs, the chip industry is bound to pass these on to downstream customers and consumers. It is understood that the previous generation Snapdragon 8 Gen3 processor chip from Qualcomm itself had a price of about $200, but the new Snapdragon 8 Gen4 processor may sell for over $250, undoubtedly raising the prices of new flagship smartphones.

In addition to advanced processes, there are rumors that Taiwan Semiconductor is facing capacity issues related to CoWoS due to the demand for AI chips from major companies such as AMD and NVIDIA. The Commercial Times cited industry sources indicating that Taiwan Semiconductor plans to add CoWoS-related equipment before the third quarter, and has requested equipment manufacturers to dispatch more engineers to enhance the capacity of the Longtan AP3, Zhunan AP6, and Central Taiwan Science Park AP5 factories.

According to sources quoted in the MoneyDJ report, it was previously estimated that Taiwan Semiconductor's CoWoS capacity is still in short supply, with a monthly capacity of 0.035 million to 0.04 million units this year. With additional outsourced capacity, next year's output could reach over 0.065 million units per month, or even higher.

In conclusion,

In general, the seven major chip giants are rushing into the 3nm process, which is not only a technological competition, but also a competition for 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, driving 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 costs will be an important issue for chip manufacturers. However, for Taiwan Semiconductor, good days are still beckoning to them.

According to Nikkei Asia, TSMC Chairman and CEO C.C. Wei mentioned 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 of the industry's prospects can be described as both broad and deep.

Moreover, as Wei Zhejia said, are 2nm customer inquiries higher than 3nm, and the demand for A16 is also strong.

Wei Zhejia stated that high-performance computing (HPC) acceleration towards small chiplet designs will not affect the adoption of 2nm. Currently, customers' demand for 2nm is even higher than for 3nm, and production capacity is expected to be higher as well. As for the most advanced A16, the demand is also strong. A16 has a strong appeal for AI server applications, and Taiwan Semiconductor is actively preparing relevant production capacity to meet customer demand.

"Institutional investors are concerned about future PC and smart phone demand, expecting single-digit growth in future PC and phone shipments, but silicon content growth will surpass shipment volume," said Wei Zhejia.

Editor / jayden