1.In the first half of the year, China’s integrated circuit exports were 650.26 billion yuan, a year-on-year increase of 20.3%
According to the General Administration of Customs of China, in the first half of 2025, our country’s import and export of goods trade will be 21.79 trillion yuan, a year-on-year increase of 2.9%. Among them, exports were 13 trillion yuan, an increase of 7.2%; imports were 8.79 trillion yuan, down 2.7%. In the first half of 2025, our country’s import and export scale will stand at 20 trillion yuan, a record high for the same period in history. From the perspective of quarterly trends, imports and exports in the second quarter increased by 4.5% year-on-year, 3.2 percentage points faster than the first quarter, and maintained year-on-year growth for seven consecutive quarters.
In the first half of 2025, our country exported 7.8 trillion yuan of mechanical and electrical products, an increase of 9.5%. Among them, automatic data processing equipment and its parts were 702.79 billion yuan, an increase of 3.0%; the export volume of integrated circuits increased by 20.6% to 167.77 billion, and the export value increased by 20.3% to 650.26 billion yuan; automobiles were 428.72 billion yuan, an increase of 9.4%; mobile phones were 357.35 billion yuan, down 7.4%.
In the first half of 2025, our country imported 3.4 trillion yuan of mechanical and electrical products, an increase of 6.3%. Among them, the number of integrated circuits increased by 8.9% to 281.88 billion, and the value increased by 8.3% to 1.38 trillion yuan, while the number of automobiles decreased by 32.4% to 224,000 units, and the value decreased by 37.1% to 83.18 billion yuan.
In 2024, the output of integrated circuit products will be 451.4 billion pieces, a year-on-year increase of 22.2%; The output of solar cells and industrial robots increased by 15.7% and 14.2% year-on-year respectively, leading the overall industrial growth rate.
IC design is the most critical link in the chip value chain in the semiconductor industry, and the revenue of this link is also worth paying attention to. According to data previously released by the Ministry of Industry and Information Technology, the revenue of China’s integrated circuit design industry in 2024 will be 364.4 billion yuan, a year-on-year increase of 16.4%. In contrast, China’s integrated circuit design revenue in 2023 will be 306.9 billion yuan, a year-on-year increase of 6.4%, that is, the growth rate of the domestic integrated circuit design industry will accelerate in 2024.
2.Intel 2nm chip TapeOut TSMC foundry!
According to reports, Intel’s next-generation client CPU flagship Nova Lake-S has been sold at TSMC’s Tape-Out fab in Taiwan. We previously speculated based on rumors that Intel would use its own in-house 18A process and use TSMC’s 2nm mass production technology. However, according to SemiAccurate, Intel has tape-out a computing module on TSMC’s N2 process, which means that Nova Lake-S’s computing module is likely to use both 18A and TSMC N2 processes. One possible reason for Intel’s decision is that Intel is building a reliable back-up solution if the 18A process cannot be delivered, or if demand is expected to be too high and internal capacity cannot be met. In any case, customers can expect the product to be delivered on time in the second half of 2026, but there may be some interesting solutions behind it.
As for the specific date, it will take several months from tape-out to final product delivery. The chips currently completed by Tape-Out are being powered up in Intel labs, running various tests to test the chip’s performance under multiple uses and check its operation correctly. Typically, it takes a few weeks to a month to power up, and final mass production will begin in a few months. It will take another two to three months to manufacture and ship after that, meaning Nova Lake-S will most likely be released in the third quarter of 2026. It should be reminded that this CPU will integrate up to 52 cores (16 P-cores, 32 E-cores, and 4 LPE cores) with an 8,800MT/s memory controller, as well as Xe3 Celestial for graphics rendering and Xe4 Druid for media and display tasks. This certainly makes it an attractive product, but it is also quite difficult to manufacture due to its heterogeneous complexity.
3.3D NOR flash memory is emerging as a new player in the market.
To truly understand the significance of this shift, it is first necessary to clarify the difference between NOR and NAND memory. NOR flash memory is known for its fast random read speeds, high reliability, and stable performance in extreme temperatures in demanding “execute-in-place” applications. It has always been a core component in application scenarios with high code storage requirements, such as automotive, cloud computing, and industrial systems. At the same time, due to its excellent write endurance, it also occupies an important position in applications that require multiple updates, such as OTA transmission.
In contrast, NAND flash memory is better suited for data storage, offering higher storage density at a lower cost per bit, but without the other benefits of NOR flash. As a result, these two storage devices each address the unique challenges faced by their respective application markets.
With the rise of artificial intelligence (AI), the Internet of Things (IoT), and edge computing, the limitations of 2D NOR flash memory are becoming more apparent. These applications and their users require both higher storage density and rely on the high reliability benefits of NOR flash. This is where the benefits of 3D NOR flash memory come into play, as it can help address these challenges by providing higher density, greater scalability, and improved reliability.
In contrast, 3D NOR flash overcomes the scalability issues inherent in 2D NOR architectures by stacking storage cells vertically. 2D NOR flash memory can achieve up to 512Mb of storage capacity on a single chip, and to further increase density, it must be achieved through multi-chip system-in-package (SIP).
As a result, 3D NOR flash memory is an attractive solution for applications that require large-scale non-volatile memory (NVM) in space-constrained environments.
3D NOR flash memory pushes the performance limits of non-volatile memory, and its core technical advantages highlight its potential to become a “game-changer” in the memory market.
4.The density is 8 times higher than that of 2D NOR
One of the most significant advantages of 3D NOR flash memory is its storage density up to 8 times that of 2D NOR flash memory. As mentioned earlier, the 3D architecture achieves a single-chip capacity of 4Gb through vertical stacking, while the maximum capacity of 2D NOR is only 512Mb. This significant increase in density, with the ability to store larger data sets within the same physical size, addresses one of the most pressing challenges in the current memory market.
Therefore, this 3D NOR architecture will provide up to 8Gb of storage capacity, making it suitable for a range of storage-demanding applications, from consumer electronics to high-end industrial systems.
5.CCTV Interview: Huang Renxun has been rooted in China for 30 years, BAT is a friend, DeepSeek is very innovative, and Huawei deserves respect
On July 16, in an exclusive interview with CCTV’s “Face to Face”, NVIDIA founder Jensen Huang had an in-depth dialogue with CCTV reporter Dong Qian on core issues such as supply chain, Chinese market, artificial intelligence and competition.
In the dialogue, Huang Renxun used “unique” to summarize the Chinese market – in the 30 years of deep cultivation, Nvidia has been deeply bound to Lenovo, Alibaba, Tencent, Baidu and Xiaomi and other companies, and he particularly appreciates China’s innovation vitality in the field of AI, believing that although H20 is not a top product, it has spawned breakthroughs such as the in-depth search for R1, confirming that “China’s innovation is unstoppable”.
“You have to admire the amazing innovation ability of DeepSeek, the R1 model they developed is a real innovation, it redesigns a lot of the way AI models work, so that they can take full advantage of the H20 architecture, which is very creative.” Huang Renxun said.
When it comes to competition, Huang shows a unique perspective – he respects opponents such as Huawei, advocates “competition and cooperation”, opposes comparing business to war, and believes that competition is the cornerstone of market prosperity, and that both enterprises and countries can find a way to coexist in competition.
Huang said, “Huawei is much larger than us, and in terms of scale, personnel and technical capabilities, they are both wide and deep. This is a company with strong chip design capabilities, system design and system software, and if we are not here, Huawei will definitely find its own solution. ”
When asked if Huawei is a competitor or a partner, Huang said: “Their achievements are admirable, the company is still very competitive, they are our competitors, but you can still admire and respect the competitors and maintain a good relationship with them.”
6.Semiconductor winners take all, with 5% of companies monopolizing $159 billion in profits
According to a report released by global consulting firm McKinsey & Company on July 20, almost all of the economic profits generated by the global semiconductor industry last year were divided among the top 5% of leading companies (based on annual sales) – including NVIDIA, TSMC, SK Hynix and Broadcom.
The report pointed out that these 5% of the top companies received up to $159 billion in economic profits, while the middle 90% of companies had a total profit of only $5 billion. The bottom 5% of companies lost $37 billion. In other words, the profits of leading companies even exceed the total economic profit of $147 billion created by the entire semiconductor industry.
This change in market structure took only two or three years. During the pandemic (2021-2022), the middle 90% of companies still earned more than $30 billion in economic profits per year, with an average annual profit of about $130 million per company. However, since the rise of the AI semiconductor boom in 2023, this number has plummeted to $38 million, and last year it fell to $17 million, a drop of 88% in two years.
McKinsey predicts that AI-related semiconductor companies will continue to grow at an annual rate of 18% to 29% until 2030, while traditional semiconductor companies that are not directly related to AI will grow at an annual rate of only 2%-3%. McKinsey analyzed: “Although some companies are using the AI value creation wave to make unprecedented profits, most companies are facing a completely different reality. ”
The key reason for the “winner-takes-all” situation is that leading companies have the right to set standards for new semiconductor products. The standards for traditional products are set by JEDEC (Joint Electronic Devices Engineering Committee), and companies develop products according to the standards. However, when the product specifications are completely different, the first company to enter the market has the right to lead the establishment of standards, thus becoming “rulemakers” and restricting the entry of latecomers.
7.The localization of semiconductor equipment has ushered in a key turning point, and Yangtze River Storage’s first national production and chemical production line will be put into trial production this year
In order to reduce dependence on foreign equipment, Yangtze Storage Technology Co., Ltd. (YMTC) has made a major breakthrough in promoting “national” manufacturing equipment, and the first nationally produced production line will be introduced into trial production in the second half of 2025.
In 2016, Yangtze River Storage was officially registered in Wuhan East Lake New Technology Development Zone, focusing on the design, manufacturing and sales of 3D NAND flash memory chips. At the end of 2022, Yangtze Storage was included in the U.S. Department of Commerce’s entity list, and in the absence of access to U.S. advanced wafer manufacturing equipment, it still relied on existing tools to maintain the development and manufacturing of advanced NAND Flash product lines, and still actively promoted production capacity expansion plans. At present, Yangtze Storage’s production capacity is close to 130,000 wafers per month, accounting for about 8% of the global production capacity, and plans to achieve a monthly production capacity of about 150,000 wafers (WSPM) in 2025, and strive to account for 15% of the global NAND flash supply by the end of 2026.
At the technical level, Yangtze River Storage has also made significant progress. Its 232-layer TLC (triple-layer cell) chip X4-9070 has achieved 294-layer equivalent density through double-layer stacking, with an interface speed of 3600MT/s. The 3D QLC X4-6080 will be launched later in 2025, possibly continuing the 294-layer stack, and mass production of 2TB 3D TLC X5-9080 and 3D QLC X5-6080 by 2026, the latter of which will support 4800MT/s high-speed interface. The next-generation architecture is expected to exceed 300 layers of stacks, thereby increasing bit output per wafer and maintaining total output growth even as process time increases and monthly wafer yields decrease.
If the trial production of the national production line is successful, it is expected to double the bit output and help Yangtze River Storage achieve its market share target. The top 3 global memory production capacity are Samsung, SK hynix, and Micron, and the production capacity forecasts of the three companies in 2025 are 660,000 pieces, 500,000 pieces and 300,000 pieces respectively. However, it must be clearly realized that from the test line to large-scale mass production is not achieved overnight, and the long-term stability of domestic equipment, process compatibility between different equipment, and cost control capabilities are all key issues that need to be solved – from yield stability to cost optimization, to product iteration, at least 3-5 years of grinding cycle is needed.
According to Morgan Stanley’s estimate of 45% domestic equipment adoption, far exceeding the national average and other major domestic fabs (SMIC, China’s largest fab, achieved a 22% localization rate in its Beijing fab and 18% in the Lingang fab), but the 45% localization rate is still well below 100%.
The trial production of Yangtze River Storage’s national production line is a “paradigm innovation” for China’s semiconductor industry in global competition. It proves that even in the case of backward single-point technology, the overall breakthrough of the industrial chain can still be achieved through the system collaboration of equipment, process and architecture. The pull-through of the first national production line is of great significance: integrated circuit production is a systematic project, which involves nine categories of core equipment and other auxiliary modules, and the pull-through of the national production line is not a breakthrough in a single link, but a breakthrough in all links.
our country In 2024, the localization rate of adhesive removal equipment, CMP equipment, cleaning equipment, etching equipment and heat treatment equipment will exceed 30%, and the localization rate of PVD/CVD/ALD, adhesive development, ion implantation, quantity detection and lithography will still be less than 20%, which are 5~20%, 5~10%, 10~20%, 1~10% and 0~1% respectively.
More notably, TechInsights’ report shows that Yangtze Storage’s latest “Xtacking 4.0” chip is on par with market leaders in terms of performance, but due to the gap between China and foreign countries in key areas such as extreme ultraviolet lithography (EUV), its continued growth will depend on its ability to close the gap between equipment capacity and output.
8.SEMI reports that total global semiconductor equipment sales in 2025 are expected to reach $125.5 billion, a record high
On July 22, 2025, SEMI pointed out in its Mid-Year Total Semiconductor Equipment Forecast – OEM Perspective that global total semiconductor manufacturing equipment sales by global original equipment manufacturers (OEMs) are expected to hit a new record of $125.5 billion in 2025, a year-on-year increase of 7.4%. Driven by the continuous migration of advanced logic, memory and technology, equipment sales are expected to further climb to $138.1 billion in 2026, achieving three consecutive years of growth.
Semiconductor Equipment Sales by Segment
After setting a record sales record of $104.3 billion in 2024, the wafer fab equipment (WFE) segment, which includes wafer processing, fab facilities, and mask/mask equipment, is expected to grow by 6.2% to $110.8 billion in 2025. This figure was revised up from SEMI’s forecast of $107.6 billion at the end of 2024, mainly driven by increased sales of foundries and memory application devices. Looking ahead to 2026, the WFE segment is expected to grow by a further 10.2% to $122.1 billion, driven by advanced logic and memory capacity expansion to support AI applications and process technology migration across key market segments.
The back-end equipment segment is expected to continue to grow on the strong recovery that began in 2024. Semiconductor test equipment sales are expected to grow by a further 23.2% in 2025, reaching a record $9.3 billion, following a 20.3% year-on-year increase in 2024. Packaging equipment sales increased by 25.4% in 2024 and are expected to grow by another 7.7% to $5.4 billion in 2025. In 2026, the expansion momentum in the back-end equipment field will continue, with test equipment sales expected to increase by 5.0% and packaging equipment sales expected to increase by 15.0%, achieving three consecutive years of growth. This growth is primarily driven by significant increases in device architecture complexity and strong demand for high performance in artificial intelligence and high-bandwidth memory (HBM) semiconductors. However, continued weakness in automotive, industrial, and consumer end markets will affect the growth of this segment to some extent.
Semiconductor Equipment Sales by Region
It is expected that by 2026, Chinese mainland, Taiwan and South Korea will continue to maintain the top three positions in equipment spending. Chinese mainland will continue to lead all regions during the forecast period, although sales are expected to decline from a record $49.5 billion in 2024. All regions except Europe are expected to see a significant increase in equipment spending from 2025. However, increasing trade policy risks could affect the pace of growth in various regions.
