1.Straight-line overtaking representative: domestic memory chip major breakthrough
According to Indian Technology reports: China’s leading NAND FLASH memory company Y company, and chip memory giant Samsung of South Korea reached a cooperation, Samsung has confirmed that from the V10, it will use the patented technology from the Chinese counterpart Y company, mainly in the new advanced packaging technology “hybrid bonding”.
Samsung has chosen to sign a hybrid bonding patent license agreement with Y Company to avoid potential patent risks!
Hybrid bonding is on the rise
Hybrid bonding equipment is the key core equipment required by Y company in this paper, which provides wafer-to-wafer (commonly known as W2W) and Die-to-wafar (commonly known as D2W) process capabilities.
At present, there are many companies at home and abroad that can provide a variety of bonding equipment, including EVG, Tokyo Precision, Shippura, ASMI, Besi, K&S and so on.
Domestic manufacturers include Xinyuan Micro, Suzhou Xinrui, Tuojing, Xinhui Lianxin (a subsidiary of Baiao Chemical), Shengmei, Qinghe Crystal Yuan and so on.
How do you view the cooperation between Samsung and Y Company
First of all, this cooperation can be seen as a great success for China on the semiconductor technology route!
In the past, Samsung usually arranged the control circuit on a single wafer, and then stacked the storage core unit above it, which is called COP (Cell on Peripheral), but when the number of NAND stacked layers exceeds 400, the bottom control circuit will bear great pressure, which will affect the reliability of NAND.
Therefore, Samsung decided to use the technology of manufacturing the memory unit and the control circuit on different wafers for the V10 product, and then merging them through hybrid bonding.
After Samsung’s research, it found that Y company has established a considerable number of patents in its NAND packaging technology route, forming a solid patent wall. For now, it is almost impossible for Samsung to bypass Y company’s patent wall and start a new business.
Therefore, Samsung chose to sign the use license of the hybrid bonding patent with Y Company and obtained the technology license from Y Company.
Therefore, we can see the great success of China’s technology route!
2. AI chip competition intensifies between China and the US
1) Chip field: the United States dominates, China breaks through
American advantage: Nvidia’s GPU dominates the global AI chip market, and its high-end products such as H100 and A100 are the core hardware for training large models. But the high cost ($15,000 to $40,000 for a single card) and supply dependence have prompted Microsoft, Meta and others to seek alternatives.
China catching up: Domestic chip companies such as Huawei (Ascend 910B), Cambrian, Haiguang and others have accelerated research and development, and the performance of some products is close to the international mainstream level. The AI reasoning chip launched by Alibaba’s Pingtou Brother is also making progress in the field of edge computing. It is worth noting that DeepSeek has optimized its algorithm and only used 2,000 low-end chips to complete training, at a cost of only 1/10 of Nvidia’s solution, showing China’s potential in algorithm innovation.
2) The Big Model Race: Open source vs. Ecology
American camp: OpenAI, Anthropic, xAI and other companies continue to finance and expand production. OpenAI plans a new round of $40 billion in financing, and its valuation may reach $300 billion; Anthropic is valued at $61.5 billion with funding from Google, Amazon and others.
China breakthrough: DeepSeek open source model overtakes ChatGPT in global downloads and technical metrics, sparking talk of a “Chinese version of Sputnik moment” The capability of domestic first-tier models (such as Tongyi Qianwen, Doubao, Kimi) has approached that of ChatGPT-4, and some Chinese scenarios perform better.
3) Application layer and ecology: China leads the penetration rate
Chinese characteristics: wechat, Tiktok and other super apps integrate AI functions to promote the application of the whole people. After DeepSeek gained access to wechat, it quickly covered hundreds of millions of users, forming a closed loop of “application scenaries-driven technology iteration”.
Us layout: OpenAI, Microsoft and others strengthen cloud services and enterprise-level solutions, while consolidating the technology ecology through open source communities (such as HuggingFace).
4) Capital flow: risk and opportunity coexist
Us financing wave: In 2024, US AI startups will receive $97 billion of investment, accounting for half of the world. In 2025, it will continue to heat up, and chip startups Positron and EnCharge AI will receive hundreds of millions of dollars in financing.
China’s capital focus: With policy support, local governments (such as Shanghai’s 60 billion AI Fund) and enterprises (such as Alibaba’s 380 billion computing power investment) increase investment, while attracting the attention of international capital. Goldman Sachs predicts that AI adoption in China will boost corporate earnings by 2.5% over the next decade, driving more than $200 billion in capital inflows.
5) Challenges and future trends
Technical bottlenecks: reliance on high-end chips and brain drain are still China’s weaknesses; The United States faces problems such as algorithmic efficiency and ethical supervision.
Open source and globalization: DeepSeek’s open source strategy accelerates technology diffusion and could reshape the global AI competitive landscape.
Policy game: The US chip embargo and data restrictions are hedging against China’s “new national system”, and the technology race will be more intense in the short term.
Summary
The Sino-US AI competition has evolved from a single technological breakthrough to a competition for the whole industrial chain, involving chips, algorithms, applications and ecological construction. Although the US still dominates in hardware and basic research, China is rapidly narrowing the gap through application scenario innovation, concentrated capital investment and policy support. In the future, the two sides may jointly promote the inclusion and globalization of AI technology in the game of open source collaboration and technology blockade.
3.Foreign media: Huawei AI chip major breakthrough!
According to foreign media “Financial Times” reported: Huawei AI chip has made a major breakthrough, Ascend 910C processor yield increased to 40%, achieve profitability, plans to further increase to 60% to reach the industry standard.
Huawei plans to produce 100,000 Ascend 910C and 300,000 Ascend 910B chips this year, although sales are lagging behind Nvidia, but still dominate the Chinese AI chip market.
Huawei has made significant breakthroughs in AI chip production. Its latest Ascend 910C processor yield has increased to 40 percent, compared with 20 percent a year ago, which is seen as a significant improvement.
Huawei plans to continue to increase the yield, aiming for 60%. This plan, if realized, will further consolidate and enhance Huawei’s position in the AI chip market. At present, Huawei has accounted for more than 75 percent of the total production of AI chips on the Chinese mainland, underscoring its dominant position in the domestic semiconductor market.
Huawei also plans to produce 100,000 Ascend 910C processors and 300,000 910B chips this year to meet the market’s growing demand for AI chips.
Huawei faces challenges as well as opportunities
On one hand, Huawei needs to convince customers to ditch Nvidia’s premium Cuda software in favor of its Ascend series of chips. This requires Huawei to continuously improve its technology and software support to win customers’ trust and favor. On the other hand, the Ascend 910B chip still has limitations in areas such as inter-chip connectivity and memory, which will require Huawei to continuously invest in research and development to improve.
Despite Huawei’s dominant position in China’s AI chip market, there are still smaller competitors. Huawei still faces difficulties in ensuring that SMIC has enough capacity to make Ascend chips. This requires Huawei to further strengthen its supply chain management and capacity building to ensure its leading position in the AI chip market.
4.Chinese companies cut 21 percent of imported chip orders again, causing a huge loss of 350 billion yuan to the United States, and spent 170 billion yuan to open three consecutive fabs against the United States
According to domestic authoritative media reports, according to an authoritative international research report data show that Chinese companies cut about 21.02% of imported chip orders, the total price of its orders as high as 350 billion. Intel, Qualcomm, these US manufacturers caused a certain degree of market impact, the stock prices of related companies began to plummet.
And from the Chinese mainland, SMIC, has thrown down 170 billion yuan to open three fabs, and fully promote the 28nm chip capacity expansion.
With the rise of domestic chips, the revenue of American companies such as Intel and Qualcomm in China has also begun to decline. According to relevant data reports, Intel’s revenue in China fell by 15% in a single quarter, while Qualcomm’s revenue in the Chinese market dropped from 65% to 48%, and the two companies fell 7.2% and 9.8% in the stock market, respectively.
The root cause of the economic damage suffered by overseas companies in the Chinese market for years is that Chinese domestic chips are gradually replacing foreign imported chips in the domestic market. In addition, the United States unilaterally imposed restrictions on the import and export of products in the Chinese market, which further encouraged Chinese enterprises to achieve the determination of domestic chip supply.
As early as 2022, SMIC has already won financial support from all walks of life, and invested 170 billion yuan to carry out the layout of the manufacturing industry chain of domestic chips. In Shanghai, Beijing and Shenzhen, SMIC has invested in the construction of three large-scale fabs to reserve capacity for the development of domestic chips.
The Shanghai facility is responsible for the 28nm process, 300mm wafer manufacturing line. The Beijing plant is responsible for mature processes of 40nm and above, and forms a connection relationship with the Shanghai plant, focusing on the capacity expansion of mature chips.
In addition, the manufacturing technology of the Shanghai and Beijing factories adopts the independently developed N+1 technology, which makes the transistor density about 1.8 times higher than that of the traditional 28nm chip. The Beijing factory also integrates the BCD process with SOI technology to develop the world’s first 40nm high-voltage power management chip compatible with 5G base stations, providing a strong chip product support for domestic 5G communication technology.
The Shenzhen plant is focused on more advanced chip manufacturing, mainly responsible for 14nm chips, and Fin FET transistor technology chip manufacturing.
According to the latest data, SMIC’s mature chip manufacturing plant has achieved a production capacity of 1 million chips on 28nm chips, which can meet 38% of the global mature chip market demand. Due to the strong demand for chips in China’s consumer market, domestic chip companies such as SMIC will first meet the demand of the domestic market, followed by the overseas market for “counterattack”.
Huawei’s flagship mobile phone has begun to use memory chips supplied by Changjiang Storage and Changxin Storage in the memory field, and the new Kirin chips are manufactured by a de-beautifying domestic supply chain.
And in the new energy vehicle el 3 supply CTP lithium iron phosphate battery, and also developed the vehicle gauge level IGBT module, reducing the cost of the car’s electric drive system by about 40%.
5.Huawei’s first self-developed PC processor Hisilicon Kirin X90 was exposed!
Huawei hisilicon Kirin X90, the first self-developed PC processor launched by Huawei Hisilicon, was first exposed on March 15, 2025!
According to the announcement of safety and reliability evaluation results released by China Information Security Evaluation Center (No. 1 in 2025), the Kirin X90 sent by Shenzhen hisilicon Semiconductor Co., Ltd. is listed, and the CPU has obtained the evaluation of safety and reliability level II. Also announced are Fei Tengyun S5000C-E, Loongson 3B6000, Loongson 3C6000, Shenwei Weixin H8000 and so on.
The Kirin X90 adopts advanced manufacturing technology, combined with Huawei hisilicon’s technical accumulation in the field of chip design, and its architecture design enables the processor to achieve excellent performance in multi-task processing and high-load computing.
The main frequency, number of cores and graphics processing units of the Kirin X90 have been significantly improved, providing users with more powerful computing power and a smoother user experience.
Kirin X90 performs well in terms of energy efficiency, allowing users to enjoy powerful performance while reducing battery consumption and extending the use time of smart devices.
Judging from the current information, the Kirin X90’s main target may be the PC sector. If it is combined with the all-self-developed Hongmeng PC system that Huawei is developing, it is expected to provide the core power for Huawei’s next generation of new PCS. Although it is currently mainly speculated that it is applied to the PC field, it is not ruled out that it may be expanded to other fields requiring high-performance processors in the future.
The emergence of Kirin X90 will further enhance Huawei’s competitiveness in the field of smart devices, especially in terms of security, power efficiency and multi-tasking capabilities, so that it has the capital to compete head-on with traditional CPU vendors such as Intel and AMD.
March 2025 is a sensitive time point, because Microsoft’s Windows operating system supply license to Huawei is about to expire, and the source pointed out that it has not been extended, which indicates that Microsoft is likely to stop the supply and say goodbye to Windows.
6.AI outbreak, how to break through local advanced packaging?
AI chips are the biggest growth point of semiconductors, and advanced packaging is the key technology for manufacturing AI chips. Previously, Nvidia’s H100 cost about $3,000, while HBM made in advanced packaging was worth $2,000. Once low-profile, back-channel technology has become the focus of competition in the industry.
In 2024, the global AI chip market will exceed $71 billion, and China’s AI chip market will be about $25 billion, accounting for about 35.2% of the world’s total. In 2025, with the landing of Deepseek and the outbreak of Chinese AI in the application side, the domestic AI chip market will have a high probability of breaking through this proportion, and the hot market will put forward more urgent requirements for domestic advanced packaging technology.
How is TSMC’s leading advanced packaging technology successful
First, advance layout, advance research and development
TSMC started developing 2.5D packages in 2012, 5-10 years earlier than most competitors, and achieved mass production as early as 2016. The 3D SoIC has also been in development since 2018 and entered mass production in 2024. This is the main reason why TSMC still has a firm grip on big customers like Nvidia and Apple.
Second, TSMC takes advantage of contract manufacturing to cut from the front to the back and form an efficient collaborative ecosystem. TSMC has established close alliances with equipment manufacturers (ASML), material manufacturers (Shin-Etsu Chemical), EDA (Synopsys), etc., forming a complete supply chain, seamlessly integrating advanced manufacturing process and advanced packaging, realizing the integration of technology and engineering, and upgrading its leading advantages to the level of industrial ecology, which not only improves research and development efficiency, but also builds stronger industrial barriers.
Third, TSMC relies on a solid cooperative relationship with customers to lock in large customers and develop new technologies according to demand. In 2017, TSMC stole Apple’s processor orders from Samsung with its InFO technology, and the cooperation between the two has made InFO an advanced technology widely used in the market. Since 2016, TSMC has exclusively contracted NVIDIA with CoWoS technology. TSMC’s short supply of production capacity has shaped today’s computing power market, which in turn has formed the cornerstone for NVIDIA GPU to rule the AI chip market.
Fourth, invest heavily in technology research and development. TSMC will spend US $6.389 billion on R&D for the full year 2024, with planned capital expenditure of US $38 billion to US $42 billion in 2025.
A series of challenges facing the domestic development of advanced packaging industry under sanctions
First, limited equipment. Advanced packaging is very dependent on equipment, such as the lack of advanced positioning bonding machine, hot pressing equipment and high-end testing equipment, which makes domestic enterprises can only do advanced packaging of chips above 7nm, and the yield and production capacity are also greatly affected.
Second, advanced packaging research and development investment is huge, domestic enterprises insufficient investment
Third, China’s advanced packaging started late, and it needs to pay the necessary time cost. TSMC’s advanced packaging began to lay out in 2012, and it has achieved the results of leading the advanced packaging market today. In addition, domestic enterprises also have to face a large number of patent barriers, as well as talent shortboards and other problems. The strengthening of these short boards requires corresponding time costs.
The domestic development of advanced packaging has certain comparative advantages
First, despite economic adjustment and trade war, the Chinese market is still the world’s largest consumer electronics market. By 2024, about 280 million domestic smartphones will be shipped, more than 10 million electric vehicles will be shipped, and China’s AI chip market will reach $28 billion.
Second, despite the lack of enterprises’ own funds, the state has provided strong policy support. Advanced packaging is a key technology supported by the third phase of the Grand Fund and the 14th Five-Year Plan.
Third, domestic packaging enterprises have a strong technical strength and industrial base. The country has close to the world’s leading sealing test capability, and the global ranking of the head enterprises is third and fourth. They have accumulated in FOWLP, 2.5D and other packaging technologies. The existing equipment is enough to support the research and development of low-end packaging, domestic enterprises have been able to complete 2.5D packaging, has been laid out FOPLP, vehicle level packaging and so on. These capabilities are enough to support them to quickly cut into the mid-end market, and gradually move towards the high-end.
The development of advanced packaging industry in China is as follows
First, concentrated resources, close coordination of major customers
Second, focus on the future, long-term layout of technological breakthroughs
With the support of industrial policy, domestic enterprises should also focus on the future, long-term layout, combined with the research and development process of domestic equipment, from the long-term, medium and near-term layout of different gradients of different advanced packaging. Consider both the urgent needs, but also consider the reserves of future technology, and gradually resolve the passive situation of technology following.
Third, straighten out the mechanism, OEM and packaging organic coordination. TSMC was not involved in the packaging business, but when the opportunity of advanced packaging appeared, it cut into advanced packaging through its own industrial position, and naturally formed a competition with packaging companies.
Fourth, layout equipment, accelerate the common growth of equipment, materials and production technology. TSMC and equipment enterprises jointly research and develop advanced packaging technology, and realize the synchronous development of advanced packaging production technology and equipment.
