1. Chip giants game four key markets!

According to TrendForce data, the world’s top 10 chip designers will have a total revenue of approximately US$249.8 billion in 2024, with the top five vendors contributing more than 90% of their total revenue. At present, IC design giants such as NVIDIA, AMD, Qualcomm, and MediaTek are actively deploying around the four key markets of mobile phones, AI PCs, automobiles, and servers; At the same time, with the rising demand for high-performance chips driven by AI and increasingly fierce market competition, the trend of industrial chain collaboration and cooperation is obvious.

With the intensification of competition in the semiconductor industry, the performance of AI-driven chips has been greatly improved, and the collaborative development trend of packaging, foundry and design in the semiconductor industry is obvious, so as to continue to explore the limits of chip architecture and performance.

At present, the collaboration between chip design companies and wafer foundries is the most basic model, and chip design companies such as Nvidia, Qualcomm, AMD, and MediaTek need TSMC, Samsung, GF, UMC and other wafer foundries to produce chip products. With the development of technology, the cooperation of the semiconductor industry is moving towards deeper collaborative optimization, especially in advanced node and packaging technology.

Traditional monolithic system-on-chips (SoCs) are approaching their physical limits in terms of scaling, and their design and manufacturing costs and difficulties are increasing dramatically. To continue to improve performance and integration, the industry is turning to more advanced solutions such as 3D integrated circuits (3D-ICs), chiplets, and heterogeneous integration. These technologies require co-optimization of expertise across multiple domains, including design, process, packaging, and systems. For example, achieving high-density interconnects and managing the thermal effects of complex multi-chip systems requires close collaboration between design houses, foundries, and packaging fabs.

In addition, the explosive growth of AI and HPC applications has placed extremely high demands on dedicated, high-performance, and energy-efficient chips (GPUs, NPUs, and AI accelerators). Downstream system vendors (e.g., cloud service providers, automakers) are increasingly looking for custom or semi-customized chip solutions for specific applications, which directly drives closer working relationships between chip design companies, foundries, and even end customers, and achieving co-optimization of systems and technologies is the key to meeting these needs.

Challenges and opportunities coexist in the global chip market, chip design, as the core of the semiconductor industry chain, is ushering in a new round of technological and ecological model changes, this war without gunpowder is being waged in the fields of mobile phones, AI PCs, automobiles and servers, technological innovation, product iteration and industrial chain collaboration are important “weapons” for manufacturers. Under the game of giants, what new changes will be ushered in in the chip design and semiconductor industry chain? We’ll see.

2. DRAM and NAND Flash market forecast for 2025

DRAM memory

Market size: According to relevant reports, DRAM revenue will reach $90.7 billion in 2024, a year-over-year increase of 75%, and the revenue is expected to be $136.5 billion in 2025, an annual increase of about 51%.

Growth factors: First, the demand for AI servers has exploded, and the demand for high-performance computing in data centers has surged, and AI servers require a large number of high-performance DRAM. The second is the recovery of demand for PCs and mobile devices, the gradual recovery of the global economy, the increase in consumer demand for PC and mobile device replacement, and the increase in DRAM configurations by manufacturers. Third, the tariff policy triggers the demand for partial inventory replenishment. Fourth, the rise of high value-added products such as HBM, which will account for 5% of DRAM bit shipments and 20% of revenue in 2024, and the market share of DDR5 and other products is also increasing.

Market Structure: In the first quarter of 2025, SK hynix topped the global DRAM market with a market share of 36.7% due to its absolute dominance in the HBM field, ending Samsung’s market dominance for more than 40 years.

NAND Flash

Market size: NAND Flash revenue will be $67.4 billion in 2024, up 77% year-over-year, and is expected to reach $87 billion in 2025, up 29% year-over-year.

Growth Factors: First, the rise of high-capacity QLC enterprise-class solid-state drives (SSDs), which have been widely adopted by cloud computing service providers in North America in inference AI servers. The second is the application of QLC UFS in smartphones, which some Chinese smartphone manufacturers will start adopting in the fourth quarter of 2024, and Apple expects to incorporate QLC into the iPhone in 2026. Third, manufacturers’ capex constraints have suppressed supply, and server demand has recovered.

Price Trends: In the second quarter of 2025, NAND Flash prices will stop falling and recover, with client SSD prices increasing by 3-8% QoQ and NAND Flash wafers increasing by 10-15%. In the sub-categories, Client SSDs have become the main growth point due to the recovery of the PC market, the prices of Enterprise SSDs are flat, and the prices of embedded storage such as eMMC/UFS are stable.

3. China must fight back, and it is not only Huawei’s Ascend chips that the United States has blocked!

On May 13, 2025 local time, the U.S. Department of Commerce officially issued a document to repeal the Biden administration’s AI proliferation rules, and at the same time announced three additional policy measures to strengthen export controls on global semiconductors, and stipulate that the use of Huawei’s Ascend chips anywhere in the world violates U.S. export control regulations.

The following are the general contents of the three policy measures:

1). Huawei Ascend chips are banned globally

Policy content: The U.S. Department of Commerce issued guidance prohibiting the use of Huawei’s Ascend chips (including the Ascend 910B, Ascend 910C, and Ascend 910D) anywhere in the world. Any act involving the sale, transfer, export, re-export, financing, ordering, purchasing, concealing, storing, using, lending, disposing of, transporting, transshipment, installation, maintenance, repair, modification, design, development, etc., of these chips is considered a violation of U.S. export controls.

Impact: This measure aims to limit China’s development in the field of artificial intelligence and high-performance computing, and further strengthen the technological blockade of Chinese high-tech enterprises.

2). Restrict the use of U.S. chips in Chinese AI models

Policy content: The United States issued guidance warning the public about the potential consequences of allowing American AI chips to be used to train and reason about Chinese artificial intelligence models. If U.S. AI chips are used to train or interfere with Chinese AI models, companies will be severely punished. This is obviously suppressing China’s DeepSeek large model.

Impact: This measure further restricts U.S.-China technological exchanges and cooperation in the field of artificial intelligence, and attempts to prevent China from using U.S. technology to develop its own AI capabilities.

3). Strengthen supply chain scrutiny

Policy content: The U.S. issues guidance to its own companies on how to protect their supply chains from transshipment strategies. U.S. companies are required to reassess their partners to guard against technology transfer risks.

Impact: This measure is designed to prevent Chinese companies from bypassing U.S. export controls through supply chains, further tightening the technology blockade against China.

The United States’ move has three deep meanings, one is to “decapitate” Chinese artificial intelligence – knock out Huawei chips” and draw salaries from the bottom of the kettle, and the other is to open the way for selling American chips The follow-up estimate is to allow the castrated version of Nvidia chips to be sold to China, which is a routine with the United States blocking Huawei 5G chips and then selling 4G Qualcomm chips to Huawei mobile phones! The third is to continue to implement the strategy of the president of Microsoft— so that American AI can be used by the whole world, in an attempt to continue to maintain the hegemony of the United States. Fourth, we will test China’s response to the extension of the long-arm jurisdiction of the United States, and continue to test China’s bottom line.

This regulation of the BIS of the United States is not only a new regulation to block China’s computing chips, but also a wanton trampling on China’s sovereignty! In this regard, China must fight back!

4. Jensen Huang: The next wave is physical artificial intelligence

On May 19, it was reported that Nvidia CEO Jensen Huang attended the Taipei International Computer Show (COMPUTEX 2025) in Taiwan, China, and delivered a keynote speech.

In his speech, Huang began by highlighting Nvidia’s transformation from a chipmaker to a leader in AI infrastructure. He noted that AI and accelerated computing are reshaping the computer industry, driving the “Fourth Industrial Revolution” and expanding from the cloud to edge computing, transforming areas such as data centers, robotics, and autonomous driving.

Huang detailed NVIDIA’s progress in AI infrastructure, including building data centers, developing specialized libraries (e.g., cuQuantum, cuDSS, etc.) to accelerate adoption in different domains, and advancing software-defined telecommunications. He also showcased AI-powered ray tracing for GeForce RTX 50 Series graphics cards, and emphasized the centrality of CUDA and related libraries in accelerated computing.

In addition, Huang introduced NVIDIA’s breakthroughs in AI inference and generative AI, including the development of inferential AI, physical AI, and agent-based AI, as well as how high-performance computing is enabled with the Grace Blackwell system. “Intelligence is much more than what you learn from a lot of data, and agent-based AI is basically understanding, thinking and acting, and it’s a digital form of robot,” he said. These will become very important in the coming years. He also announced plans to build giant AI supercomputers in partnership with Foxconn and TSMC, and launched NVLink Fusion technology to support the construction of semi-custom AI infrastructure.

Huang also showcased several new products, including DGX Spark and DGX workstations, which are designed to provide developers and businesses with powerful AI computing capabilities. He also introduced the RTX Pro enterprise server and the Omniverse platform, highlighting the promise of AI in enterprise IT and how digital twins can advance industrial automation and robotics.

5.SEMI Report: The semiconductor industry in the first quarter of 2025 is typically seasonal, but atypical changes may occur due to tariffs and other impacts

Despite heightened trade policy risks, current data for the first quarter of 2025 suggests that electronics and integrated circuit (IC) sales are not directly affected by the new tariffs. Electronics sales in the first quarter of 2025 fell 16% sequentially and were flat year-over-year, in line with traditional seasonal patterns. IC sales declined 2% sequentially but increased significantly by 23% year-over-year, reflecting continued investment in artificial intelligence and high-performance computing infrastructure. “While electronics and IC sales in the first quarter of 2025 were not directly impacted by the new tariffs, uncertainty over global trade policies has prompted some companies to speed up shipments while others pause investments, and this push-pull dynamic could lead to atypical seasonality in the industry for the rest of the year as the industry adjusts to changing supply chains and tariff patterns,” said Clark Tseng, senior director of market analysis at SEMI. “Semiconductor capital expenditures (CapEx) decreased 7% sequentially but increased 27% year-over-year as manufacturers continued to invest heavily in advanced logic, high-bandwidth memory (HBM) and advanced packaging to support AI-driven applications. In the first quarter of 2025, memory-related CapEx surged 57% year-over-year, while non-memory-related CapEx increased 15% year-over-year, highlighting the industry’s focus on innovation and resilience.

Fab equipment (WFE) spending grew 19% year-over-year in Q1 2025 and is expected to grow another 12% in Q2, driven by strong investments in advanced logic and memory production that support the rapid adoption of AI semiconductors. Test equipment orders grew 56% year-over-year in the first quarter and are expected to grow 53% in the second quarter, reflecting the increasing complexity and stringent performance requirements of AI and HBM chip testing.

Packaging and test equipment also saw double-digit growth, benefiting from the industry’s push for higher-density integration and advanced packaging solutions. Boris Metodiev, Director of Market Analysis at Tech Insights, said: “The WFE market is poised for steady growth, driven by government investment and advances in semiconductors, particularly in the areas of artificial intelligence and emerging technologies. However, geopolitical uncertainties, including export restrictions and potential tariffs, could affect this positive trajectory. “In line with the growth in capital equipment investment, global fab capacity is ramping up and is expected to exceed 42.5 million wafers per quarter (in 300mm wafer equivalent) in the first quarter of 2025, up 2% sequentially and 7% year-over-year. Chinese mainland continues to lead capacity expansion among all regions, although growth is expected to slow in the coming quarters. Notably, Japan and Taiwan recorded the strongest quarterly capacity growth, driven by significant investments in power semiconductor manufacturing in Japan and capacity ramp-up at leading foundries in Taiwan.

Looking ahead, SEMI and Tech Insights expect an atypical seasonal pattern in the industry in 2025 as companies navigate the dual challenges of trade policy uncertainty and supply chain adaptation. While demand for AI and data center technology remains a bright spot, other areas may see investment delays or demand shifts due to the market’s response to evolving tariffs and geopolitical uncertainty.