Recently, TrendForce, a world-renowned market research firm, organized a seminar titled "The Global Expansion of Semiconductors in the AI Era: Big Predictions for the Technology Industry in 2025". In this seminar, experts had an in-depth discussion on the future development of a series of high-tech industries, including foundry services, high-bandwidth memory (HBM), NAND flash memory (Flash), AI power management integrated circuits (PMICs), AI servers, panel-level packaging technology, liquid cooling and heat dissipation solutions, and other high-tech industries.
Rapid advances in artificial intelligence (AI) are widely recognized as a key force in driving industrial transformation and reshaping the global economic landscape. It has had a broad and far-reaching impact on the development of several industries. In this prediction, TrendForce experts provide a detailed analysis of how AI technology interacts with the semiconductor industry and how it is expected to shape the future trends of the semiconductor industry. These discussions included the role of AI technology in improving the efficiency of chip design and manufacturing, driving the development of new storage solutions, enhancing the data processing capabilities of servers, optimizing power management and cooling technologies, and how to integrate AI capabilities into consumer electronics such as PCs. These analyses address not only the challenges and opportunities at the technical level, but also the impact of market demand, the competitive landscape of the industry, and macroeconomic trends on the semiconductor industry.
First. The output value of wafer foundries has increased, and the demand for AI application is large
The rise of AI applications has driven the urgent demand for high-performance computing chips for two consecutive years, making high-computing applications the main driving force for advanced process technology and the development of the entire wafer foundry industry. Starting in 2025, not only AI chip vendors and cloud service providers (CSPs) are developing their own chips, but memory vendors are also actively seeking collaboration with advanced process foundry partners to optimize HBM and logic chip compatibility. In this AI-driven technology race, the fab and its upstream and downstream IP development, design services, and back-end packaging and testing ecosystems have become indispensable resources, and the industry's focus is no longer limited to the advancement of front-end process technology.
From the perspective of the overall trend of the wafer foundry industry, in addition to the business opportunities brought by advanced processes, the demand for power management by AI may also inject new vitality into mature processes with stable long-term demand. At the same time, the development trend of cloud AI and edge AI in the wafer foundry industry in 2025 will also become a hot spot for the industry. As various application fields complete their two-year inventory adjustment cycle in 2024, TrendForce predicts that the global foundry industry is expected to grow by 20% in 2025. Among them, TSMC is expected to continue to maintain its leading position, while other foundries are also expected to achieve an annual growth rate of nearly 12%.
Second.AI demand for servers is growing
The rapid growth of the global AI market is driving significant growth in AI server shipments. In 2024, global AI server shipments are expected to grow by 42% in 2024, driven by strong demand from cloud service providers (CSPs) and brand customers for AI infrastructure. This growth momentum is expected to continue until 2025, when AI server shipments are expected to grow by about 28%, driven by demand from cloud operators and sovereign clouds, bringing AI servers to nearly 15% of the global server market. In the long run, with the continuous advancement of cloud-based AI training and inference application services, AI servers are expected to account for nearly 19% of the global server market by 2027. In terms of AI chip suppliers, it is expected that by 2024, NVIDIA's market share in the GPU market will be close to 90%, and its new Blackwell platform will gradually increase in volume in the first half of 2025 and become the mainstream of NVIDIA's high-end GPU shipments.
In addition, as AMD, Intel and other CSPs actively develop next-generation AI chips, it is expected that the shipment growth momentum will increase in 2025, which in turn will drive the shipment of CoWoS and HBM to double. These development trends show that the market for AI servers and related components is ushering in opportunities for rapid expansion, while also facing fierce market competition and technical challenges.
Third.AI server drives liquid cooling and heat dissipation
In the cloud computing space, leading players such as Google, AWS, and Microsoft are aggressively building new data centers around the world and accelerating the deployment of AI servers. As processor performance increases, so does its thermal design power (TDP), such as NVIDIA's GB200 NVL72 rack solution, which has a TDP of about 140kW, a high power level that requires liquid cooling technology to effectively cool down. It is expected that in the early days, liquid cooling technology will become mainstream by means of water cooling and air exchange. It is estimated that by 2025, with the widespread deployment of GB200 rack solutions, the penetration rate of liquid cooling technology for AI chips is expected to grow from 11% to 24% in 2024.
Google is one of the first companies to adopt liquid cooling technology in the field of high-end AI-specific chips developed by cloud service providers, while other manufacturers mainly rely on air-cooling systems. The growing focus on ESG (Environmental, Social, and Governance) by governments and regulators around the world has accelerated the shift from air-cooled to liquid-cooled cooling technologies. The penetration rate of liquid cooling technology is expected to increase year by year, which has prompted power and cooling solution providers and system integrators to increase their investment in the AI liquid cooling market, forming a new industry competitive landscape.
Fourth. HBN is an important component of AI servers
The high-bandwidth memory (HBM) market is currently in a phase of rapid growth, especially in the continuous deployment and upgrade of AI servers, where the demand for HBM as a key component is increasing. HBM plays a vital role in the increase in the computing power and memory capacity of GPUs, which also drives further upgrades in HBM's specifications and capacity. For example, Nvidia's Blackwell platform is expected to feature up to 192GB of HBM3e memory, while AMD's MI325 may support more than 288GB of capacity.
Due to the complexity of HBM's production process, the yield needs to be improved, which leads to higher production costs, resulting in an average selling price of about three to five times that of traditional DRAM products. With the gradual expansion of HBM3e's mass production and production capacity, its contribution to revenue is expected to increase quarter-on-quarter. This means that HBM technology not only technically meets the needs of high-performance computing, but also gradually shows its commercial potential in the economy, indicating that HBM will play an increasingly important role in the field of high-performance computing in the future.
Fifth. AI is applied to personal PCs
With the continuous advancement of artificial intelligence technology, AI functions are gradually becoming the standard configuration of laptops. The market penetration rate of AI-powered laptops is expected to reach 21.7% by 2025, and by 2029, this proportion will be close to 80%. This means that in the future, most laptops will be AI-capable, allowing users to provide users with smarter and more personalized services. The development of AI technology will greatly improve the user experience, especially in speech recognition and natural language processing. Users will be able to interact with the computer more intuitively through voice commands, enjoying a more natural and convenient operating experience. In addition, AI technology can provide more accurate business insights through data analysis and help enterprises make more informed decisions. Although the application form of AI technology is mostly known and highly dependent on cloud services, with the maturity of the technology and the improvement of market acceptance, as well as the growth of user demand for related products, the application prospect of AI in the field of notebook computers is still worth looking forward to.
In addition, the development of AI technology will also promote innovation in the laptop industry, bringing new opportunities to the mature and stable laptop industry. Breakthrough AI applications are expected to emerge in the future, providing consumers with more valuable options. These applications may include smarter personal assistants, more efficient workflow management, more advanced creative tools, and more, driving continued growth and innovation in the laptop industry.
