NVIDIA overtakes Apple to become TSMC’s largest customer, underscoring its dominance in the age of AI In January 2026, an event symbolic of the semiconductor industry’s power structure was reported: on the back of explosive demand for AI chips, NVIDIA overtook Apple, long TSMC’s largest customer, to take the top spot in terms of revenue contribution. The company has taken the top spot in terms of revenue contribution. TSMC CEO C.C. Wei said that AI demand is “real and just getting started” and predicted that server AI processor sales will grow by an average of 50% per year over the next five years. To meet AI demand, TSMC plans to increase capital expenditures from a record $38 billion to $42 billion in 2026, and will accelerate investment in advanced packaging technologies, especially 2nm process and CoWoS (Chip-on-Wafer-on-Substrate). In particular, it is accelerating investment in advanced packaging technologies such as 2nm process and CoWoS (Chip-on-Wafer-on-Substrate). Meanwhile, NVIDIA CEO Jensen Huang, while acknowledging the company’s extreme dependence on TSMC’s manufacturing capacity, has called for increased capacity throughout the supply chain to help resolve supply shortages. This move could affect the price negotiation power and supply priority that Apple has enjoyed so far as the most preferred customer, and TSMC has two huge growth engines, “AI” and “smartphones,” which further solidify its pricing power.
Memory Market Enters “ Super Cycle “; Supply Shortages and Prices Soar through 2028 The DRAM and NAND flash memory market has entered an unprecedented “super cycle,” driven by intense demand from AI data centers. Micron, SK Hynix, Samsung Micron warns that this supply shortage will continue until at least 2028, and DRAM prices are expected to rise by up to Some estimates suggest that DRAM prices could rise by up to 50% in the first quarter of 2026 alone. The impact has been far-reaching, with Samsung reportedly planning to double consumer SSD and DRAM prices, and Chinese smartphone makers Xiaomi and OPPO forced to cut their 2026 shipment targets due to soaring memory costs. Furthermore, Samsung’s distributors have reportedly raised prices by 80%, and this “chipflation” is beginning to seriously affect the purchasing power of the average consumer through price pass-through on end products.
U.S. and Taiwan Agree on Historic Tariff Pact, Conditioned on Massive Investment Led by TSMC The United States and Taiwan have agreed on a new trade pact to strengthen the supply chain for the semiconductor industry. The core of the agreement, known as the “Taiwan Model,” is that in exchange for the U.S. lowering its tariffs on Taiwan-made semiconductors from 20% to 15%, Taiwanese companies, including TSMC, will make manufacturing investments totaling $250 billion (approximately 40 trillion yen) in the United States. This agreement is the result of the Trump administration’s policy of returning to manufacturing and Taiwan’s geopolitical security needs. Specifically, TSMC will accelerate the construction of a factory in Arizona, and there are also reports of plans to add five more factories, for a total of more than 10 giant fab complexes. This would give the U.S. domestic production capacity for cutting-edge semiconductors, while Taiwan would gain access to the U.S. market and political backing. However, there is growing concern in Taiwan over the “hollowing out of the industry” due to the outflow of cutting-edge technology and talented human resources to the U.S. The question for the future is how to balance economic interests and national security.
HBM4 Development Competition Intensifies ; SK Hynix and Samsung Battle for Dominance of Next-Generation Memory The development competition over HBM4, the next-generation memory that determines the performance of AI chips, is intensifying between SK Hynix and Samsung. SK Hynix, which currently holds the lead in the HBM market, is expected to supply about two-thirds of HBM4 for NVIDIA’s next-generation GPUs, and leads its competitors in advanced technologies such as 16-layer stacking technology (16-Hi). Samsung, on the other hand, has taken a gamble by allocating a large portion of its latest 1c-nanometer DRAM production capacity to HBM4, with the aim of starting mass production in early 2026. Samsung has introduced “custom HBM” technology to customize logic dies according to customer requirements, and is strengthening its partnerships with TSMC and NVIDIA. Micron is also looking to expand its share of the HBM market and is accelerating capital investments in Singapore and the U.S. It is no exaggeration to say that the winner of this race will dominate the next-generation AI semiconductor market, as HBM4 is the key to dramatically improving the learning efficiency and inference speed of AI models.
China ‘s Domestic Production of Semiconductors Accelerates; “De-NVIDIA” and Self-Sufficiency Advances under U.S. Regulations Amid tightening U.S. export restrictions, China is moving at breakneck speed to become more self-reliant in its semiconductor industry. The local production ratio of semiconductor production equipment in China has already surpassed the 35% target set for 2025, and domestic manufacturers such as NAURA and AMEC are rapidly increasing their technological capabilities. In the field of AI chips, Huawei, Alibaba (T-Head), Baidu, and others are accelerating the development and mass production of domestically produced chips to replace NVIDIA products, and Huawei’s Ascend series and others are already taking over NVIDIA’s share. Furthermore, the Chinese government has secretly launched a national project known as the “Chinese version of the Manhattan Project” and is investing a huge amount of money in research and development for the domestic production of EUV lithography technology and HBM. While the U.S. is partially approving the export of NVIDIA’s latest H200 chip, China is pushing forward with building its own supply chain based on “technological decoupling,” and is expected to achieve significant domestic self-sufficiency by 2026.
Micron Expands Global Production Capabilities, Investing Simultaneously in Singapore, India, and the U.S. U.S. memory giant Micron is undertaking a massive expansion of its production capacity around the world to meet the explosive growth in demand for AI. In Singapore, the company has begun construction of a $24 billion state-of-the-art NAND flash memory plant, which is expected to be operational by 2028. In Gujarat, India, a back-end (packaging and testing) factory has begun commercial production, diversifying our supply base in Asia. Micron has also begun construction of a mega-fab in New York State that will house up to four manufacturing buildings and, together with its headquarters plant in Idaho, will dramatically increase domestic production capacity. These moves are part of a broader effort to reduce geopolitical risks. These moves represent an ambitious global strategy aimed at simultaneously diversifying geopolitical risk and increasing market share in the AI and data center markets.
Japan’s Rapidus and its photoelectric fusion technology are attracting global attention and becoming the hub of a next-generation ecosystem. Rapidus, a Japanese next-generation semiconductor manufacturer, is steadily preparing for mass production of 2nm chips in 2027. In addition to building a factory in Hokkaido, the company is reportedly working with Foxconn and Sharp to produce AI servers domestically, as well as collaborating with NVIDIA in GPU design and manufacturing. Also, Japan’s strength in optoelectronic fusion (IOWN/CPO) technology is attracting attention, with NTT, Sony, Kyocera, and others accelerating development. Photoelectric fusion is expected to dramatically reduce power consumption in data centers (by 70% according to NVIDIA’s calculations), and when combined with Rapidus semiconductors, Japan could become a key player in next-generation AI infrastructure. The government is fully supporting this move, and materials manufacturers such as JX Nippon Mining & Metals are also planning to invest in and supply Rapidus, thus creating an all-Japan ecosystem.
Intel’s Struggles and Strategic Shift, Separation of Manufacturing Units and Focus on Packaging Intel, the former semiconductor champion, continues to face difficult business conditions, posting a net loss of approximately $600 million in its fourth quarter 2025 financial results. The turnaround of its foundry business, which has been underway under CEO Pat Gelsinger, is halfway through, and the company is betting its fortunes on the launch of its cutting-edge “18A” process. At the same time, the company is partially slowing the pace of its manufacturing node expansion and shifting the focus of its strategy to “advanced packaging” technology, for which demand is rapidly increasing, and technology licensing to UMC and others. There are also reports that NVIDIA will outsource some of its chip production to Intel in 2028, which, if realized, would be a major turning point for Intel’s manufacturing division. However, there are still many challenges to a full revival, as the pivotal “AI PC” strategy is being fueled by memory supply shortages.
The race to commercialize next-generation technologies “glass substrates” and “CPO” is in full swing. Attention is rapidly turning to “glass substrates” and “CPO (Co-Packaged Optics)” as next-generation technologies to break through the limits of miniaturization in improving semiconductor performance. Compared to conventional organic substrates, glass substrates are superior in terms of flatness and heat resistance, and enable finer circuit formation and larger packages, so Intel, Samsung, SK Hynix, and BOE of China are rushing to develop and prepare for mass production. CPO, on the other hand, is a technology that converts electrical signals into optical signals for chip-to-chip connections, which improves data transfer speeds and reduces power consumption. Taiwanese supply chains and Japanese materials and equipment manufacturers (Canon, Tatsumo, etc.) are also developing products for these new technologies, and 2026 is expected to be a critical year for these technologies to move from the laboratory to the mass production line.
Trump Administration’s Tariff Policies Pressure Korean, Taiwanese Firms, Force Supply Chain Restructuring Protectionist trade policies by the U.S. Trump administration (or the influence of the incoming administration) are weighing heavily on semiconductor companies worldwide. In particular, the warning of “100% tariffs on companies that do not invest in the U.S.” poses a major threat to South Korean memory makers such as Samsung and SK Hynix. The U.S. is trying to create economic security and jobs by attracting AI chip and memory manufacturing facilities to the U.S. In response, the South Korean government and companies are struggling to respond, forcing them to increase investment in the U.S. and raise the ratio of local production. Taiwan is also facing the need to decouple from the Chinese market while strengthening its ties with the U.S. through tariff agreements, and political pressure is a major factor in determining companies’ investment decisions and supply chain arrangements.
AI Trends and Prospects for Application in the Manufacturing Industry in the Last Week of January 2026
