Semiconductor News 20260404

In today’s semiconductor news, we report on the vivid intersection of massive global investment and structural changes in the supply chain against the backdrop of the rapid expansion of AI infrastructure: from TSMC’s plan to build 12 plants in Arizona, USA, to Microsoft’s 1.6 trillion yen investment in its data hub in Japan, manufacturing and data infrastructure in certain The manufacturing and data infrastructure in certain regions is being strengthened at a rapid pace. On the other hand, as indicated by the 50% rise in SSD prices and the price hike of FC-BGA substrates by Samsung Electro-Mechanics, intense inflationary pressure due to tight supply of materials is becoming apparent. Furthermore, geopolitical decoupling is accelerating the shift to proprietary technological hegemony, as evidenced by the share of domestically produced AI chips in the Chinese market surpassing 40%.

Main News

1. TSMC to Build 12 Semiconductor Fabs and 4 Packaging Facilities in Arizona, USA
   According to press reports, TSMC is moving forward with a major plan to build 12 semiconductor manufacturing plants and 4 packaging facilities in Arizona, USA. The move is believed to be part of Taiwan’s $50 billion investment plan for the U.S. and is seen as a move to restructure the supply chain to avoid geopolitical risks and to firmly expand production capacity for advanced semiconductors and AI chips in the U.S.
2. Microsoft to invest a massive 1.6 trillion yen to expand AI data centers in Japan
   Microsoft has announced a 1.6 trillion yen investment to build data centers and other AI infrastructure in Japan, in collaboration with Softbank, Sakura Internet, and others. This is in response to the rapid increase in demand for computing resources and data processing that will accompany the spread of generative AI. Full-scale infrastructure investment in the Japanese market by foreign IT giants will strongly drive the semiconductor and telecommunications ecosystems in the country.
3. Domestically Produced Companies’ Share of China’s AI Chip Market Surpasses 40%; NVIDIA Retreats
   The share of domestically produced semiconductor manufacturers in China’s AI chip market has surged to a level of over 40%. This is the result of the rapid development of alternative chips in China against the backdrop of strict U.S. export controls, and a clear retreat in the market share of NVIDIA, which has dominated the market until now. This highlights the reality that the decoupling between the U.S. and China is accelerating China’s semiconductor ecosystem independence.
4. SSD prices soared 50% in the January-March period due to a surge in demand for AI applications, underscoring a clear supply crunch
   Generation SSD prices soared 50% quarter-on-quarter in the January-March period due to a surge in demand for high-capacity storage in conjunction with the rapid expansion of AI infrastructures. This is causing a serious supply crunch, especially for enterprise SSDs for AI servers, and while it is a strong profit-improving factor for memory makers, it is also putting strong manufacturing cost pressure on set makers.
5. SEMI Forecasts Semiconductor Equipment Investment to Exceed $150 Billion in 2027
   The International Semiconductor Equipment and Materials Institute has released its latest forecast that global semiconductor equipment investment will surpass $150 billion in 2027, driven by strong demand for AI semiconductors and governments’ efforts to make supply chains self-sustaining. The strong demand for AI semiconductors and governments’ massive subsidy policies aimed at making the supply chain more independent and bringing the industry back to Japan are expected to create a strong tailwind that will create special demand and a boom for equipment manufacturers and related component suppliers in the medium to long term.
6. PSMC Collaborates with European CEA-Leti and Others on Silicon Photonics for Next-Generation AI
   PSMC, a Taiwanese foundry, has collaborated with European research organizations CEA-Leti and CEA-List to develop RISC-V architecture, MicroLED and silicon photonics for next-generation AI in a multi-year agreement to integrate 3D stacking, a groundbreaking technology development partnership to eliminate data transfer speed and power consumption bottlenecks faced by AI data centers.
7. Samsung Electro-Mechanics Raises Prices of FC-BGA Substrates Due to Supply Shortage Caused by AI Demand
   Samsung Electro-Mechanics has decided to raise prices of its high-end FC-BGA package substrates for AI semiconductors, for which demand is exploding. This is a direct result of the fact that the company’s production capacity for high multilayer substrates, which are essential for advanced packaging technology, is not keeping up with the strong market demand, causing a chain of inflation that spills over into higher manufacturing costs for semiconductor chip makers and end products.
8. Next-Generation Memory HBM4 Verification to be Completed in Q2, Establishing Supply Structure of Three Major Memory Makers
   It has been reported that the quality verification of the next-generation broadband memory HBM4, which will greatly influence the performance of AI chips, is expected to be completed by customers in the second quarter. This will solidify the framework for full-scale mass production and supply by the three major memory makers, SK Hynix, Samsung Electronics, and Micron, and will bring the fierce development competition among memory companies for the next-generation dominance of AI infrastructure into a new phase.
9. Murata Manufacturing, Daifuku, and other semiconductor-related infrastructure plants in Japan completed construction one after another.
   A new production building at Izumo Murata Manufacturing was completed in Izumo City, Shimane Prefecture, putting in place a mass production system to meet the growing demand for multilayer ceramic capacitors. Daifuku has also completed construction of a new plant at its Shiga Works to produce transport and storage systems for semiconductor production lines.
10. The announcement of the new TurboQuant technology upset memory stocks, but the demand outlook remains strong
    The announcement of TurboQuant, a new algorithmic technology that has the potential to significantly reduce inference memory demand for AI models, caused a temporary shakeup in the stock market for memory-related companies. However, analysis by industry analysts and others suggests that the market’s concerns are limited, as the bullish outlook for memory demand as a hardware, backed by the rapid expansion of the overall AI infrastructure, remains unchanged.

News List