May 21, 2025 5 min read AI

Japan’s Semiconductor Lag Exposes a Two-Speed Global Market

Sign up for ARPU: Stay ahead of the curve on tech business trends.

Japan is investing billions to rebuild its semiconductor industry, aiming to bolster global supply chain resilience and reclaim a leading position. Yet, a report from Nikkei Asia reveals a surprising paradox: only three of seven new chip plants built or acquired in fiscal years 2023 and 2024 have begun mass production as of April. The reason? Sluggish demand for chips used in applications other than artificial intelligence. This unexpected lag starkly highlights that the global semiconductor market, often painted with a broad brush of “boom,” is in fact operating at two distinct speeds.

Why are some chip markets booming while others are struggling?

The primary engine of growth and demand in the current semiconductor market is artificial intelligence. Specifically, the rapid development and deployment of large language models and generative AI applications require immense computational power, driving unprecedented demand for high-performance processors like AI accelerators (GPUs), advanced CPUs, and specialized custom silicon. These chips typically rely on the most cutting-edge manufacturing processes (advanced nodes like 3nm and 2nm) and require sophisticated packaging and high-bandwidth memory, pushing the boundaries of the supply chain.

In stark contrast, demand remains relatively weak in more traditional semiconductor markets. These include chips for personal computers, smartphones, consumer electronics, and some industrial applications, often manufactured on older or “mature” process nodes (like the >40nm nodes Japanese companies are currently limited to, or the 12nm process at the new TSMC Japan fab). Sales in these segments have been cyclical or slowed, leading to inventory backlogs and lower capacity utilization rates globally. While the AI segment is experiencing exponential growth, the traditional segments still account for a large portion of the overall chip market, and their current softness impacts manufacturers focused on these areas.

Who benefits from the AI-driven demand?

The AI boom disproportionately benefits companies at the forefront of high-performance computing and advanced manufacturing. NVIDIA, with its dominant market share (around 95% of the AI GPU market), is the prime example, seeing its revenue and market value surge. TSMC, as the leading foundry for cutting-edge chips, is essential for manufacturing these high-end processors for fabless designers like NVIDIA, Apple, and AMD, leading to their advanced node capacity being fully booked despite rapid expansion. Equipment providers like ASML, which hold a monopoly on advanced lithography needed for the latest nodes, are also critical enablers of this boom.

Furthermore, the massive demand and high costs associated with AI hardware are prompting major tech companies (hyperscalers like Google, Amazon, Microsoft, Meta) to invest billions in developing their own custom AI chips optimized for specific workloads, often manufactured at leading foundries, adding another layer to the demand for advanced capacity.

What parts of the industry are facing weak demand?

The Japan news provides concrete examples of segments currently facing headwinds. Renesas is seeing sluggish demand for power semiconductors used in electric vehicles and industrial applications, delaying mass production at a reopened plant. Rohm and Sanken Electric have also pushed back the start of full-scale production at acquired or new facilities focused on power semiconductors. Kioxia, a major memory chip producer, delayed bringing a new fabrication facility online until later in 2025, waiting for the memory market to recover. Sony Group, strong in image sensors for smartphones, is cautiously adding equipment to a new fab due to slower iPhone sales and competition from Chinese handset makers switching to local suppliers.

These examples reflect a broader trend. Globally, demand weakness outside of AI has resulted in average fab capacity utilization rates hovering around 60% to 70% in 2024, significantly below the 80% to 90% considered healthy for the industry. This disparity between surging demand in the advanced AI segment and muted demand elsewhere is creating challenges for companies whose primary focus or current investments lie outside the AI spotlight.

What does this “two-speed” market mean for chipmakers and government strategies?

The divergent market conditions complicate investment decisions for chipmakers and highlight challenges for government strategies aimed at building domestic capacity. While governments, including Japan’s (~10 trillion yen in support by FY2030, ~9 trillion yen expected investment), are providing significant funding for new fabs to enhance economic security and resilience, the Japan case shows that simply building facilities is not enough. For these fabs to be profitable and contribute meaningfully to global supply, there must be sufficient market demand for the specific types of chips they produce.

Regions or companies primarily focused on segments currently experiencing weak demand, or those significantly lagging in the advanced nodes (<7nm) required for cutting-edge AI chips, face the risk of new fabs remaining underutilized or delaying operations, despite geopolitical impetus. Even TSMC’s first fab in Japan, focused on 12nm and 16nm nodes, is reportedly not utilizing its full capacity yet, and the start of its second, more advanced fab (aiming for mass production in FY2027), was pushed back. This suggests that bridging the technology gap and aligning strategic investments with viable market demand across all segments remains a challenge in the two-speed semiconductor world.