Samil PwC: "K-Semiconductors Must Enter the AI Semiconductor Ecosystem"

Report Highlights Opportunities for K-Semiconductors in the Expanding AI Semiconductor Ecosystem
Three Key Areas for Growth: Next-Generation Memory, Foundry Services, and Neural Processing Units
Call for a "Semiconductor Team Korea" to Strengthen Industry Competitiveness

As global competition for dominance in AI semiconductors, often referred to as the "brains of artificial intelligence (AI)," intensifies, there are growing calls for the Korean semiconductor industry?which has been heavily focused on memory?to actively participate in the AI semiconductor ecosystem.

On June 10, Samil PwC announced the release of its report, "Semiconductors Powered by AI: Opportunities for K-Semiconductors to Leap Forward in AI." The report examines the opportunities available to domestic companies within the AI semiconductor supply chain and provides strategic directions for realizing these opportunities.

The report finds that, aside from memory products such as high-bandwidth memory (HBM), Korea has a limited presence in the AI semiconductor sector, where US-China technological competition is escalating. The report points out, "While domestic companies maintain an edge in high value-added memory demand, their competitiveness in the non-memory segment?which is three times larger than the memory market?is weak. As a result, they are unable to fully benefit from the rapid growth of the AI semiconductor market."

The report identifies three main areas within the AI semiconductor ecosystem where domestic companies may find opportunities: next-generation memory solutions such as Compute Express Link (CXL), foundry (semiconductor contract manufacturing) services, which are expected to see increased demand as companies move away from reliance on Nvidia, and neural processing units (NPUs), which are more energy-efficient than graphics processing units (GPUs). Among these, CXL is highlighted as a promising field where Korea could secure technological leadership in memory following HBM. CXL is a semiconductor interface technology that efficiently connects different types of devices, helping to overcome the limitations of existing memory. Major domestic companies such as Samsung Electronics and SK Hynix are accelerating the development and mass production of CXL-based DRAM.

The report also notes that, despite TSMC's overwhelming market share in the foundry sector, opportunities still exist for domestic companies. It states, "As big tech companies seek to reduce their dependence on Nvidia GPUs by developing their own chips, and as Chinese companies also intensify their own semiconductor development efforts, opportunities may arise for domestic companies to secure new foundry customers."

Furthermore, the report forecasts that if AI services become lightweight enough to run on edge devices, the NPU market will open up in the mid to long term. NPUs, which process big data using artificial neural networks, are considered next-generation technology that can address the inefficient computing performance and high power consumption of GPUs. Since no single company has yet emerged as a clear technology leader in the NPU field, many companies are actively pursuing technological development in this area.

Finally, the report proposes the formation of a "Semiconductor Team Korea" to serve as a public foundry and enhance the competitiveness of the domestic semiconductor industry. Citing TSMC's role as a public foundry during its early days, the report argues that a Korean-style public foundry could achieve economies of scale and support the commercialization of domestic fabless companies, thereby driving growth across the Korean semiconductor industry as a whole.

The report also emphasizes the need to pursue software development alongside hardware in order to create a lock-in effect for K-Semiconductors. As a representative example, Nvidia has expanded its AI semiconductor ecosystem through CUDA, a programming tool specialized for its GPUs. The report advises, "While securing GPUs is important for expanding AI infrastructure, in the mid to long term, government support for the development of fundamental technologies is essential for domestic semiconductor self-reliance."

Jung Jaeguk, Technology, Media, and Telecommunications (TMT) Industry Leader (Partner) at Samil PwC, stated, "Whereas the semiconductor industry has traditionally been viewed from the perspective of information and communications technology and manufacturing, it is now time to seek new growth engines in future promising fields such as autonomous vehicles and digital twins." He predicted, "AI semiconductor competitiveness will become the core infrastructure competitiveness spanning all industries."