AI Semiconductor Market to Reach KRW 1,200 Trillion... "Next 3-5 Years Will Be Critical for Korean Companies' Survival"

EXIM Bank Economic Research Institute:
“Korean Firms Have Technology, But Lag in Commercialization and Ecosystem Development”


NPU Opportunities Expand as Inference AI Spreads


“Government Must Spur Early Demand and Support Overseas Supply Chain Entry”

As global competition intensifies over the data center artificial intelligence (AI) semiconductor market, which is expected to grow to KRW 1,200 trillion by 2030, experts believe that the next three to five years will be a crucial turning point for the survival of Korean companies. It is pointed out that the government should foster early demand in the domestic market and support domestic fabless players with technological capabilities in breaking into overseas AI supply chains.

According to the Overseas Economic Research Institute of the Export-Import Bank of Korea on September 27, the market for AI semiconductors for data centers is projected to grow from USD 124 billion (approximately KRW 170 trillion) in 2024 to USD 860 billion (approximately KRW 1,200 trillion) by 2030. The average annual growth rate is expected to reach 38%.

Nvidia currently dominates the market. As of last year, Nvidia held a 78.2% share of the data center AI semiconductor market, far ahead of Google (4.7%), AMD (4.1%), Intel (3.7%), and Huawei (2.6%).

However, the institute notes that as the AI market shifts its focus from "training" to "inference," latecomers may find new opportunities. As generative AI services become more widespread, demand is rapidly increasing for inference operations that generate answers to actual user inquiries, and the burden from semiconductor prices and electricity costs in data center operations is also rising.

In particular, neural processing units (NPUs), which have lower prices and power consumption compared to Nvidia's graphics processing units (GPUs), are gaining attention as alternatives. NPUs are semiconductors specialized for AI inference, and domestic companies are intensively targeting this market.

Leading the field are FuriosaAI and Rebellions. FuriosaAI, founded in 2017, began mass production of its second-generation NPU, “Renegade (RNGD),” earlier this year. Rebellions, launched in 2020, is also scheduled to release its second-generation product, “REBEL 100,” in the second half of this year.

The main issue is commercialization experience and ecosystem development. The institute assessed that domestic companies have secured chip design technology, but still lag global leaders by at least three years in terms of capital strength, customer base, software ecosystem, and practical data center deployment experience. Since success in the AI semiconductor market depends not only on chip performance, but also on software compatibility, stable supply capabilities, and customer usage experience, it is crucial to secure reference customers as early as possible.

In this context, the government is advised to expand its role beyond supporting technology development to “supporting market entry.” Providing tax benefits or subsidies to companies adopting Korean-made AI semiconductors would help mitigate the risk of early adoption. Overseas, it is important to strengthen government-to-government cooperation to help domestic companies gain access to local data center supply chains.

The Middle East is being highlighted as a key target market. With countries in the Middle East expanding their investment in AI infrastructure, there is rising demand for alternative suppliers due to the risks of U.S. AI semiconductor export controls, the high prices of Nvidia GPUs, and limitations in power and available sites.

The institute stressed that as the structure of major players and supply chains in the global AI semiconductor market could become largely fixed within the next three to five years, it is vital for Korean companies to gain actual customers and commercialization experience as quickly as possible during this period.

This content was produced with the assistance of AI translation services.