South Korea AI: Is the AI story beyond memory compelling?

• South Korea's AI investment case extends well beyond memory chips, and the market has not yet priced in the earnings potential of its broader ecosystem.

• Physical AI is the most compelling layer — Korea's manufacturing base lets Hyundai, LG, and Rainbow Robotics commercialise faster and convert progress into earnings.

• Sovereign models secure data self-sufficiency, while a domestic data-centre buildout — SK Telecom–Nvidia's 2GW project — deepens rather than replaces the memory case.

• KOSPI's forward P/E sits below its historical average, and implied upside reached 66.6% as of 12 August 2026 after a recent pullback.

• Five major constituents span consumer discretionary, industrials, and communication services — roughly 4.9% of the index — giving KOSPI a diversified, memory-independent driver.

When discussing South Korea's AI development, investors tend to focus on memory chip production. SK Hynix and Samsung Electronics, which together hold over 70% of the global HBM market, have driven much of South Korea's export and stock market gains. However, should sentiment on the memory cycle shift, both stocks would be directly hit, dragging down the broader Korean market. In fact, while other listed South Korean companies have smaller market capitalizations and less eye-catching gains than Samsung Electronics and SK Hynix, their AI initiatives are an integral part of South Korea's technology ecosystem and are equally worth understanding for investors looking to build exposure to South Korea's AI theme.

Furthermore, as we noted in our second-half South Korea outlook, the comprehensive AI positioning of South Korean technology companies should allow them to fully benefit from the rollout of AI applications. Although the KOSPI may stay volatile in the near term, the AI applications discussed below could provide a new catalyst for Korean equities over the longer term.

Related article: Upgrade to 4 Stars: Can Narratives Beyond Memory Become the New Catalyst for South Korea Market?

South Korea's double halt: Do South Korea's fundamentals still hold after the storm?

AI Models: Not Their Strength, But Still Necessary

Unlike the US and China, South Korea's AI models attract less attention and score lower on performance. According to Artificial Analysis, both HyperCLOVA X from Korean internet company Naver (035420.KS) and K-EXAONE from LG score below their Chinese and US counterparts, while Motif 3, developed by Korean AI startup Motif Technologies, ranks only in the middle of the pack. This suggests that South Korea's strategic AI advantage does not lie in large language models.

Figure 1: Artificial Analysis AI Index Ranking

Source: Artificial Analysis
Data as of 13 August 2026.

That said, despite lagging other countries, South Korea still treats AI model development as an important strategy. Under government backing, the Ministry of Science and ICT (MSIT) leads a sovereign AI foundation model project, selecting elite teams from various companies to build internationally competitive AI models. The core of the program lies in the word “sovereign” — ensuring technological and data self-sufficiency and thereby reducing reliance on foreign models. Although a performance gap remains, having a sovereign model is still important for data security reasons, particularly for a country seeking to integrate AI into government functions and key industries. In addition, a foundation model is a necessary condition for the orderly operation of domestic AI infrastructure. HyperCLOVA X, for example, is trained primarily on the Korean language, giving it a better grasp of Korean grammar and cultural context than foreign models, and making it better suited to serving local use cases.

AI Infrastructure: The Foundation for All Applications

Without sufficient computing power and electricity, neither AI models nor physical AI can be deployed, which is why investment in AI infrastructure by the South Korean government and corporations is becoming increasingly important.

South Korean telecom operators and electronics manufacturers have already begun building infrastructure. SK Telecom (NYSE: SKM) has announced a partnership with Nvidia to build a large-scale data center in South Korea with capacity of up to 2 gigawatts, using Vera Rubin chips paired with SK Hynix memory. Samsung SDS (018260.KS) is leading the national AI computing center project, building data centers together with Kakao (035720.KS), KT (NYSE: KT), and Naver. Separately, companies such as Doosan Enerbility (034020.KS) and HD Hyundai Electric (267260.KS) are responsible for supplying the power equipment for these facilities.

Beyond laying the groundwork for downstream AI applications, building this infrastructure is itself boosting demand for AI servers and memory chips. While the market currently focuses mostly on how capital spending by overseas hyperscale cloud providers drives South Korean memory sales, domestic demand should not be underestimated either. The national AI computing center project, for instance, is part of South Korea's “AI highway” plan, and the government's backing offers a degree of protection for the long-term demand faced by both memory suppliers and data center builders. Seen this way, for SK Hynix and Samsung Electronics, the “South Korea AI story beyond memory” is not a substitute for memory as an investment opportunity, but rather a piece that makes the South Korea memory investment case more complete.

Physical AI: Maximizing South Korea's Advanced Manufacturing Edge

Compared with AI models and infrastructure buildout, we believe the investment case for physical AI applications may be more compelling.

Beyond semiconductors, South Korea's manufacturing capabilities in autos, home appliances, and displays are also highly advanced. This long-accumulated technological strength means that in advancing physical AI, South Korea does not need to rebuild its production systems, supply chains, or sales channels from scratch, and can instead fully leverage its manufacturing base to upgrade existing industries. In addition, since some companies' core businesses are facing considerable challenges — for example, Hyundai Motor (005387.KS) and LG Electronics (066570.KS) must compete with Chinese rivals in the auto and appliance markets, which may squeeze their room for expansion — advancing physical AI could represent a new growth opportunity for these companies.

Hyundai Motor is a case in point. After completing its full acquisition of Boston Dynamics, Hyundai Motor Group's entire Atlas robot production capacity for this year has already been allocated to Hyundai and Kia's own plants, and the company also plans to build a factory in the US by 2028 capable of producing 30,000 robots a year. This reflects years of accumulated capability that has given Hyundai the ability to carry out large-scale global factory deployment, allowing it to move its physical AI strategy quickly from the lab to mass production, and thereby secure an early market position and showcase its technological edge. LG Electronics, meanwhile, is building on its home appliance and smart factory experience to create an AI ecosystem through the production of household robots, integrating physical AI into its existing core business as a new growth driver. Rainbow Robotics, listed on the KOSDAQ, focuses on core components for humanoid robots and is helping its largest shareholder, Samsung Electronics, advance its robotics business.

Worth noting that, compared with the infrastructure and large models discussed above, the path to commercial monetization is clearer for physical AI. South Korea already has a production base capable of absorbing advanced manufacturing technology, along with long-established sales channels and brand recognition. Companies developing physical AI therefore do not need to build the underlying foundation from scratch, allowing their commercialization progress to be validated more quickly and reflected directly in financial results — providing new support for KOSPI earnings.

Investment Implications

The KOSPI is currently still highly dependent on memory chips as its main driver, and its valuation is also largely tied to the outlook for the memory cycle. However, as shown in the analysis above, South Korea's broader AI development and the earnings potential of related companies still have meaningful room to grow, and this part of the outlook has not yet been fully digested by the market — suggesting KOSPI valuations may be underestimated. More importantly, the companies mentioned above — Hyundai Motor, Kia, Naver, Doosan Enerbility, and LG Electronics — are all significant KOSPI constituents, with market caps ranking among the top 25 of all index constituents and together accounting for approximately 4.9% of the index. Most of these companies are also spread across different sectors, including consumer discretionary, industrials, and communication services. Compared with the highly concentrated memory story, the broader AI application story is markedly more diversified and less correlated, providing KOSPI with a second growth driver that does not rely entirely on the memory cycle.

The forward P/E ratio of KOSPI is currently well below its historical average, leaving even more room for upside ahead. We are therefore maintaining our earnings and target price forecasts for South Korean equities; as of 12 August 2026, the implied upside, following a recent pullback, has risen to 66.6%.

Figure 2: 12-Month Forward P/E of KOSPI

Table 1: KOSPI Forecast Earnings and Implied Upside              

	2025A	2026E	2027E	2028E
EPS (KRW)	248.9	627.6	787.2	842.9
EPS growth rate	19.4%	152.2%	25.4%	7.1%
P/E ratio	26.4	10.5	8.4	7.8
Dividend yield	1.0%	2.6%	3.2%	3.5%
Target Price at the end of 2028 (Based on 13x Forward P/E)				10,958
Potential upside				66.6%
Source: Bloomberg L.P., iFAST Compilations. Data as of 12 August 2026.