World's top HBM maker, but still chasing the AI big leagues

President Lee Jae-myung's "San Francisco AI Declaration" has put a spotlight on South Korea's current standing in artificial intelligence, as the government unveiled a blueprint to develop the country into a global AI production hub.

South Korea boasts world-class competitiveness in high-bandwidth memory chips, AI patents and manufacturing data, but still trails US tech giants significantly in GPU capabilities, AI computing infrastructure and platform ecosystems. Analysts say the central challenge for the country's bid to join the "AI G3" — alongside the United States and China — is how quickly it can translate that manufacturing edge into AI infrastructure and service competitiveness.

The country holds the No. 1 position in HBM, ranks among the global leaders in AI patents, fields competitive AI models and has accumulated vast manufacturing data — yet the gaps with American big tech in GPU supply, AI computing infrastructure and platform ecosystems remain substantial.

Experts say the success or failure of the government's AI G3 strategy ultimately hinges on how swiftly South Korea can convert its world-class manufacturing competitiveness into AI infrastructure and service capabilities.

The government has already set joining the AI G3 as a core national goal. The San Francisco AI Declaration centers on building a global production base linking semiconductors, AI data centers and physical AI.

International indicators suggest South Korea's AI fundamentals are solid. According to the AI Index 2026 published by Stanford University's Human-Centered AI institute, South Korea ranked third globally in the number of "notable AI models" developed, behind only the United States and China.

Leading domestic AI models include Naver's HyperCLOVA X, LG Group AI Research's Exaone, SK Telecom's A.X K1, Kakao's Kanana and Upstage's Solar.

These companies are expanding the domestic AI ecosystem across search, telecommunications, manufacturing and the public sector, drawing on Korean-language large language models and advances in reasoning and agentic AI.

South Korea has also registered 14.31 AI patents per 100,000 people, topping the world for two consecutive years. The country has consistently placed among the leading AI nations in major assessments including the Tortoise Media Global AI Index, the IMF AI Preparedness Index and the World Intellectual Property Organization Global Innovation Index, demonstrating strong competitiveness across the board.

The greatest asset of South Korea's AI ecosystem is by far its robust hardware and manufacturing capabilities. The HBM market in particular is in effect dominated by South Korean companies.

According to market research firm Counterpoint Research, SK Hynix held a 58 percent share of the global HBM market by first-quarter sales this year, with Samsung Electronics in second place at 21 percent. Combined, the two companies control 79 percent of the market — a near-total grip on the memory chips at the heart of AI semiconductors.

HBM is an essential component in major AI accelerators, including those made by Nvidia, and is considered the foundation of South Korea's strategic advantage in the global AI supply chain.

Decades of accumulated process and sensor-control expertise from electronics, telecommunications and manufacturing operations have fed directly into AI patents, building an extensive intellectual property portfolio.

South Korea's edge could grow even sharper as the era of "physical AI" — centered on factory equipment and robots — takes hold. Physical AI improves with greater exposure to process, sensor and quality data gathered on real manufacturing floors, meaning South Korea's world-class industrial base becomes an increasingly powerful advantage.

The vast industrial data accumulated across Hyundai Motor and Kia's automotive operations, HD Hyundai and Hanwha Ocean's shipbuilding, and Samsung SDI and LG Energy Solution's battery businesses could give South Korea a competitive edge in optimizing production processes, enabling autonomous manufacturing and advancing AI models for industrial robotics.

This combination of on-site data and hardware dominance forms the core engine underpinning the "global AI production base" vision articulated in the San Francisco Declaration.

South Korea also carries significant weaknesses in AI, however. While it commands an overwhelming lead in HBM, it lags far behind US big tech in GPUs — the processing heart of AI computation — and in the massive infrastructure needed to tie them together.

With Nvidia holding a commanding grip on the GPU market, Microsoft, Google and Amazon are each pouring vast sums — ranging from tens of trillions to around 100 trillion won ($68.1 billion) annually — into building AI data centers.

By contrast, the AI investment budgets of South Korea's leading players — Naver, LG Group AI Research and Upstage among them — remain in the low trillions of won, a stark difference in scale.

Unlike the United States, which delivers AI services to a global audience through Google Search, Apple's iPhone and social media platforms such as Facebook, South Korea's AI ecosystem remains largely confined to the Korean-language domestic market — another barrier the country must overcome.

Spreading AI across the broader industrial economy is just as important as developing cutting-edge technology. Major indicators including those from the OECD show that AI adoption among South Korean companies is rising steeply, but small and medium-sized enterprises with fewer than 250 employees still lag behind leading nations such as Germany and Ireland.

Data standardization challenges and high upfront costs continue to hamper digital transformation at smaller firms. For the "AI for All" initiative embedded in the San Francisco Declaration to deliver real results, policies that lower the barriers to AI access — extending beyond large conglomerates to public administration, healthcare and small and medium-sized manufacturers — must accompany the broader push.