Korean AI Services: HyperCLOVA X, Kakao, SKT, Exaone — 2026-10-22
South Korea’s sovereign AI push faces a critical juncture as the government confirms a two-track strategy for frontier model development and industrial diffusion, backed by a 4.7 trillion won budget. Meanwhile, domestic app usage data reveals ChatGPT nearing 25 million monthly users in Korea, while local challenger Zeta claims the top spot for total usage time. NVIDIA continues to deepen its industrial ties with Korean tech giants, expanding component partnerships and infrastructure investments.
Korean AI Services: HyperCLOVA X, Kakao, SKT, Exaone — 2026-10-22
Top developments
Government confirms two-track AI strategy with 4.7 trillion won budget
The Ministry of Science and ICT (MSIT) announced on October 5 that South Korea will invest 4.7 trillion won (approx. $3.5 billion) to develop state-led frontier AI models, splitting the push into "global top-tier model development" and "domestic AI industry diffusion" tracks. This move addresses confusion over the Sovereign AI Foundation Model Project, with the ministry clarifying that no separate GPU budget was allocated for the sovereign model project this year, instead concentrating support on 10,000 Nvidia Vera Rubin GPUs and data resources. The government will select two elite teams for the "Independent AI Foundation Model" (DokpaMo) project by February 2027, aiming to foster homegrown capabilities while avoiding the pitfalls of previous fragmented efforts.
ChatGPT nears 25 million users in Korea; Zeta leads in usage time
According to a September 2026 ranking by WiseApp·Retail released on October 13, OpenAI’s ChatGPT maintained its No. 1 position in South Korea with 24.7 million monthly active users (MAU). Despite ChatGPT’s dominance in user count, domestic chatbot Zeta recorded the highest total usage time, exceeding ChatGPT’s usage by more than double, signaling strong engagement among local users for Korean-language optimized services. This data underscores the competitive landscape for Daily desk services like HyperCLOVA X and Kakao Kanana, where user retention and depth of engagement are becoming as critical as raw user numbers.
NVIDIA expands partnerships with Korean component makers
NVIDIA is deepening its industrial ties with South Korea by tapping local component makers for AI chips and robots, as reported on October 22. Samsung Electro-Mechanics and LG Innotek are expanding their collaborations, leveraging expertise in semiconductor substrates and sensing technology to support NVIDIA’s hardware ecosystem. This move complements NVIDIA’s earlier $1 billion investment in Naver’s AI data center and its expanded accord with SK Group, solidifying a "Korea AI Stack" that integrates local hardware supply chains with global AI infrastructure demands.
NIPA secures 19,000 GPUs for domestic AI ecosystem
Park Yoon-kyu, head of the National IT Industry Promotion Agency (NIPA), announced that South Korea has secured 19,000 graphics processing units (GPUs) this year to accelerate AI infrastructure expansion. These GPUs will be supplied quickly to startups, small firms, and research institutions to foster a homegrown AI ecosystem, addressing the critical resource bottleneck for smaller players in the Korean AI market. This initiative aligns with the broader national goal to decentralize AI compute access beyond the major conglomerates like Naver and SKT.
Local view
Local media outlets are highlighting the strategic pivot in Korea’s AI policy. The Korea JoongAng Daily notes that the government’s decision to split the AI push into two tracks is an attempt to clarify the "Sovereign AI Foundation Model Project," which had faced confusion over budget allocations. Seoul Economic Daily emphasizes that while the 4.7 trillion won investment is significant, the focus on "frontier AI" and "industrial diffusion" reflects a pragmatic approach to catching up with US and Chinese models while ensuring domestic industries benefit.
Context & numbers
- Sovereign AI Budget: 4.7 trillion won ($3.5 billion) allocated for frontier AI model development.
- GPU Allocation: 10,000 Nvidia Vera Rubin GPUs concentrated for frontier model support; 19,000 GPUs secured by NIPA for broader ecosystem use.
- User Metrics: ChatGPT MAU in Korea: 24.7 million; Zeta MAU: ~4 million (but leads in total usage time).
- Infrastructure: South Korea plans 18.4 GW of AI data centres by 2035, with grid capacity identified as the primary constraint.
On the radar
- Final Selection of DokpaMo Teams: The government is scheduled to select the final two teams for the Independent AI Foundation Model project in February 2027.
- AI for All Pilot Services: SK Telecom, Kakao, and KT are set to begin pilot services for the "AI for All" free public AI initiative this month, with a contest scheduled for October to determine the best operator.
- G-Star 2026: KRAFTON will showcase its lineup including "Tarae: The Unbound" and "Age Twisters" at G-Star 2026 in Busan from November 19-22, highlighting the growing intersection of AI and gaming tech.
This content was collected, curated, and summarized entirely by AI — including how and what to gather. It may contain inaccuracies. Crew does not guarantee the accuracy of any information presented here. Always verify facts on your own before acting on them. Crew assumes no legal liability for any consequences arising from reliance on this content.
Powered by
