top of page

South Korea’s $390 Million AI Gambit: How LG’s Exaone 4.0 Is Challenging OpenAI and Google

South Korea has entered a decisive new phase in the global artificial intelligence (AI) race. In 2025, the country launched its most ambitious sovereign AI initiative to date, pledging ₩530 billion (about $390 million) to fund five domestic players building large-scale foundational models. This is not a simple tech upgrade. It is a national strategy to cut dependence on foreign AI technologies, safeguard data sovereignty, and position the country as a global hub of next-generation AI. From LG AI Research’s Exaone 4.0 to SK Telecom’s A.X, Naver Cloud’s HyperCLOVA X, and Upstage’s Solar Pro 2, the project represents a bold experiment in nurturing a full-stack AI ecosystem tuned to local language, culture, and business priorities. Each player is pursuing a different edge—efficiency, specialization, infrastructure, or ecosystem building—but together they signal a transformation in how AI might evolve beyond today’s general-purpose models. Why South Korea is Building a Sovereign AI Ecosystem The Ministry of Science and ICT’s decision to fund five companies marks a clear pivot away from dependence on global AI giants like OpenAI and Google. Every six months, the government reviews progress, cuts underperformers, and reallocates funds to frontrunners until only two remain to lead the sovereign AI drive. This “performance-based funnel” approach mirrors how private equity or venture capital accelerates innovation, but at a national scale. Three drivers stand out: Data Security and Sovereignty: AI systems trained on foreign platforms may export sensitive data. By developing its own foundational models, South Korea ensures tighter control over strategic datasets, including defense, healthcare, and industrial IP. Local Language and Culture: While GPT-4 or Gemini are optimized for global use, they often underperform in Korean-language contexts. Local players like SK Telecom and Upstage report measurable gains in Korean input efficiency and benchmark performance. Industrial Integration: Unlike consumer-focused AI in the U.S., South Korea’s models are explicitly designed to plug into high-value industries—semiconductors, mobility, finance, and advanced materials—where it already leads globally. This focus on sovereignty and specialization gives South Korea a potential advantage over larger but more generalized AI ecosystems. LG AI Research and Exaone 4.0: Hybrid Reasoning for Real-World Impact At the heart of this strategy is LG AI Research, the R&D arm of South Korea’s LG Group. Launched in December 2020 as part of LG’s digital transformation strategy, the company has developed a family of large language models under the Exaone brand. Its latest release, Exaone 4.0, is a hybrid reasoning AI model combining broad language processing with advanced reasoning capabilities introduced in its earlier Exaone Deep model. According to LG AI Research, Exaone 4.0 outperforms comparable models from Alibaba, Microsoft, and Mistral AI in science, math, and coding benchmarks, though it still trails DeepSeek’s top model. Unlike mainstream systems like ChatGPT or Gemini, Exaone 4.0 is designed primarily for business-to-business (B2B) use cases rather than consumer chatbots. Honglak Lee, co-head of LG AI Research and a former research scientist at Google Brain, describes the strategy this way: “We’re not just offering an inference engine. We aim to provide an end-to-end system that integrates the key functionalities enterprises actually need—so they can immediately plug it into their workflow. Every enterprise has unique operational needs. That’s why we’re designing our solution to be flexible—able to combine and configure different parts based on each customer’s environment.” Exaone 4.0 is already available for research and academic use on Hugging Face and supports Korean, English, and now Spanish. This multilingual reach is part of LG’s plan to move beyond local markets while staying efficient. Exaone Ecosystem: From Vision-Language Models to Enterprise Agents In July 2025, LG AI Research unveiled a broader Exaone ecosystem at its AI Talk 2025 event. The lineup includes: Exaone 4.0 Vision Language (VL): A multimodal AI model capable of interpreting both text and images, reportedly edging out Meta’s Llama 4 Scout in performance tests. Exaone Path 2.0: A healthcare-focused model designed to diagnose patient conditions in minutes. ChatExaone: An AI agent used internally by LG employees to support corporate workflows. Exaone Data Foundry: A platform that accelerates data generation, reportedly doing in one day what would take 60 experts three months. On-Premise Full-Stack Agent: Deployed in isolated, secure environments without exposing sensitive data, enabling enterprises to run autonomous agents entirely within their own infrastructure. This modular approach allows businesses to combine different components—data generation, reasoning, compliance agents—into a tailored AI system. Crucially, the on-premise version runs on chips developed by FuriosaAI, a South Korea–based startup manufacturing neural processing units (NPUs). FuriosaAI’s RNGD accelerator reportedly delivers 2.25 times the inference performance of competing GPUs and can generate up to 3.75 times as many tokens per rack at the same power limits, offering both speed and energy efficiency. Toward Physical and Web-Native AI Agents LG AI Research is also laying groundwork for “physical AI”—embedding perception, reasoning, and action loops into robots—and building Web-native agents capable of crawling, extracting, and acting on information from online sources. A prototype, the Nexus Agent, assesses legal compliance of datasets by autonomously navigating the Internet. This kind of agentive architecture moves beyond today’s static generative models toward autonomous decision-making systems. EXAONE Business Intelligence: The World’s First Fully Autonomous Financial AI Agent In September 2025, LG AI Research took its ambitions global by launching EXAONE Business Intelligence (EXAONE-BI) at the London Stock Exchange. Powered by LSEG data, EXAONE-BI is billed as the world’s first financial AI agent capable of performing end-to-end analysis, forecasting, and report generation without human intervention. The system integrates four specialized AI agents: AI Agent Function AI Journalist Collects external data such as news, corporate disclosures, and macroeconomic indicators, then synthesizes it with context. AI Economist Uses the compiled data to forecast future trends. AI Analyst Analyzes forecasts alongside internal indicators to identify key factors and anomalies, explaining them in plain language. AI Decision-Maker Compares scenarios and generates a final score. At launch, LG and LSEG showcased an AI-Powered Equity Forecast Score, a 1–100 signal predicting four-week asset returns. Scores ≤50 suggest decline; >50 suggest growth. Each score includes a commentary generated by LG’s language model, highlighting key drivers such as earnings or sentiment, making forecasts more transparent and actionable. Unlike traditional financial AI tools, which often act as auxiliary summarizers, EXAONE-BI provides expert-level insights on more than 5,000 U.S. stocks daily. It aims to usher in the era of “Agentive AI,” where multiple AI agents collaborate to deliver domain-specific intelligence autonomously. Woohyung Lim, co-head of LG AI Research, emphasized the strategic importance: “EXAONE-BI will evolve into a superintelligent agent where individual agents collaborate to create collective intelligence. This partnership is a powerful signal of Korean AI’s growing global competitiveness.” Todd Hartmann, Group Head of Data & Feeds at LSEG, added: “Our collaboration with LG AI reflects LSEG’s commitment to providing trusted data that supports more effective decision-making. Working together, we see opportunities to deliver greater value to customers and drive innovation across the financial ecosystem.” SK Telecom: Leveraging Telecom Infrastructure for AI Scale South Korea’s telco giant SK Telecom (SKT) offers a different model for scaling AI. Having launched its personal AI agent A. (A-dot) in 2023, SKT rolled out its large language model A.X in July 2025. Built on Alibaba Cloud’s open-source Qwen 2.5, A.X 4.0 comes in two versions: a 72-billion-parameter model and a lighter 7B model. Key differentiators include: Local Language Efficiency: Processes Korean inputs about 33% more efficiently than GPT-4o. Open-Source Ecosystem: Open-sourced A.X 3.1 earlier in 2025, encouraging community development. Integrated Services: The A. service features AI call summaries, auto-generated notes, and has 10 million subscribers as of August 2025. SKT’s edge lies in combining its telecom infrastructure with AI research. It is investing in GPUaaS, South Korea’s largest GPU-based service, and building a new hyperscale AI data center with AWS. Partnerships extend to Korean AI chipmaker Rebellions, government agencies, universities, and international research groups like MIT’s MGAIC, which applies foundation models to advanced manufacturing, battery, and semiconductor innovation. As Taeyoon Kim, head of the foundation model office at SK Telecom, explains: “With our telecom infrastructure, extensive user base and proven service like A., we bring AI directly into everyday life, whether in customer service, mobility, or manufacturing.” Naver Cloud: Building an AI Full Stack Naver Cloud, the cloud services arm of South Korea’s leading internet company, launched its large language model HyperCLOVA in 2021 and upgraded to HyperCLOVA X two years later. Alongside its chatbot CLOVA X and generative AI search engine Cue, Naver recently introduced a multimodal reasoning model, HyperCLOVE X Think. Naver stands out as Korea’s only company—and one of the few globally—with a genuine “AI full stack”: it builds the model, runs the data centers, provides cloud services, and integrates AI into consumer products such as search, shopping, maps, and finance. Its AI Shopping Guide, for example, offers personalized recommendations based on real-world purchasing behavior. Business-facing offerings like CLOVA Studio and CLOVA Carecall extend this capability to third parties. A Naver spokesperson summarized the company’s philosophy: “The true power of LLMs is to serve as connectors linking legacy systems and siloed services to improve usefulness.” Rather than chasing parameter counts, Naver emphasizes perfecting its model “recipe” and securing the capital to scale, arguing that sophistication at comparable sizes can outperform brute-force approaches. Upstage: A Startup in the Frontier Model Club Upstage is the only startup among the five funded players but has already made its mark with Solar Pro 2, launched in July 2025. Recognized as a frontier model by Artificial Analysis, Solar Pro 2 has just 31 billion parameters but outperforms larger global models on major Korean benchmarks. Upstage targets a Korean language performance of 105% of the global standard and is building specialized models for industries such as finance, law, and medicine. Soon-il Kwon, executive vice president at Upstage, explains the company’s differentiator: “Solar Pro 2 has outperformed global models on major Korean benchmarks. With this project, Upstage aims to achieve a Korean language performance of 105% of the global standard.” By focusing on real business impact rather than just benchmarks, Upstage is fostering an ecosystem of “AI-native” startups that could rival the big tech incumbents. How South Korea’s Strategy Differs from the U.S. and China The U.S. and China dominate AI by sheer scale of investment, data, and compute power. South Korea’s strategy is different: Selective Scaling: Funding only the top-performing models every six months creates a survival-of-the-fittest dynamic. Industry-Specific Models: Prioritizing integration with sectors like semiconductors, biotech, and finance. Hardware Efficiency: Leveraging domestic chipmakers like FuriosaAI and Rebellions to optimize energy and token generation. Agentive Architecture: Moving beyond static LLMs toward autonomous agents capable of forecasting, compliance checking, and decision-making. This approach positions South Korea as a “smart scaler” in AI—a country that may not outspend OpenAI or Google but could outmaneuver them in efficiency, localization, and industrial integration. Conclusion: A New Model for AI Sovereignty and Innovation South Korea’s sovereign AI initiative marks a turning point in how countries think about artificial intelligence. By funding a competitive cohort of companies, the government is catalyzing innovation in a way that aligns with national priorities—data security, industrial competitiveness, and cultural relevance. Models like LG’s Exaone 4.0, EXAONE-BI, and Upstage’s Solar Pro 2 show that cutting-edge AI does not have to be massive to be effective. Meanwhile, SK Telecom’s telecom-based AI services and Naver’s full-stack ecosystem illustrate the power of embedding AI into existing infrastructures. As countries and corporations alike grapple with the risks and opportunities of AI, South Korea’s experiment offers a blueprint for building sovereign, specialized, and efficient AI systems. It also signals that the next wave of competition may not be about who has the largest model, but who can build the smartest, most autonomous, and most integrated one. For readers seeking deeper analysis on the intersection of AI, national strategy, and emerging technologies, the expert team at 1950.ai, led by Dr. Shahid Masood, regularly provides insights on how innovations like LG’s Exaone ecosystem and agentive AI are shaping global industries. Their work highlights the critical role of data sovereignty, efficient infrastructure, and cross-sector integration in the future of AI. Further Reading / External References How South Korea Plans to Best OpenAI, Google, Others with Homegrown AI – TechCrunch LG Launches EXAONE Business Intelligence to Boost Global Financial Markets – Intlbm Exaone 4.0 and LG AI Research’s Strategic Roadmap – IEEE Spectrum

South Korea has entered a decisive new phase in the global artificial intelligence (AI) race. In 2025, the country launched its most ambitious sovereign AI initiative to date, pledging ₩530 billion (about $390 million) to fund five domestic players building large-scale foundational models. This is not a simple tech upgrade. It is a national strategy to cut dependence on foreign AI technologies, safeguard data sovereignty, and position the country as a global hub of next-generation AI.


From LG AI Research’s Exaone 4.0 to SK Telecom’s A.X, Naver Cloud’s HyperCLOVA X, and Upstage’s Solar Pro 2, the project represents a bold experiment in nurturing a full-stack AI ecosystem tuned to local language, culture, and business priorities. Each player is pursuing a different edge—efficiency, specialization, infrastructure, or ecosystem building—but together they signal a transformation in how AI might evolve beyond today’s general-purpose models.


Why South Korea is Building a Sovereign AI Ecosystem

The Ministry of Science and ICT’s decision to fund five companies marks a clear pivot away from dependence on global AI giants like OpenAI and Google. Every six months, the government reviews progress, cuts underperformers, and reallocates funds to frontrunners until only two remain to lead the sovereign AI drive. This “performance-based funnel” approach mirrors how private equity or venture capital accelerates innovation, but at a national scale.


Three drivers stand out:

  • Data Security and Sovereignty: AI systems trained on foreign platforms may export sensitive data. By developing its own foundational models, South Korea ensures tighter control over strategic datasets, including defense, healthcare, and industrial IP.

  • Local Language and Culture: While GPT-4 or Gemini are optimized for global use, they often underperform in Korean-language contexts. Local players like SK Telecom and Upstage report measurable gains in Korean input efficiency and benchmark performance.

  • Industrial Integration: Unlike consumer-focused AI in the U.S., South Korea’s models are explicitly designed to plug into high-value industries—semiconductors, mobility, finance, and advanced materials—where it already leads globally.


This focus on sovereignty and specialization gives South Korea a potential advantage over larger but more generalized AI ecosystems.


LG AI Research and Exaone 4.0: Hybrid Reasoning for Real-World Impact

At the heart of this strategy is LG AI Research, the R&D arm of South Korea’s LG Group. Launched in December 2020 as part of LG’s digital transformation strategy, the company has developed a family of large language models under the Exaone brand. Its latest release, Exaone 4.0, is a hybrid reasoning AI model combining broad language processing with advanced reasoning capabilities introduced in its earlier Exaone Deep model.


According to LG AI Research, Exaone 4.0 outperforms comparable models from Alibaba, Microsoft, and Mistral AI in science, math, and coding benchmarks, though it still trails DeepSeek’s top model. Unlike mainstream systems like ChatGPT or Gemini, Exaone 4.0 is designed primarily for business-to-business (B2B) use cases rather than consumer chatbots.

Honglak Lee, co-head of LG AI Research and a former research scientist at Google Brain, describes the strategy this way:

“We’re not just offering an inference engine. We aim to provide an end-to-end system that integrates the key functionalities enterprises actually need—so they can immediately plug it into their workflow. Every enterprise has unique operational needs. That’s why we’re designing our solution to be flexible—able to combine and configure different parts based on each customer’s environment.”

Exaone 4.0 is already available for research and academic use on Hugging Face and supports Korean, English, and now Spanish. This multilingual reach is part of LG’s plan to move beyond local markets while staying efficient.


Exaone Ecosystem: From Vision-Language Models to Enterprise Agents

In July 2025, LG AI Research unveiled a broader Exaone ecosystem at its AI Talk 2025 event. The lineup includes:

  • Exaone 4.0 Vision Language (VL): A multimodal AI model capable of interpreting both text and images, reportedly edging out Meta’s Llama 4 Scout in performance tests.

  • Exaone Path 2.0: A healthcare-focused model designed to diagnose patient conditions in minutes.

  • ChatExaone: An AI agent used internally by LG employees to support corporate workflows.

  • Exaone Data Foundry: A platform that accelerates data generation, reportedly doing in one day what would take 60 experts three months.

  • On-Premise Full-Stack Agent: Deployed in isolated, secure environments without exposing sensitive data, enabling enterprises to run autonomous agents entirely within their own infrastructure.


This modular approach allows businesses to combine different components—data generation, reasoning, compliance agents—into a tailored AI system. Crucially, the on-premise version runs on chips developed by FuriosaAI, a South Korea–based startup manufacturing neural processing units (NPUs). FuriosaAI’s RNGD accelerator reportedly delivers 2.25 times the inference performance of competing GPUs and can generate up to 3.75 times as many tokens per rack at the same power limits, offering both speed and energy efficiency.


Toward Physical and Web-Native AI Agents

LG AI Research is also laying groundwork for “physical AI”—embedding perception, reasoning, and action loops into robots—and building Web-native agents capable of crawling, extracting, and acting on information from online sources. A prototype, the Nexus Agent, assesses legal compliance of datasets by autonomously navigating the Internet. This kind of agentive architecture moves beyond today’s static generative models toward autonomous decision-making systems.


EXAONE Business Intelligence: The World’s First Fully Autonomous Financial AI Agent

In September 2025, LG AI Research took its ambitions global by launching EXAONE Business Intelligence (EXAONE-BI) at the London Stock Exchange. Powered by LSEG data, EXAONE-BI is billed as the world’s first financial AI agent capable of performing end-to-end analysis, forecasting, and report generation without human intervention.

The system integrates four specialized AI agents:

AI Agent

Function

AI Journalist

Collects external data such as news, corporate disclosures, and macroeconomic indicators, then synthesizes it with context.

AI Economist

Uses the compiled data to forecast future trends.

AI Analyst

Analyzes forecasts alongside internal indicators to identify key factors and anomalies, explaining them in plain language.

AI Decision-Maker

Compares scenarios and generates a final score.

At launch, LG and LSEG showcased an AI-Powered Equity Forecast Score, a 1–100 signal predicting four-week asset returns. Scores ≤50 suggest decline; >50 suggest growth. Each score includes a commentary generated by LG’s language model, highlighting key drivers such as earnings or sentiment, making forecasts more transparent and actionable.


Unlike traditional financial AI tools, which often act as auxiliary summarizers, EXAONE-BI provides expert-level insights on more than 5,000 U.S. stocks daily. It aims to usher in the era of “Agentive AI,” where multiple AI agents collaborate to deliver domain-specific intelligence autonomously.


Woohyung Lim, co-head of LG AI Research, emphasized the strategic importance:

“EXAONE-BI will evolve into a superintelligent agent where individual agents collaborate to create collective intelligence. This partnership is a powerful signal of Korean AI’s growing global competitiveness.”

Todd Hartmann, Group Head of Data & Feeds at LSEG, added:

“Our collaboration with LG AI reflects LSEG’s commitment to providing trusted data that supports more effective decision-making. Working together, we see opportunities to deliver greater value to customers and drive innovation across the financial ecosystem.”

SK Telecom: Leveraging Telecom Infrastructure for AI Scale

South Korea’s telco giant SK Telecom (SKT) offers a different model for scaling AI. Having launched its personal AI agent A. (A-dot) in 2023, SKT rolled out its large language model A.X in July 2025. Built on Alibaba Cloud’s open-source Qwen 2.5, A.X 4.0 comes in two versions: a 72-billion-parameter model and a lighter 7B model.

South Korea has entered a decisive new phase in the global artificial intelligence (AI) race. In 2025, the country launched its most ambitious sovereign AI initiative to date, pledging ₩530 billion (about $390 million) to fund five domestic players building large-scale foundational models. This is not a simple tech upgrade. It is a national strategy to cut dependence on foreign AI technologies, safeguard data sovereignty, and position the country as a global hub of next-generation AI. From LG AI Research’s Exaone 4.0 to SK Telecom’s A.X, Naver Cloud’s HyperCLOVA X, and Upstage’s Solar Pro 2, the project represents a bold experiment in nurturing a full-stack AI ecosystem tuned to local language, culture, and business priorities. Each player is pursuing a different edge—efficiency, specialization, infrastructure, or ecosystem building—but together they signal a transformation in how AI might evolve beyond today’s general-purpose models. Why South Korea is Building a Sovereign AI Ecosystem The Ministry of Science and ICT’s decision to fund five companies marks a clear pivot away from dependence on global AI giants like OpenAI and Google. Every six months, the government reviews progress, cuts underperformers, and reallocates funds to frontrunners until only two remain to lead the sovereign AI drive. This “performance-based funnel” approach mirrors how private equity or venture capital accelerates innovation, but at a national scale. Three drivers stand out: Data Security and Sovereignty: AI systems trained on foreign platforms may export sensitive data. By developing its own foundational models, South Korea ensures tighter control over strategic datasets, including defense, healthcare, and industrial IP. Local Language and Culture: While GPT-4 or Gemini are optimized for global use, they often underperform in Korean-language contexts. Local players like SK Telecom and Upstage report measurable gains in Korean input efficiency and benchmark performance. Industrial Integration: Unlike consumer-focused AI in the U.S., South Korea’s models are explicitly designed to plug into high-value industries—semiconductors, mobility, finance, and advanced materials—where it already leads globally. This focus on sovereignty and specialization gives South Korea a potential advantage over larger but more generalized AI ecosystems. LG AI Research and Exaone 4.0: Hybrid Reasoning for Real-World Impact At the heart of this strategy is LG AI Research, the R&D arm of South Korea’s LG Group. Launched in December 2020 as part of LG’s digital transformation strategy, the company has developed a family of large language models under the Exaone brand. Its latest release, Exaone 4.0, is a hybrid reasoning AI model combining broad language processing with advanced reasoning capabilities introduced in its earlier Exaone Deep model. According to LG AI Research, Exaone 4.0 outperforms comparable models from Alibaba, Microsoft, and Mistral AI in science, math, and coding benchmarks, though it still trails DeepSeek’s top model. Unlike mainstream systems like ChatGPT or Gemini, Exaone 4.0 is designed primarily for business-to-business (B2B) use cases rather than consumer chatbots. Honglak Lee, co-head of LG AI Research and a former research scientist at Google Brain, describes the strategy this way: “We’re not just offering an inference engine. We aim to provide an end-to-end system that integrates the key functionalities enterprises actually need—so they can immediately plug it into their workflow. Every enterprise has unique operational needs. That’s why we’re designing our solution to be flexible—able to combine and configure different parts based on each customer’s environment.” Exaone 4.0 is already available for research and academic use on Hugging Face and supports Korean, English, and now Spanish. This multilingual reach is part of LG’s plan to move beyond local markets while staying efficient. Exaone Ecosystem: From Vision-Language Models to Enterprise Agents In July 2025, LG AI Research unveiled a broader Exaone ecosystem at its AI Talk 2025 event. The lineup includes: Exaone 4.0 Vision Language (VL): A multimodal AI model capable of interpreting both text and images, reportedly edging out Meta’s Llama 4 Scout in performance tests. Exaone Path 2.0: A healthcare-focused model designed to diagnose patient conditions in minutes. ChatExaone: An AI agent used internally by LG employees to support corporate workflows. Exaone Data Foundry: A platform that accelerates data generation, reportedly doing in one day what would take 60 experts three months. On-Premise Full-Stack Agent: Deployed in isolated, secure environments without exposing sensitive data, enabling enterprises to run autonomous agents entirely within their own infrastructure. This modular approach allows businesses to combine different components—data generation, reasoning, compliance agents—into a tailored AI system. Crucially, the on-premise version runs on chips developed by FuriosaAI, a South Korea–based startup manufacturing neural processing units (NPUs). FuriosaAI’s RNGD accelerator reportedly delivers 2.25 times the inference performance of competing GPUs and can generate up to 3.75 times as many tokens per rack at the same power limits, offering both speed and energy efficiency. Toward Physical and Web-Native AI Agents LG AI Research is also laying groundwork for “physical AI”—embedding perception, reasoning, and action loops into robots—and building Web-native agents capable of crawling, extracting, and acting on information from online sources. A prototype, the Nexus Agent, assesses legal compliance of datasets by autonomously navigating the Internet. This kind of agentive architecture moves beyond today’s static generative models toward autonomous decision-making systems. EXAONE Business Intelligence: The World’s First Fully Autonomous Financial AI Agent In September 2025, LG AI Research took its ambitions global by launching EXAONE Business Intelligence (EXAONE-BI) at the London Stock Exchange. Powered by LSEG data, EXAONE-BI is billed as the world’s first financial AI agent capable of performing end-to-end analysis, forecasting, and report generation without human intervention. The system integrates four specialized AI agents: AI Agent Function AI Journalist Collects external data such as news, corporate disclosures, and macroeconomic indicators, then synthesizes it with context. AI Economist Uses the compiled data to forecast future trends. AI Analyst Analyzes forecasts alongside internal indicators to identify key factors and anomalies, explaining them in plain language. AI Decision-Maker Compares scenarios and generates a final score. At launch, LG and LSEG showcased an AI-Powered Equity Forecast Score, a 1–100 signal predicting four-week asset returns. Scores ≤50 suggest decline; >50 suggest growth. Each score includes a commentary generated by LG’s language model, highlighting key drivers such as earnings or sentiment, making forecasts more transparent and actionable. Unlike traditional financial AI tools, which often act as auxiliary summarizers, EXAONE-BI provides expert-level insights on more than 5,000 U.S. stocks daily. It aims to usher in the era of “Agentive AI,” where multiple AI agents collaborate to deliver domain-specific intelligence autonomously. Woohyung Lim, co-head of LG AI Research, emphasized the strategic importance: “EXAONE-BI will evolve into a superintelligent agent where individual agents collaborate to create collective intelligence. This partnership is a powerful signal of Korean AI’s growing global competitiveness.” Todd Hartmann, Group Head of Data & Feeds at LSEG, added: “Our collaboration with LG AI reflects LSEG’s commitment to providing trusted data that supports more effective decision-making. Working together, we see opportunities to deliver greater value to customers and drive innovation across the financial ecosystem.” SK Telecom: Leveraging Telecom Infrastructure for AI Scale South Korea’s telco giant SK Telecom (SKT) offers a different model for scaling AI. Having launched its personal AI agent A. (A-dot) in 2023, SKT rolled out its large language model A.X in July 2025. Built on Alibaba Cloud’s open-source Qwen 2.5, A.X 4.0 comes in two versions: a 72-billion-parameter model and a lighter 7B model. Key differentiators include: Local Language Efficiency: Processes Korean inputs about 33% more efficiently than GPT-4o. Open-Source Ecosystem: Open-sourced A.X 3.1 earlier in 2025, encouraging community development. Integrated Services: The A. service features AI call summaries, auto-generated notes, and has 10 million subscribers as of August 2025. SKT’s edge lies in combining its telecom infrastructure with AI research. It is investing in GPUaaS, South Korea’s largest GPU-based service, and building a new hyperscale AI data center with AWS. Partnerships extend to Korean AI chipmaker Rebellions, government agencies, universities, and international research groups like MIT’s MGAIC, which applies foundation models to advanced manufacturing, battery, and semiconductor innovation. As Taeyoon Kim, head of the foundation model office at SK Telecom, explains: “With our telecom infrastructure, extensive user base and proven service like A., we bring AI directly into everyday life, whether in customer service, mobility, or manufacturing.” Naver Cloud: Building an AI Full Stack Naver Cloud, the cloud services arm of South Korea’s leading internet company, launched its large language model HyperCLOVA in 2021 and upgraded to HyperCLOVA X two years later. Alongside its chatbot CLOVA X and generative AI search engine Cue, Naver recently introduced a multimodal reasoning model, HyperCLOVE X Think. Naver stands out as Korea’s only company—and one of the few globally—with a genuine “AI full stack”: it builds the model, runs the data centers, provides cloud services, and integrates AI into consumer products such as search, shopping, maps, and finance. Its AI Shopping Guide, for example, offers personalized recommendations based on real-world purchasing behavior. Business-facing offerings like CLOVA Studio and CLOVA Carecall extend this capability to third parties. A Naver spokesperson summarized the company’s philosophy: “The true power of LLMs is to serve as connectors linking legacy systems and siloed services to improve usefulness.” Rather than chasing parameter counts, Naver emphasizes perfecting its model “recipe” and securing the capital to scale, arguing that sophistication at comparable sizes can outperform brute-force approaches. Upstage: A Startup in the Frontier Model Club Upstage is the only startup among the five funded players but has already made its mark with Solar Pro 2, launched in July 2025. Recognized as a frontier model by Artificial Analysis, Solar Pro 2 has just 31 billion parameters but outperforms larger global models on major Korean benchmarks. Upstage targets a Korean language performance of 105% of the global standard and is building specialized models for industries such as finance, law, and medicine. Soon-il Kwon, executive vice president at Upstage, explains the company’s differentiator: “Solar Pro 2 has outperformed global models on major Korean benchmarks. With this project, Upstage aims to achieve a Korean language performance of 105% of the global standard.” By focusing on real business impact rather than just benchmarks, Upstage is fostering an ecosystem of “AI-native” startups that could rival the big tech incumbents. How South Korea’s Strategy Differs from the U.S. and China The U.S. and China dominate AI by sheer scale of investment, data, and compute power. South Korea’s strategy is different: Selective Scaling: Funding only the top-performing models every six months creates a survival-of-the-fittest dynamic. Industry-Specific Models: Prioritizing integration with sectors like semiconductors, biotech, and finance. Hardware Efficiency: Leveraging domestic chipmakers like FuriosaAI and Rebellions to optimize energy and token generation. Agentive Architecture: Moving beyond static LLMs toward autonomous agents capable of forecasting, compliance checking, and decision-making. This approach positions South Korea as a “smart scaler” in AI—a country that may not outspend OpenAI or Google but could outmaneuver them in efficiency, localization, and industrial integration. Conclusion: A New Model for AI Sovereignty and Innovation South Korea’s sovereign AI initiative marks a turning point in how countries think about artificial intelligence. By funding a competitive cohort of companies, the government is catalyzing innovation in a way that aligns with national priorities—data security, industrial competitiveness, and cultural relevance. Models like LG’s Exaone 4.0, EXAONE-BI, and Upstage’s Solar Pro 2 show that cutting-edge AI does not have to be massive to be effective. Meanwhile, SK Telecom’s telecom-based AI services and Naver’s full-stack ecosystem illustrate the power of embedding AI into existing infrastructures. As countries and corporations alike grapple with the risks and opportunities of AI, South Korea’s experiment offers a blueprint for building sovereign, specialized, and efficient AI systems. It also signals that the next wave of competition may not be about who has the largest model, but who can build the smartest, most autonomous, and most integrated one. For readers seeking deeper analysis on the intersection of AI, national strategy, and emerging technologies, the expert team at 1950.ai, led by Dr. Shahid Masood, regularly provides insights on how innovations like LG’s Exaone ecosystem and agentive AI are shaping global industries. Their work highlights the critical role of data sovereignty, efficient infrastructure, and cross-sector integration in the future of AI. Further Reading / External References How South Korea Plans to Best OpenAI, Google, Others with Homegrown AI – TechCrunch LG Launches EXAONE Business Intelligence to Boost Global Financial Markets – Intlbm Exaone 4.0 and LG AI Research’s Strategic Roadmap – IEEE Spectrum

Key differentiators include:

  • Local Language Efficiency: Processes Korean inputs about 33% more efficiently than GPT-4o.

  • Open-Source Ecosystem: Open-sourced A.X 3.1 earlier in 2025, encouraging community development.

  • Integrated Services: The A. service features AI call summaries, auto-generated notes, and has 10 million subscribers as of August 2025.


SKT’s edge lies in combining its telecom infrastructure with AI research. It is investing in GPUaaS, South Korea’s largest GPU-based service, and building a new hyperscale AI data center with AWS. Partnerships extend to Korean AI chipmaker Rebellions, government agencies, universities, and international research groups like MIT’s MGAIC, which applies foundation models to advanced manufacturing, battery, and semiconductor innovation.


As Taeyoon Kim, head of the foundation model office at SK Telecom, explains:

“With our telecom infrastructure, extensive user base and proven service like A., we bring AI directly into everyday life, whether in customer service, mobility, or manufacturing.”

Naver Cloud: Building an AI Full Stack

Naver Cloud, the cloud services arm of South Korea’s leading internet company, launched its large language model HyperCLOVA in 2021 and upgraded to HyperCLOVA X two years later. Alongside its chatbot CLOVA X and generative AI search engine Cue, Naver recently introduced a multimodal reasoning model, HyperCLOVE X Think.


Naver stands out as Korea’s only company—and one of the few globally—with a genuine “AI full stack”: it builds the model, runs the data centers, provides cloud services, and integrates AI into consumer products such as search, shopping, maps, and finance. Its AI Shopping Guide, for example, offers personalized recommendations based on real-world purchasing behavior. Business-facing offerings like CLOVA Studio and CLOVA Carecall extend this capability to third parties.


A Naver spokesperson summarized the company’s philosophy:

“The true power of LLMs is to serve as connectors linking legacy systems and siloed services to improve usefulness.”

Rather than chasing parameter counts, Naver emphasizes perfecting its model “recipe” and securing the capital to scale, arguing that sophistication at comparable sizes can outperform brute-force approaches.


Upstage: A Startup in the Frontier Model Club

Upstage is the only startup among the five funded players but has already made its mark with Solar Pro 2, launched in July 2025. Recognized as a frontier model by Artificial Analysis, Solar Pro 2 has just 31 billion parameters but outperforms larger global models on major Korean benchmarks. Upstage targets a Korean language performance of 105% of the global standard and is building specialized models for industries such as finance, law, and medicine.

Soon-il Kwon, executive vice president at Upstage, explains the company’s differentiator:

“Solar Pro 2 has outperformed global models on major Korean benchmarks. With this project, Upstage aims to achieve a Korean language performance of 105% of the global standard.”

By focusing on real business impact rather than just benchmarks, Upstage is fostering an ecosystem of “AI-native” startups that could rival the big tech incumbents.


How South Korea’s Strategy Differs from the U.S. and China

The U.S. and China dominate AI by sheer scale of investment, data, and compute power. South Korea’s strategy is different:

  • Selective Scaling: Funding only the top-performing models every six months creates a survival-of-the-fittest dynamic.

  • Industry-Specific Models: Prioritizing integration with sectors like semiconductors, biotech, and finance.

  • Hardware Efficiency: Leveraging domestic chipmakers like FuriosaAI and Rebellions to optimize energy and token generation.

  • Agentive Architecture: Moving beyond static LLMs toward autonomous agents capable of forecasting, compliance checking, and decision-making.


This approach positions South Korea as a “smart scaler” in AI—a country that may not outspend OpenAI or Google but could outmaneuver them in efficiency, localization, and industrial integration.


A New Model for AI Sovereignty and Innovation

South Korea’s sovereign AI initiative marks a turning point in how countries think about artificial intelligence. By funding a competitive cohort of companies, the government is catalyzing innovation in a way that aligns with national priorities—data security, industrial competitiveness, and cultural relevance.


Models like LG’s Exaone 4.0, EXAONE-BI, and Upstage’s Solar Pro 2 show that cutting-edge AI does not have to be massive to be effective. Meanwhile, SK Telecom’s telecom-based AI services and Naver’s full-stack ecosystem illustrate the power of embedding AI into existing infrastructures.


As countries and corporations alike grapple with the risks and opportunities of AI, South Korea’s experiment offers a blueprint for building sovereign, specialized, and efficient AI systems. It also signals that the next wave of competition may not be about who has the largest model, but who can build the smartest, most autonomous, and most integrated one.


For readers seeking deeper analysis on the intersection of AI, national strategy, and emerging technologies, the expert team at 1950.ai, led by Dr. Shahid Masood, regularly provides insights on how innovations like LG’s Exaone ecosystem and agentive AI are shaping global industries. Their work highlights the critical role of data sovereignty, efficient infrastructure, and cross-sector integration in the future of AI.


Further Reading / External References

Comments

bottom of page