South Korea’s AI Renaissance & Exclusive Interview with 'Korea's Sequoia'
Deep Dive into Korea's AI Ecosystem: Innovation, Investment, and the Building of Sovereign AI
Hi all,
I was in Seoul, South Korea, for a few days last week and had the opportunity to meet with both old and new friends. My J-school classmate, who has been a badass foreign correspondent, covering everything from the Japanese stock market to the Thai political unrest, has recently moved to the Asia bureau of the New York Times in Seoul, so it was such a treat to meet up with her and learn about her journey. I also met up with a few investor contacts in Korea, in the early VC space as well as later public sector, to get a better sense of the overall ecosystem.
Overall, it was a super fruitful trip, filled with kimchi and beef, and a wealth of knowledge from people on the ground about the shifts and changes in the economy, administration, and obviously, the AI & tech space.
Thus, I think today, we can start with the macro context to really understand why Korea is where it is with its AI strategy.
Macro Background
For those of you who might not have been following, Korea experienced a tumultuous few years with its changing administrations. Two presidents were charged with corruption and misuse of power and were then impeached. It was a scandalous decade, and thus, domestically, people were a bit battered and disappointed. Internationally, K-pop continued to make waves with Blackpink performing at Coachella, and Korean films and shows continuing to push boundaries.
However, its economy struggled, and diplomatic and trade relations between South Korea, Japan, and China were also somewhat rocky, given the shifts in approach and philosophy by the three major economies in East Asia. Now, with a new president who just resumed office in June this year, we’re seeing some major efforts being made to stabilize its economy and push forward its technological capabilities. This is all critical information to better understand the investment space and the technology/ AI ecosystem in South Korea.
Governance
South Korea's political landscape over the past decade has been marked by significant presidential scandals, impeachments, and shifting economic and foreign policy priorities. The country witnessed two presidential impeachments — Park Geun-hye in 2017 for corruption and embezzlement, and Yoon Suk Yeol in 2025 for misuse of martial law and illegal arrest of judges. These incidents have reshaped governance. Economically, South Korea faces challenges including slowing growth and high household debt, while foreign policy navigates complex U.S.-China competition. The 2025 snap election brought progressive leader Lee Jae-myung to power, signaling a pragmatic recalibration in both domestic and international arenas.
The new president and government were only officially instated in June this year and thus have said that, as the administration is putting together their agenda and priorities, they had to skip the recent NATO meeting in The Hague, Netherlands, where South Korea has traditionally been a major participant.
South Korea-China-U.S. Trilateral Relations
Historically, South Korea adopted an approach called “anmi kyeongjung,” which translates to "security with the U.S., economy with China." For context, the U.S. has had a security alliance with South Korea since the 1950s, whereas China is South Korea's largest trading partner, and the two nations have long had a bilateral trade agreement.
Citing the Asia Pacific Foundation of Canada, the new South Korean President Jae-myung is “ Lee is expected to depart from Yoon’s U.S.-centric, Japan-friendly regional security initiatives based on shared values and instead resume cautious engagement with China and Russia to maximize South Korea’s manoeuvrability in an increasingly polarized regional order.”
While I walked from the Imperial Palace to Myeongdong, a shopping district just a ~7km walk away, along the streets, I encountered small-scale protests against U.S. interference in Korean politics and businesses, as well as gatherings supporting a stronger U.S. alliance. It seems that the country is undergoing an internal reassessment of pragmatism versus idealism and reevaluating its diplomatic strategy.
Software
But this conundrum doesn’t seem to affect the startups, at least not for the AI developers. From people on the ground, some South Korean AI startups and tech companies are taking an apolitical approach to LLMs, and they’re open to using everything from DeepSeek to OpenAI. While others are choosing not to use either Chinese or U.S. LLMs due to compliance and internal security considerations altogether.
For most businesses, they are leaning into what think tank Carnegie Endowment for International Peace coined as “shifting from engagement-centered idealism to strategic pragmatism” rather than picking one over the other, given that there is currently no strong domestic contender to replace international LLM options.
In fact, for startups, it’s less of a political choice but a strategic business case. Robbie Oh, tech investor at leading VC Altos Ventures, said to AI Proem during a meeting.
“Korean founders are clever and highly strategic. While they may publicly highlight their use of US-based LLMs, their actual decisions tend to be much more practical. Instead of focusing on brand or origin, they compare performance, price, customization flexibility, API accessibility, and service reliability.
In this context, Chinese LLMs are emerging as competitive alternatives. Models like DeepSeek, Zhipu, and Moonshot offer high token throughput, low cost, and relatively open fine-tuning capabilities. These features are attractive not only for experimentation but also for real-world service deployment.
U.S.-based models such as GPT-4, Claude, and Mistral (French) still lead in technical credibility and have stronger communities for support. However, many startups now use hybrid strategies. For example, some run basic operations through a Chinese model and send premium tasks to GPT-based systems.
Startups we meet at Altos rarely rely on just one model. They are building multi-model workflows, optimizing for cost and performance, and designing user experiences tailored to specific markets. In Korea, where the speed of iteration is high and technology adoption is fast, this approach works especially well.”
Hardware
Meanwhile, the U.S. government is reportedly considering revoking export control waivers for major chipmakers operating in China, which includes Korean semiconductor companies such as SK Hynix and Samsung. This move means that the U.S. will require license approvals for future equipment imports and technology updates, which will significantly impact Korean chipmakers' operations, as their Chinese fabs contribute a sizable share of their DRAM and NAND production.
For context, SK Group is a conglomerate that has businesses leading in electric vehicle batteries, semiconductor technology & materials (SK hynix that you’re probably most familiar with), and pharmaceutical development & manufacturing.
Many know of the major players like Samsung Electronics and SK Hynix, which dominate the global memory chip market. Together, they hold close to 70% of the overall memory chip market, with SK Hynix recently overtaking Samsung as the world's top DRAM chipmaker with a 36% market share compared to Samsung's 34%. But these are all within the infrastructure layer of the stack.
According to the latest report from Futurum Group, SK Hynix faces the most substantial risk as its Wuxi fab facility produces approximately 40% of the company’s DRAM, and the Dalian fab contributes 25% of NAND output. License delays or denial could disrupt node migration to 1a nm in DRAM and limit production capacity in DRAM and NAND in China. Meanwhile, Samsung’s Xi’an fab contributes an estimated 40–50% of its total NAND output and ~10% of global NAND supply. The restriction may delay the firm’s ongoing technology transition for its NAND products from V6 (128-layer) to V8 (236-layer) and V9 (286-layer) NAND, as well as the future expansion of production capacity. [See full report here, and shout out to
]
The Building of Sovereign AI
In the early months of this year, due to the political uncertainty, the aftermath of dealing with long-COVID impacts, and increasing international competition in key economic drivers, consumer sentiment was low, and everyone looked to the new potential presidential candidates for clear guidance and direction to usher the country forward.
Both the left and right parties emphasized their “shared perception that AI is a future growth driver for Korea,” though different during the election campaigns, according to Korean news outlet ChosunBiz.
Coming up on top was the Democratic Party’s Lee Jun-seok, who became the 14th president of South Korea. Despite President Lee’s political philosophy that advocates for a smaller government, he has actively pushed for the development of technologies like AI in the role of corporations, and reforming regulations that limit technological development is the responsibility of the new government.
Lee has consistently emphasized the importance of AI policy and made a pledge of 100 trillion won (USD 735 billion) into the sector as one of his signature campaign pledges. His AI platform also included a proposal to develop a “sovereign AI” tailored specifically to South Korea’s cultural and historical context.
All of this is unfolding in front of our eyes. On June 15, the presidential office tapped Ha Jung-woo, head of the AI Innovation Center at Naver, as the country’s first-ever senior presidential secretary dedicated exclusively to AI policy. For context, Naver is one of Korea’s largest tech companies, and its products consist of a powerful search engine, popular blogging and community platforms, a digital comics service, and the nation’s most popular messaging app (LINE). You can think of it as Korea’s equivalent to China’s Tencent plus Baidu.
According to Altos Ventures, Naver began developing its own large language model, HyperCLOVA, in May 2021.
“Korea has a strong sense of technological sovereignty, and there is considerable public resistance to government use of foreign AI services. Naver has aligned its strategy with this sentiment and positioned itself accordingly. As a key player in Korea’s digital rise, Naver was the first private company to advocate for AI sovereignty. Since the recent change in government, two former Naver executives have been appointed to cabinet-level roles, increasing the company’s political influence.
Naver has since upgraded its model to HyperCLOVA X and expanded into multimodal capabilities similar to GPT-4. While not officially part of the government’s Sovereign AI initiative, Naver is its closest private-sector partner from a technical and infrastructure perspective. Naver’s unique position as one of the only domestic firms with both cloud and search technology in-house has made this possible.
If HyperCLOVA becomes widely used by startups in Korea, it would mark a deeper shift beyond just model performance. It would represent a fundamental change in the country’s platform strategy.”
This newly established AI-focused unit within the government will be at the core of leading the nation’s sovereign AI efforts, according to Chosun Daily. And on the same day, SK Group unveiled plans to build South Korea’s largest AI data center in Ulsan’s Mipo National Industrial Complex in partnership with Amazon Web Services, indicating clear top-down efforts to push forward Korea’s AI capabilities.
And just a few days ago, according to a Bloomberg report, Korea Investment Corp’s CEO, Park Il Young, stated that the nation’s sovereign investment arm, which manages US$206.5 billion in assets, is exploring tech investment opportunities in China, specifically in the AI sector.
And he highlighted, in particular, plans to increase allocations to tech startups and venture capital (VC) funds as part of its broader push to deepen exposure to artificial intelligence (AI) and other disruptive technologies.
VCs and AI Startups
So this finally brings me back to my meeting with Altos Ventures. This VC was referred to as “the Sequoia of Korea” by one of their LPs when I spoke to him, and it has backed some of South Korea’s most successful internet companies. It was founded in 1996 by three founders: Han Kim, Ho Nam, and Anthony Lee, two of whom are of Korean heritage. Although its headquarters are based in Silicon Valley, the Korean ties have meant that the fund introduced this Silicon Valley capital playbook to Korea, in some ways bringing back their expertise.
The VC has been widely recognized for being the first investor in fintech unicorn Toss, and an early investment into Danggeun Market, also known as Karrot in some regions, a large second-hand marketplace.
I met with Robbie Oh, an investor at Altos in Korea who focuses on Web3 and AI investments. Below are some of his insights into the space.
“Korea’s AI industry is advancing at a remarkable pace, especially when it comes to the speed of adoption and real-world consumer impact. The B2C sector is particularly responsive, where AI adoption tends to happen quickly and with significant user engagement. For instance, soon after ChatGPT’s launch, Korea became one of the fastest-growing markets for OpenAI’s paid subscription service, second only to the United States.
This mirrors the rapid uptake of smartphones in the past and reflects a broader cultural tendency in Korea: new technologies are embraced not as passing trends, but as tools for real productivity. The result is a tech-savvy user base that is highly open to experimentation and quick iteration. Even under the previous administration, there was growing interest in AI at the government level. Under the new administration led by President Lee, this momentum has accelerated. The government has announced plans to invest what amounts to nearly 100 trillion Korean won(nearly 77 billion USD) by 2027, marking an unprecedented national commitment to building AI infrastructure and securing technological sovereignty.”
He highlighted that, overall, Korea has had very strong B2C startups, and consumer adoption of new technology is very high. Despite the understanding that overall East Asia populations have been very receptive to new tech adoption, I was still shocked to hear that OpenAI’s second-largest market (enterprise and pro) is South Korea.
So far, most of this investment has been directed toward infrastructure, such as data centers, AI semiconductors, and national LLM development. While this tactical, infrastructure-led approach has laid important groundwork, creative productization and consumer-facing services are still in the early stages.
B2C AI
From Altos’s perspective, this is a pivotal moment. The priority now is not just to develop the core technology, but to enable fast cycles of product validation and user feedback. Korea is uniquely positioned as one of the few markets where such rapid experimentation is not only possible but culturally encouraged.
Confirming my research, he stated that there is an expected USD 1-2 billion injection into GPU purchasing for the nation’s sovereign AI research. While AI vertical startups are still in their nascent stages, Korea has yet to see a leading LLM startup rise domestically. He’s seeing more and more innovative startups in the space that are integrating DeepSeek and OpenAI alike, trying out new creative ways to engage with consumers.
One interesting example, reflecting the current societal and cultural shifts in Korea, is the success of AI in the avatar business. As Oh shared, one of the most widely adopted AI apps is a companion chatbot that acts like one's boyfriend. As Korea's marriage rate has significantly decreased in recent years, and fertility rate has dropped to as low as 0.78 births per woman as of 2022, in contrast to 1.66 in the U.S. and 1.80 in Japan, according to the World Bank, one of the lowest amongst developed nations globally.
“The rise of companionship AI shows that it addresses emotional and social needs. For example, AI boyfriend chatbots and avatar-based chat services have gone beyond demos and secured loyal user bases. These products reflect broader social trends, including low marriage and birth rates, and respond to growing emotional gaps,” said the early-stage tech investor.
Automation
Meanwhile, as we know, Korea is a large manufacturer of cars and automation, and he’s seeing AI-driven simulation use cases being explored by startups where AI is being utilized in the design stage, largely boosting efficiency. Robbie Oh added,
“This has enabled the quick adoption of AI simulation tools and AI-integrated design software in industrial settings. These are not tech showcases, but practical tools that solve real problems.
We believe Korea’s greatest opportunity lies in its cultural sensitivity and boldness in experimentation. The market may be small, but it is one of the fastest environments for turning ideas into products with global potential.”
Finally, he added, it’s a talent shortage in South Korea that is the biggest bottleneck right now. Most of the smartest young people want to become doctors, as it's still seen as the most prestigious and stable profession in the country. But eventhough the eventual AI market isn’t as big as China or the U.S. (for context the most widely used app KaKaoTalk has ~49 million MAU that is nearly the whole population), he is confident in consumer adoption and innovation in a future AI-driven world, especially in niche categories because East Asian consumers as a whole has always been very receptive and willing to try new technology, just see everything from Line to Kakao, Groupon to Toss.
China, Japan, and South Korea’s consumer-facing internet apps are significantly ahead of American consumer apps in terms of creativity and functionality, and this trend is likely to continue in the AI era as well.