South Korea's 'AI Squid Game' Eliminates Tech Giants in High-Stakes AI Race
Late last year, over a thousand people packed a Seoul convention center. Elite engineers presented their newest artificial intelligence developments. The auditorium buzzed with complex technical terms. For Korean tech executives and investors, this event was essential. Rumors swirled that the science minister skipped a parliamentary hearing about a data breach at Coupang Inc. He reportedly came to watch the demonstrations instead. This marked a rare, high-stakes contest to find the best homegrown AI models.
Tens of thousands watched the event via livestream. Cheers and shouts erupted each time contestants took the spotlight. For some teams, this appearance would be their last. People now call this competition the "AI Squid Game." The name references the popular Netflix survival drama. The government-sponsored tournament started in August. It will run for more than a year. Its goal is to identify leaders in South Korea's push to become an AI powerhouse.
A Ruthless Format Inspired by Pop Culture
The format is ruthless, much like the show that became a symbol of Korea's cultural influence. Teams' AI foundation models face evaluation every six months. A panel of judges, overseen by the Ministry of Science and ICT, decides who gets eliminated. On a recent Thursday, the judges made surprising cuts. They eliminated the unit of Naver Corp. Officials criticized it for using foreign technology. They also dismissed the AI subsidiary of NCSoft Corp. This was the only team led by a woman. The decision came after evaluations of its model's performance.
This development shocked many. Observers expected only one team to be cut. According to the science ministry, subsidiaries of LG and SK Group remain in the running. AI startup Upstage also stays in the contest. The ministry announced it would select an additional team for the next stage. Two winners are expected in early 2027.
"It may look like a spectacle, but this is a high-stakes contest that will shape Korea's future," said Chanjun Park. He is an assistant professor in the School of Software at Soongsil University in Seoul.
Global Ambitions in a Competitive Field
The government bets that fierce competition will drive rapid innovation. It aims to bolster a homegrown AI industry. Currently, US and Chinese players dominate the field. Officials state the ultimate goal is to develop indigenous open-source models. These models should match global frontier models like ChatGPT and DeepSeek.
South Korea joins a growing list of nations determined not to fall behind in AI. France also aims to become a third AI hub. Its government invests billions in AI projects and promotes local players. Saudi Arabia and the UAE leverage massive capital. They acquire advanced computing infrastructure and fund domestic models like Falcon LLM. Other countries, including Singapore, Japan, Canada, and India, work on their own sovereign AI strategies.
Korean officials show strong commitment to building homegrown foundation models. The country is already a heavy AI user. It has one of the world's biggest markets for ChatGPT among paying users. South Korea also has the world's highest density of industrial robots. This sector faces radical changes driven by AI technology.
Positioning as a 'Full-Stack AI Country'
Science and ICT Minister Bae Kyung-Hoon was appointed in July. He believes South Korea can capitalize on its unique standing. He calls it a "full-stack AI country." The nation possesses its own advanced memory chips, domestic cloud services, and applications.
"There are several countries vying to be No. 3, but we don't see ourselves as just another contender in that pack," Minister Bae said. "We believe we have a real shot at becoming a serious global player — one capable of challenging the top two."
Contenders and Emerging Controversies
For participants, this contest could make or break their AI ambitions. The government provides access to GPUs and datasets. Winning teams gain an influential and possibly lucrative position in the AI ecosystem. Eliminated teams lose access to the GPUs. Since the announcement, Naver's shares have fallen. Shares in LG affiliates and SK Telecom Co. have rallied.
LG AI Research is widely seen as the top contender. Its model, named K-Exaone, achieved the highest overall score in the latest stage. SK Telecom also draws attention due to its model's sheer scale. Its A.X K1 model boasts 519 billion parameters. That is roughly double the scale of K-Exaone. It nears the level of the latest DeepSeek models, though it trails the most advanced ChatGPT version.
Generally, higher parameters indicate more sophisticated models with greater reasoning power. Parameters reflect the number of variables used in machine learning. However, this can also mean heavy computing and energy requirements. Upstage, a five-year-old startup aiming for an IPO, remains in the race. It unveiled a powerful and cost-efficient large language model.
The government plans to soon accept applicants for a fourth contestant. The next round of evaluations is expected mid-year. Teams eliminated earlier are allowed to apply. Recently, two startups cut last year, Motif Technologies and Trillion Labs, announced they would re-enter the race.
Debate Over 'Homegrown' AI Definitions
The competition sparked debate over what constitutes real, homegrown AI. The government emphasizes that AI foundation models must be trained end-to-end. They require proprietary learning processes, data, and architecture built from the ground up. As teams fight to match global leaders, claims persist that some used foreign technology. This contradicts the official requirement to build models from scratch.
The controversy peaked when a rival firm accused Upstage's Solar Open model. It claimed the model closely resembled a Chinese model, GLM-4.5-Air. Upstage denied the charge and convened a public verification session. This forced an apology from the accuser. Yet scrutiny spread to other participants.
Critics accused Naver Cloud of incorporating vision encoders and weights similar to architectures from Chinese labs like Alibaba. SK Telecom's model also faced scrutiny for alleged architectural similarities to a Chinese model. Building on existing architectures is common practice in AI development. For many, the time and cost to create a frontier model from zero are prohibitive. Many players lean on foundational research by leaders like Meta Platforms Inc.
In eliminating Naver Cloud, government officials told reporters its partial use of an existing Chinese model violated program requirements. They clarified they saw no reason to disqualify SK Telecom or Upstage. Naver said it respected the panel's decision.
High Pressure and Personal Sacrifices
Even without disputes, participants say emotions run high. Engineers on different teams work around the clock. They inspect dataset quality and tweak models to boost performance. This work can be both repetitive and time-consuming.
Lee Jinsik leads LG AI Research's Exaone Lab. He said he has been working so late that months have passed since he saw his toddler awake. Yet he and other project leaders express excitement. They feel part of a competition with major consequences for Korea's technology industry and broader economy.
"This is nerve-racking. But this is exactly how K-pop took off," said Upstage CEO Sung Kim. He compared it to audition shows that launched early stars of Korea's famous music scene. "I'm pretty sure the same thing will happen in the field of AI and science in Korea."
A Familiar Pattern of National Ambition
South Korea is not alone in pursuing its own AI capabilities. Yet this initiative follows a familiar national pattern. The country of 51.6 million rose from the ruins of the Korean War. It achieved this through a unique synergy of state-led investment and private-sector ambition. By channeling funding into critical industries, the government repeatedly transformed a once-impoverished nation into a thriving capitalist democracy.
In 1968, South Korea still depended on US food aid. Then-President Park Chung Hee pushed through the Seoul-Busan highway. Many dismissed the project as a reckless fantasy at the time. Chung Ju Young, the late founder of Hyundai Group, backed it anyway. He later built the world's biggest shipbuilding company in Korea. This helped catalyze the country's industrial takeoff.
Three decades later, President Kim Dae Jung directed another pivot. He laid ultra-fast broadband that sparked the country's digital rise. Now, President Lee Jae Myung frames artificial intelligence as the next great leap. In a November budget speech, Lee pledged to build an "AI Highway." This national infrastructure includes compute power, data, and the rollout of 6G technology.
By invoking the same logic that built roads and the web, Korea bets it can once again engineer its way to the top. This time, the private sector acts as the primary engine for the state-led initiative. Chaebol and budding startups lead the charge. SK Group Chairman Chey Tae-won made an online video appearance at the December event. He emphasized how important a homegrown sovereign AI model is for the country's future.
Pressure, Pride, and an Engineer's Dream
All that attention brings pressure as well as pride, said Kim Tae Yoon. He leads SK Telecom's foundation model team. He says his hair has gone gray since the contest started. In a break from tradition, he was too busy to help with the annual harvest at his parents' tangerine farm on Jeju Island.
"To be involved in a project this important is a rare honor," Kim said. "It's an engineer's dream."
