South Korea bets on sovereign AI model in record budget

South Korean President Lee Jae Myung unveiled the country’s largest ever budget on Monday, pledging to “boldly invest in AI and future growth engines” with a record push for a sovereign AI model.

Total spending for 2027 will rise 12.8% to 820.9 trillion won ($610.14 billion), including 21.3 trillion won for semiconductors, physical AI and AI data centers, nearly double the 2026 level, CNBC reported. A separate 4.7 trillion won will go toward developing a sovereign AI model with what Lee described as the “world’s highest performance.”

The budget is the largest sovereign AI bet of any country, and it signals that South Korea is recycling the chip export boom into domestic AI capability. The 21.3 trillion won for semiconductors and AI infrastructure is nearly double the previous year’s spending, and the 4.7 trillion won for a sovereign model is a direct investment in building a Korean alternative to the U.S. AI labs that dominate the frontier. The bet is that the country that makes the world’s chips should also make the world’s AI models.

The sovereign AI model

The 4.7 trillion won for a sovereign AI model is the headline number. Lee described the target as the “world’s highest performance,” which is an ambitious claim for a country that does not currently have a frontier AI lab. The investment is a hedge against dependence on U.S. AI companies like OpenAI and Anthropic, and it positions South Korea alongside France, the UAE, and other countries that are building sovereign models to ensure they have control over the AI capabilities that their economies and militaries will depend on.

The sovereign model strategy is a bet that the chip expertise of Samsung and SK Hynix can be translated into model expertise. The chips that South Korea produces are the hardware that every AI lab in the world uses, and the government is betting that the same engineering talent can build the software that runs on those chips. The 2,000 specialists that Lee pledged to train in semiconductor design and the regional chip industry are the workforce that would build the sovereign model, and the university contract departments that link students to Samsung and SK Hynix are the pipeline.

The risk is that building a frontier AI model is not a hardware problem. The U.S. labs that lead the frontier have spent years accumulating the software expertise, the data, and the research talent that goes into training a large language model, and the sovereign model will need to match that expertise to achieve the “world’s highest performance” that Lee described. The 4.7 trillion won is a large investment, but it is small compared to the tens of billions that OpenAI and Anthropic have raised, and the gap in software expertise is not closed by chip manufacturing capability.

The semiconductor and data center push

The 21.3 trillion won for semiconductors, physical AI, and AI data centers is the spending that is most likely to pay off, because it builds on South Korea’s existing strength. The country is the world’s largest producer of memory chips, and the AI buildout is driving demand for the HBM chips that Samsung and SK Hynix make. The budget nearly doubles the 2026 spending on these areas, and the data center investment positions South Korea as a potential hub for AI computing in Asia.

The government also plans to invest heavily in talent, including training 2,000 specialists to “lead semiconductor design and the regional semiconductor industry,” while expanding AI education for students, job seekers and existing workers. South Korea has set up university “contract departments,” where students can enroll in company-linked degrees that offer pathways to jobs at firms like Samsung Electronics and SK Hynix before graduation. Companies help design courses, provide financial support and recruit qualifying graduates.

The contract department model is the right approach to the talent problem, because it aligns the curriculum with the needs of the companies that will hire the graduates. The 2,000 specialists are a small number relative to the global AI talent pool, but they are targeted at the specific intersection of semiconductor design and AI that is South Korea’s comparative advantage, and the pipeline from university to Samsung and SK Hynix is the mechanism that turns the training into industrial capacity.

The export boom

The record spending plan comes as the IMF forecasts South Korea’s economy will grow 3.3% in 2026, its fastest pace since the post-pandemic rebound in 2021. South Korea’s strong economic growth has been driven largely by semiconductor exports as global demand for computing power rises with the AI boom.

Exports surged 93.7% from a year earlier to a record $115.3 billion in October, according to data released by the Korea Customs Service. Shipments totaled $930 billion in the first 10 months of 2026, surpassing the $900 billion mark for the first time. The country posted a $42.5 billion trade surplus in October, with imports rising 36% year on year to $72.8 billion. Lee said South Korea was within reach of $1 trillion in annual exports, citing strong performance in key industries including semiconductors.

The 93.7% export growth is the number that funds the budget. South Korea is running a massive trade surplus driven by chip exports, and the government is recycling that surplus into domestic investment in AI infrastructure and sovereign model development. The $1 trillion export target is within reach, and the budget assumes that the chip boom will continue to generate the revenue that funds the spending. If the AI buildout slows and chip demand falls, the export growth that funds the budget will weaken, and the sovereign AI investment will become a fiscal burden rather than a growth bet.

Lee said the higher spending would result in “virtually no net increase in government debt,” with newly issued bonds used to repay existing debt. The budget also increases spending on defense and social security spending for the youth. Defense expenditure is set to rise 8.2% to 73.3 trillion won. “We will rapidly transform into a smart, elite, and powerful military by establishing a hybrid manned and unmanned system based on AI, robots, and drones,” Lee said, CNBC reported.

What I think

The budget is the right bet for South Korea, because it recycles the chip export boom into the capabilities that the country will need to remain competitive in the AI economy. The 21.3 trillion won for semiconductors and data centers is the spending that is most likely to generate a return, because it builds on the existing strength of Samsung and SK Hynix and it positions South Korea as an AI infrastructure hub in Asia. The contract department model is the right talent strategy, because it links the curriculum to the companies that will hire the graduates and it creates a pipeline from university to industry.

The sovereign AI model is the bet that is hardest to justify. The 4.7 trillion won is a large investment, but building a frontier AI model requires software expertise, data, and research talent that South Korea does not currently have in the same concentration as the U.S. labs. The “world’s highest performance” target is aspirational, and the more realistic outcome is a model that is competitive in Korean-language tasks and in the specific applications that the Korean military and industry need. That is a worthwhile goal, because it reduces dependence on U.S. AI companies and it gives South Korea control over the AI capabilities that its economy and defense will rely on. But it is not the same as building the world’s best model, and the budget should be judged on the realistic outcome, not the aspirational claim.

The defense spending is the component that is most closely tied to the sovereign model. The 8.2% increase and the commitment to a “hybrid manned and unmanned system based on AI, robots, and drones” is a military doctrine that requires AI capabilities, and the sovereign model is the hedge against relying on U.S. AI for military applications. The combination of the chip export boom, the AI infrastructure investment, and the sovereign model development is a coherent industrial strategy, and the budget is the financial expression of that strategy.

The risk is the chip cycle. The 93.7% export growth is the revenue that funds the budget, and the budget assumes that the growth continues. If the AI buildout slows and chip demand falls, the export growth will weaken, the trade surplus will narrow, and the government will face a choice between cutting the AI investment or running a deficit. The IMF’s 3.3% growth forecast is based on the chip boom continuing, and the budget is a bet on that forecast. If the forecast is right, the budget is affordable. If the forecast is wrong, the sovereign AI investment becomes a fiscal burden, and the “virtually no net increase in government debt” claim does not hold. The bet is the right one to make, because the cost of not investing in AI is larger than the cost of investing and being wrong, but the risk is real and it is the one that the budget does not hedge.