South Korea's 2027 Budget Nearly Doubles AI Spending, Eyes Sovereign Model at World-Best Performance
South Korea's president has proposed a record 821-trillion-won budget for 2027, with 21.3 trillion won — nearly double the previous year — allocated to semiconductors and AI, and a stated goal of building a sovereign AI model with the world's highest performance.
President Lee Jae Myung on November 2 delivered his annual budget speech to the National Assembly and laid out what South Korea intends to spend in 2027: 821 trillion won in total, a year-over-year increase of 12.8% and the largest budget in the country’s history. The headline number is notable, but the structural shift underneath it is larger. Of that 821 trillion, 21.3 trillion won will flow into semiconductors, physical AI, and AI data centres — nearly double the previous year’s allocation for the same sectors.
Lee framed the increase around three mega-projects, which the government clusters under the AI-and-chips umbrella: semiconductor manufacturing, physical AI (robotics and embodied systems rather than purely software-based models), and the data-centre infrastructure needed to train and run them. The 21.3 trillion won figure covers all three, and Lee asked the National Assembly for bipartisan support to pass the budget as presented.
A separate line item within that spending caught a different kind of attention. 4.7 trillion won has been earmarked for frontier AI development, aimed squarely at building homegrown models that can compete at what the government describes as the top tier of global performance. Lee put a sharper label on it: the programme’s stated goal is to acquire a sovereign AI model with the world’s highest performance.
“Sovereign AI” is a term that has been gaining traction across several countries — France, the UAE, India, and others have used it to describe models built domestically with national strategic value rather than reliance on foreign providers. South Korea’s version appears to go further by attaching a performance claim to it rather than just a sovereignty claim. Whether 4.7 trillion won — roughly $3.4 billion at current exchange rates — is enough to reach the top of global model rankings is a separate question, but the budget puts a concrete number behind the ambition in a way that framing alone does not.
The context for the broader spending push is the country’s existing entrenchment in the chip supply chain. South Korean memory makers already dominate global DRAM and NAND production, and the government has treated AI-era compute demand as an opportunity to extend that position upward into logic chips, packaging, and the end-to-end AI stack. Doubling the budget year-over-year is less a pivot than a deepening — though the frontier-model line item does represent something genuinely new in the spending mix.
Lee’s coalition holds a majority in the National Assembly, which makes passage of the budget likely, though not automatic — the speech was also a request for bipartisan support rather than a simple announcement, suggesting some negotiation remains. For the global AI landscape, though, the more immediate signal is that another major economy has moved from policy documents and policy speeches into line-item budgets with real attached figures. The sovereign-AI model and the physical-AI robotics push are both bets on specific technical bets that other countries are also placing, but the scale of spending South Korea has committed makes it one of the most aggressive state-level investments in the category to date.