South Korea just decided it can't afford to watch the AI race from the sidelines anymore.
On Tuesday, the country's Ministry of Science and ICT announced a 4.7 trillion won ($3.5 billion) state-backed programme to develop a frontier AI model, starting in March 2027. It's one of the biggest government bets on AI ever made by a country — and it has a clear target: the leading Chinese open-source models.
THE PLAN
Reuters reports the programme works like a national startup pitch. After parliament approves the 2027 budget in December, the ministry will pick a lead developer through a competitive tender — potentially as early as February. The money isn't grants: it's equity investment from the state, combined with private funding, so the returns can eventually flow back to the public.
To give the project a fighting chance, the government plans to concentrate computing chips, data, and AI talent on it. Participation is open to single companies, consortiums, or special-purpose vehicles — startups, big firms, universities, and researchers can all bid for the lead role.
SEPARATE FROM THE HOME-GROWN PROGRAMME
This is a new track, not a replacement. South Korea's existing national AI foundation model programme continues, with two final teams to be chosen in a third-stage review this February. That programme's focus will shift toward models tailored to specific industries and real-world applications, while the new frontier project chases top-tier performance.
WHY NOW
According to Second Vice Minister Ryu Je-myung, speaking at a press briefing Friday, the government and domestic companies agree that global competition for advanced AI models has entered "a completely different phase" since the preview of Anthropic's Claude Mythos emerged in April. The old competition-based approach, he said, needs investment "on an entirely different scale."
He's not wrong about the scale race. Europe's Mistral raised 3 billion euros in September at a 21 billion euro valuation. Chinese open models like DeepSeek and Qwen keep shipping at a fraction of US training costs. The frontier is getting expensive to hold — and governments are starting to notice.
AIPOST'S TAKE
The AI race is quietly becoming a nations' race. Yesterday's breakthrough was a startup announcement; today's is a $3.5 billion sovereign bet. The real question isn't whether Korea can fund a frontier model — it's whether a national programme can move at the speed of the labs it's chasing. March 2027 is the starting gun, not the finish line.
Comments
Post a Comment