#South Korea is betting 4.7 trillion won on frontier AI, as the AI race escalates from a corporate contest to a national one.
South Korea’s Ministry of Science and ICT plans to launch a frontier AI model development project in March 2027, with a total budget of 4.7 trillion won, or about $3.5 billion.
After the National Assembly approves the 2027 budget in December, the project plans to select lead developers through a public tender. The winners could be announced as early as next February.
What’s really worth watching this time isn’t just the amount of money being invested, but how South Korea plans to do it.
The government plans to combine public equity investment with private funding, while channeling resources into computing chips, data, and talent.
Participation won’t be limited to large corporations. Startups, corporate consortia, universities, research teams, and special-purpose entities will also be eligible.
This signals a shift in the AI race:
In the past, companies like OpenAI, Google, and Anthropic competed to build the best models;
Now, the contest increasingly looks like a competition between nations for computing power, chips, data, talent, and capital.
South Korea has some clear advantages: a strong semiconductor industry, with Samsung and SK hynix already playing important roles in the AI chip supply chain, including HBM.
The potential ripple effects of this 4.7 trillion won investment across the industry could look like this:
Government funding → frontier AI models → demand for computing power → GPUs/HBM → servers and data centers → power infrastructure → AI applications.
So the biggest beneficiaries may not just be AI model companies, but also AI chipmakers, HBM suppliers, server manufacturers, data centers, power providers, and related infrastructure companies.
But the risks are also clear.
Investment in AI model development is growing rapidly. The key question is whether it can ultimately generate commercial revenue.
If the spending continues without producing leading models and practical industry applications, pressure to justify the returns on capital expenditure will keep mounting.
My view is that South Korea isn’t simply trying to catch up in AI. It wants to extend its semiconductor advantage further upstream into the AI industry.
In the future, the AI race may be about more than model parameters. It may come down to who has cheaper, more stable, and more abundant computing power—and who can actually turn it into productivity.
For investors, the AI supply chain is still worth watching, but the amount of government investment alone isn’t enough to focus on.
What matters more is whether it leads to actual orders, revenue, and profits.
Which do you think will benefit first from South Korea’s bet on frontier AI: HBM, data centers, or power infrastructure?
South Korea’s Ministry of Science and ICT plans to launch a frontier AI model development project in March 2027, with a total budget of 4.7 trillion won, or about $3.5 billion.
After the National Assembly approves the 2027 budget in December, the project plans to select lead developers through a public tender. The winners could be announced as early as next February.
What’s really worth watching this time isn’t just the amount of money being invested, but how South Korea plans to do it.
The government plans to combine public equity investment with private funding, while channeling resources into computing chips, data, and talent.
Participation won’t be limited to large corporations. Startups, corporate consortia, universities, research teams, and special-purpose entities will also be eligible.
This signals a shift in the AI race:
In the past, companies like OpenAI, Google, and Anthropic competed to build the best models;
Now, the contest increasingly looks like a competition between nations for computing power, chips, data, talent, and capital.
South Korea has some clear advantages: a strong semiconductor industry, with Samsung and SK hynix already playing important roles in the AI chip supply chain, including HBM.
The potential ripple effects of this 4.7 trillion won investment across the industry could look like this:
Government funding → frontier AI models → demand for computing power → GPUs/HBM → servers and data centers → power infrastructure → AI applications.
So the biggest beneficiaries may not just be AI model companies, but also AI chipmakers, HBM suppliers, server manufacturers, data centers, power providers, and related infrastructure companies.
But the risks are also clear.
Investment in AI model development is growing rapidly. The key question is whether it can ultimately generate commercial revenue.
If the spending continues without producing leading models and practical industry applications, pressure to justify the returns on capital expenditure will keep mounting.
My view is that South Korea isn’t simply trying to catch up in AI. It wants to extend its semiconductor advantage further upstream into the AI industry.
In the future, the AI race may be about more than model parameters. It may come down to who has cheaper, more stable, and more abundant computing power—and who can actually turn it into productivity.
For investors, the AI supply chain is still worth watching, but the amount of government investment alone isn’t enough to focus on.
What matters more is whether it leads to actual orders, revenue, and profits.
Which do you think will benefit first from South Korea’s bet on frontier AI: HBM, data centers, or power infrastructure?