South Korea's Ministry of Science and ICT confirmed it will proceed as planned with its homegrown AI foundation model project through February next year, despite some speculation it could be halted.

Separately from the project, the ministry said nothing has been decided on whether to set up a special-purpose company (SPC) to develop frontier-level AI models. It stressed the need to concentrate funding on developing top-tier AI models.

In a press explanation released on Sunday, the ministry said there has been no decision on halting the homegrown AI foundation model project or on whether to establish an SPC related to developing top-tier AI models.

Earlier, some observers raised speculation that the government was reviewing a halt to the project and that discussions were under way on developing frontier AI models through a separate SPC.

With the project proceeding as planned, the government is expected to select a final 2 companies in February after a third-round evaluation.

After that, policy focus is likely to tilt toward frontier AI development.

The ministry said in the press explanation that it is deeply aware of the need to concentrate financial support on developing top-tier AI models.

As part of that, the government has reviewed options including establishing an SPC for frontier AI development or using an equity investment approach as one of its policy options.

Vice Prime Minister Bae Kyung-hoon (배경훈), who also serves as science and ICT minister, also cited in a July media briefing the need to pursue development of a frontier AI model at a global top level separately from the homegrown AI foundation model project.

Bae estimated at the time that about 10,000 of Nvidia's next-generation graphics processing units (GPUs), Vera Rubin, would be needed for frontier AI model development, and that the GPU cost alone would total about 3.5 trillion won. He said it would be difficult in practice for the government to concentrate resources of that scale on a specific company, and that it was reviewing ways to raise funding through investment-type research and development, such as government equity investment or setting up an SPC.