US Startup to Build South Korea’s Biggest AI Data Center

Reflection AI is partnering with Seoul-based conglomerate Shinsegae Group to build a huge new AI data center in South Korea.

The companies revealed their intentions in a media statement on March 16, in which they confirmed they had signed a memorandum of understanding for the project.

The partners did not release financial details of the deal, although the investment is likely to run to several billion dollars. They also did not disclose the location.

However, the companies said the data center will consume up to 250 megawatts -- roughly the same as a small city -- while local media in South Korea reported Shinsegae billed it as the “largest AI-dedicated data center built by a single company” in the country.

Each party will bring its own expertise to the project. Reflection, founded in 2024 by two former Google Deep Mind researchers and based in New York City, will provide the technical know-how, including Nvidia graphics processing units (GPUs), open weight foundation models and full-stack engineering. The AI chip giant was one of Reflection’s biggest backers when the startup raised $2 billion in October.

Related:OpenAI Now Valued at $852B After Funding Round

Shinsegae, meanwhile, will take responsibility for the infrastructure, including site selection, physical architecture, power, permitting and financing. Korean media say the conglomerate -- previously known for credit cards and department stores -- is also likely to use the data center as a springboard into more AI business ventures with the ultimate vision of positioning itself as "Korea’s Amazon"

The agreement is notable for a number of other reasons. It's a major step forward in the effort to provide South Korea with sovereign frontier AI capabilities built and operated in the country. This sovereign AI campaign was also recently boosted by Hyundai Motor Group’s commitment to spend $6.1 billion on an innovation hub for robotics and AI in the port city of Gunsan.

Also, by using Reflection’s open weight models -- with which users can download and run them locally without needing access to code or training data -- South Koreans will have visibility into the tech underpinning their critical systems. Reflection is said to be developing models customized for Korean language and culture.

“Every country is realizing that AI sovereignty is existential, which requires open models. We have a narrow window to ensure the foundation of intelligence remains open and accessible to all, rather than controlled by a few," Misha Laskin, CEO of Reflection, posted on X.

There is also a geopolitical aspect to the deal, with the agreement demonstrating the potential of President Donald Trump’s AI export program, in which the U.S. administration aims to strengthen relationships with allies by promoting the export of American AI tech packages.

Related:Microsoft Commits $1B to Thailand's AI future

The program is clearly designed to prevent Chinese dominance of the sector, and the media release for the Reflection-Shinsegae deal acknowledged: “The partnership reinforces a vital geostrategic alliance between the United States and a valued Indo-Pacific partner.”

About the Author

Contributing Writer

Graham Hope has worked in automotive journalism in the U.K. for 26 years, including spells as editor of leading consumer news website and weekly Auto Express and respected buying guide CarBuyer.