SK's AI Megaproject: Inside Korea's Largest Data Center by 2027

SK Group is partnering with Amazon Web Services to build Korea's largest AI-dedicated data center in Ulsan. The massive facility will house up to 60,000 graphics processing units when completed in 2027. It features innovative hybrid cooling technology to handle the extreme heat from high-performance AI processors. The project represents a major investment in South Korea's AI infrastructure capabilities.

Key Points: SK Group AWS Build Korea's Largest AI Data Center in Ulsan

  • Facility will house 60,000 GPUs, 60x current SK Telecom capacity
  • Uses hybrid air-liquid cooling for high-performance processors
  • Covers 66,000 square meters with five-story structure
  • AWS CEO personally visited site during APEC Summit inspection
2 min read

SK AI Data Center to become Korea's largest AI facility by 2027

SK Group and AWS are constructing Korea's largest AI data center in Ulsan, featuring 60,000 GPUs and advanced hybrid cooling technology by 2027.

"The DLC method precisely controls chip temperatures, ensuring a stable training environment - SK ecoplant manager"

Seoul, November 3

Dozens of workers and five pieces of heavy equipment busily engaged in foundation work at a construction site of SK AI Data Center in Ulsan in the Ulsan Mipo Industrial Complex late last month, as per a report by Pulse, the English service of Maeil Business News Korea.

This facility is SK Group's AI Data Center, currently being built in collaboration with Amazon Web Services (AWS). Though still in the excavation phase, by 2027, it will become the country's largest AI-dedicated data center.

"The site covers more than 20,000-pyeong (66,000 square meters) and the completed building will be five stories above ground, reaching a height of 47.7 meters. When completed in 2027, SK AI Data Center Ulsan will be capable of housing up to 60,000 graphics processing units (GPUs), over 60 times the capacity of SK Telecom Co.'s currently operating Haein cluster, which has about 1,000 GPUs," the report said.

The main focus of this data center is cooling efficiency. Unlike conventional server data centers, this facility is designed to run not only CPUs but also GPUs and MPUs in parallel.

Creating an entirely new cooling and power infrastructure was the top priority as the high-performance processors generate more than 10 times the heat and power consumption of CPU-based systems.

The report stated that SK introduced a hybrid cooling system that combines air and liquid cooling. The Ulsan data center uses direct liquid cooling technology, in which coolant is circulated directly over the chips inside the servers.

The refrigerant comes into direct contact with heat-generating components such as GPUs and CPUs.

"The DLC method precisely controls chip temperatures, ensuring a stable training environment," the report quoted an on-site manager at SK ecoplant.

He added that SK leveraged synergies among its affiliates to establish a reliable power supply network.

"AWS is also heavily invested in this project. Matt Garman, who recently took over as CEO of the U.S. cloud giant, personally visited the Ulsan construction site to inspect progress while attending the APEC CEO Summit," the report said.

- ANI

Share this article:

Reader Comments

P
Priya S
The cooling technology is fascinating! Direct liquid cooling for GPUs makes so much sense given the heat generation. Hope Indian data centers are also adopting such advanced cooling solutions. Our climate demands it even more!
A
Aditya G
While this is impressive, I wonder about the environmental impact. 60,000 GPUs will consume enormous electricity. Hope they're using renewable energy sources. Sustainable AI development is crucial for our planet 🌍
S
Sarah B
Great to see AWS collaborating on such projects! This shows how global partnerships can accelerate AI infrastructure development. Indian companies should form similar alliances to boost our tech capabilities.
K
Karthik V
The timeline is ambitious - 2027 completion. Hope they can maintain quality while meeting deadlines. India's AI mission should also set such clear targets with proper execution plans. Jai AI! 🇮🇳
M
Meera T
The power supply network collaboration between SK affiliates is smart thinking. Indian conglomerates should learn from this synergy approach rather than working in silos. Together we can achieve more! 🤝

We welcome thoughtful discussions from our readers. Please keep comments respectful and on-topic.

Leave a Comment

Minimum 50 characters 0/50