SK expands cooperation with big tech on AI infrastructure, speeding up mega projects [President Lee's U.S. visit]

SK Group has built AI alliances with global big tech companies, including Microsoft, Anthropic and Amazon Web Services (AWS), as well as NVIDIA, with which it is pursuing AI infrastructure cooperation worth a total of $500 billion, or about 731 trillion won. Based on a partnership framework spanning AI factories, AI memory and AI data centers, the company aims to strengthen the execution of its recently announced three mega projects and expand the AI infrastructure ecosystem.
According to industry sources on the 26th, SK said it reached agreements on strategic cooperation centered on AI semiconductors and AI data centers with NVIDIA, Microsoft (MS) and Anthropic at the AI Summit held in San Francisco, United States, on the 24th local time. The move is intended to strengthen the execution of one of the three mega projects announced last month: building a nationwide AI data center network and creating a semiconductor belt. SK plans to accelerate the projects through cooperation with global big tech companies.
The core of the partnership is an AI infrastructure alliance with NVIDIA. The two companies signed a letter of intent to pursue broad cooperation worth more than $500 billion, covering everything from building AI factories to supplying AI memory.
SK Telecom will work with NVIDIA on the supply of next-generation graphics processing units (GPU) and investment cooperation to build Korea's largest AI factory with a capacity of up to 2 gigawatts (GW). This builds on the domestic GW-scale AI factory plan announced by the two companies in June. The plan calls for phased construction starting next year, using NVIDIA's AI factory platform DSX Platform and Rubin, a next-generation AI system equipped with SK hynix's HBM4.
SK hynix will also expand long-term cooperation on AI memory with NVIDIA. The two companies agreed to jointly develop and optimize next-generation HBM needed for the era of Agentic AI and Physical AI, going beyond large language model (LLM) training, in order to strengthen the AI memory ecosystem.
SK is also significantly expanding cooperation with global AI companies beyond NVIDIA. SK hynix is pursuing long-term supply cooperation with Microsoft (MS) for memory used in AI servers. In response to surging demand for AI services, the company plans to supply high-performance memory for data centers over the medium to long term and jointly identify memory solutions optimized for AI workloads.
SK Telecom signed a comprehensive memorandum of understanding with Anthropic to build a GW-scale AI data center in South Korea. Anthropic will take part in SK Telecom's AI data center project and discuss specific areas of cooperation. SK Telecom is also expanding AI infrastructure cooperation with AWS. Based on its experience co-building the Ulsan AI data center, the company plans to expand AI data centers to major domestic hubs and secure GW-scale AI computing capacity in the long term.
Chey Tae-won also met directly with OpenAI CEO Sam Altman to discuss AI infrastructure cooperation, including memory. The two sides have continued their strategic partnership since signing an agreement in October last year on memory supply and the construction of an AI data center in the southwest region.
At the summit, Chey introduced meetings with executives from global big tech companies including NVIDIA, Broadcom Inc., Anthropic and OpenAI, and said, "The scale of AI chips and computing power demanded by companies is growing much faster than expected." He added, "The investment scale announced so far is still a realistic figure when future demand is taken into account." He went on to say, "AI is still only in its early stages," and "South Korea must leap forward as an AI infrastructure hub linking the United States, Europe and Asia, based on more memory and AI data centers."
[email protected] Lim Subin Reporter