Microsoft is strengthening its partnership with South Korean memory chip company SK Hynix as part of a broader strategy to reduce its reliance on NVIDIA’s AI hardware.
This is reported by South Korean media based on sources close to Microsoft’s private CEO Summit 2026 in Redmond. Neowin cites those media outlets in its own article.
During the multi-day event, approximately one hundred international executives and policymakers are gathering to discuss generative AI, cloud infrastructure, and the growing demand for AI data centers. According to reports, SK Hynix CEO Kwak Noh-Jung is also participating in discussions with Microsoft CEO Satya Nadella and co-founder Bill Gates.
According to insiders, the presence of the SK Hynix CEO underscores how important the company has become within Microsoft’s own AI chip strategy. Microsoft has long been trying to gain more control over the infrastructure behind AI services, in part by deploying its own chips alongside NVIDIA GPUs.
The Maia 200 plays a key role in this effort. This is Microsoft’s proprietary inference accelerator for AI workloads, which went into operation earlier this year at a data center in Des Moines, Iowa. According to those involved, the hardware delivers a better price-performance ratio than previous generations of AI systems within Microsoft’s infrastructure.
According to reports, SK Hynix is the exclusive supplier of the high-bandwidth memory for the Maia 200. The chip features six memory stacks of 36 GB each, providing a total capacity of 216 GB and a memory bandwidth of 7 TB per second. Such configurations are designed to run large AI models faster and more consistently without delays caused by memory bottlenecks.
Competition in AI infrastructure is intensifying
In addition to memory for Microsoft’s own AI hardware, SK Hynix also supplies DRAM, NAND flash, and high-bandwidth memory to other major players in the AI market. The company collaborates with NVIDIA, among others, and also maintains relationships with cloud companies such as Google and Amazon Web Services.
These developments are part of a broader shift in the AI sector, where hyperscalers are increasingly attempting to design their own chips and infrastructure. By doing so, they hope to better control the costs of AI services and reduce their dependence on a limited number of suppliers.
According to Korean media, the CEO of LG Uplus is the only other South Korean telecom or ICT executive attending the summit in Redmond.