Summary
OpenAI is deepening its presence in South Korea across both enterprise adoption and AI infrastructure.
ChatGPT Enterprise adoption in Korea has increased roughly 28-fold over the past year, while weekly active use of ChatGPT Work and Codex has risen about 14-fold in six months. At the same time, OpenAI is working toward local inference capacity and exploring a broader infrastructure footprint in Korea.
Upstream, Samsung Electronics and SK hynix are participating in OpenAI’s Stargate infrastructure initiative through advanced memory cooperation, while OpenAI says its next-generation semiconductor collaboration with Samsung is also progressing.
For Korean capital markets, the key distinction is increasingly clear: memory-related AI demand is already visible in earnings, while the broader OpenAI-linked infrastructure opportunity remains largely strategic optionality.
Key Takeaways
- ✓ ChatGPT Enterprise adoption in Korea has increased roughly 28-fold over the past year, showing rapid penetration of generative AI into Korean corporate workflows.
- ✓ Samsung Electronics and SK hynix are positioned within OpenAI’s infrastructure strategy through advanced memory cooperation under the Stargate initiative.
- ✓ OpenAI says its next-generation semiconductor cooperation with Samsung is progressing, although the technical scope, manufacturing role, and commercial structure have not been disclosed.
- ✓ For Korean investors, advanced memory provides the clearest near-term earnings exposure. Data centers, local inference infrastructure, and broader enterprise AI offer potentially meaningful but less quantifiable optionality.
- ✓ The central investment question is whether strategic partnerships with OpenAI eventually create incremental earnings beyond what is already being generated by the broader global AI infrastructure cycle.
What Happened
OpenAI marked the first anniversary of its Korean operation by outlining a broader expansion of its business and infrastructure ambitions in the country.
ChatGPT Enterprise adoption in Korea has increased roughly 28-fold over the past year, while weekly active users of ChatGPT Work and Codex have grown about 14-fold over the past six months. The figures point to rapid adoption of OpenAI tools across Korean enterprises and professional workflows.
OpenAI is also working toward local inference capacity in Korea and discussing the infrastructure required to support it. Specific investment amounts, data-center locations, and implementation timelines have not yet been disclosed.
Upstream, OpenAI said cooperation with Samsung Electronics on next-generation semiconductor technology is progressing, while the exact architecture, manufacturing role, and commercial structure of the partnership remain undisclosed.
Samsung Electronics and SK hynix are also participating in OpenAI’s Stargate initiative through advanced memory cooperation. Under the partnership framework, the companies plan to expand production capacity toward a targeted scale of up to 900,000 DRAM wafer starts per month.
The 900,000-wafer figure indicates the potential scale of future AI infrastructure requirements. It should not be interpreted as contracted OpenAI demand or a guaranteed procurement volume.
The developments show that OpenAI’s relationship with Korea is expanding beyond software adoption into semiconductors and physical infrastructure, although much of that expansion remains at an early or exploratory stage.
Why It Matters
For Korean public-market investors, the significance of OpenAI’s expansion depends on where economic value ultimately accrues.
The clearest near-term transmission mechanism is memory. Larger AI clusters require increasing volumes of high-bandwidth memory and server DRAM, placing Samsung Electronics and SK hynix directly inside the global infrastructure spending cycle. Unlike many other parts of the OpenAI relationship, this demand is already visible in reported earnings.
The next layer is semiconductor optionality. Samsung’s next-generation chip cooperation with OpenAI could broaden its exposure beyond memory, but the project remains too undefined to model as a specific revenue stream or foundry opportunity.
Further downstream, local inference and data-center development could expand the addressable opportunity into cloud, networking, construction, power, and enterprise infrastructure. These areas remain much less visible in current earnings and should therefore be treated as optionality rather than established financial contribution.
For investors, the distinction is critical. Memory exposure is already generating earnings. The broader OpenAI-linked infrastructure opportunity still depends on execution, capital deployment, and eventual commercial contracts.
OpenAI therefore matters not simply as a potential customer, but as part of a much larger structural expansion in global AI infrastructure demand.
Korea’s relationship with OpenAI is moving from adoption toward infrastructure.
The first phase was demand. Korean consumers and enterprises became major users of generative AI. The next phase could bring Korea deeper into the physical stack required to operate AI at scale, beginning with memory and potentially expanding into semiconductors, data centers, and enterprise infrastructure.
For investors, however, these layers should not be valued equally.
Samsung Electronics and SK hynix already have measurable exposure through the broader AI memory cycle. OpenAI-specific semiconductor, inference, and data-center projects remain considerably earlier in their commercial development.
The investment case therefore rests on two different sources of value: earnings that already exist and optionality that could emerge.
Korea’s strategic importance to OpenAI is increasing. The question is how much of that strategic relevance can ultimately be converted into recurring revenue, higher margins, and durable returns on capital for Korean listed companies.