OpenAI announced at a press conference in Seoul on September 9, 2026 that it is expanding its collaboration with Samsung Electronics. The expanded partnership covers joint development of next-generation custom AI chips, supply of high-bandwidth memory, and cooperation on data center infrastructure. As OpenAI pushes to bring more of its chip supply in-house, Samsung is becoming an indispensable partner on both the manufacturing and research sides.

OpenAI Accelerates Its Push Into Custom Silicon

Harrison Kim, General Manager of OpenAI Korea, described the joint production and research on next-generation chips as "one of the areas where we have gained the most recognition" in the companies' collaboration. OpenAI had already approached Samsung and SK Hynix as early as October 2025 to supply chips for its large-scale infrastructure initiative, Stargate, and this announcement takes that relationship a step further.

OpenAI previously developed its custom inference chip, Jalapeño, together with Broadcom, and began mass production with TSMC in June 2026. OpenAI CFO Sarah Friar has said that by using the company's own AI models in the design process, the Jalapeño chip was completed in just nine months. The new work with Samsung is seen as targeting a successor generation of chips, though neither company has disclosed technical specifications.

An Industry-Wide Move Away From GPU Dependence

OpenAI is far from alone in pursuing its own chip designs. Google has its TPUs, Amazon has Trainium, and Meta has its MTIA chips, reflecting a broader trend among major AI companies to reduce their reliance on NVIDIA GPUs. By deepening its relationship with Samsung on top of its existing work with Broadcom, OpenAI appears to be spreading chip design, manufacturing, and memory sourcing across multiple partners to secure supply chain stability and stronger negotiating leverage on pricing.

Memory and Data Center Infrastructure Are Also in Play

Running large language models requires not just processing power but also fast memory. Samsung and SK Hynix have already signed letters of intent to supply memory for the Stargate data center project, and Samsung has begun shipping samples of its next-generation HBM4E high-bandwidth memory to major customers. Beyond the core semiconductor business, other Samsung group companies, including Samsung SDS, Samsung C&T, and Samsung Heavy Industries, are said to be exploring broader infrastructure partnerships, including floating data centers.

Deep ChatGPT Adoption Inside Samsung Is Also a Factor

The expanded partnership is also underpinned by how deeply ChatGPT has been adopted inside Samsung. Kim described Samsung as "one of the largest-scale deployments of ChatGPT globally," noting that employees in Samsung's Device eXperience division, which oversees mobile devices and home appliances, can use ChatGPT alongside Google's Gemini and Anthropic's Claude for their daily work. The number of ChatGPT Enterprise users in South Korea is said to have grown roughly 28-fold year over year as of the end of August 2026. South Korea has also become the largest market for paying ChatGPT subscribers outside the United States, with adoption spreading across legal, sales, and finance functions.

Summary

The relationship between OpenAI and Samsung now extends well beyond simple component sourcing, spanning chip design, memory, data center infrastructure, and even software adoption. For OpenAI, which is racing to bring more of its silicon in-house, its ties with Samsung's manufacturing scale and leading-edge memory technology look set to become an important pillar supporting the Stargate initiative. Attention now turns to when the two companies will reveal concrete details of the next-generation chip.

References