OpenAI is quietly building a chip supply chain that doesn’t run through Nvidia. The latest piece of that strategy, as reported by SamMobile, is a new agreement with Samsung Electronics to manufacture a custom AI chip for OpenAI, announced by OpenAI Korea alongside a broader expansion of the two companies’ collaboration.
The chip doesn’t have a public name yet, and OpenAI hasn’t confirmed what it will do specifically. But the context makes the direction clear. OpenAI’s first custom inference chip, codenamed Jalapeno, was designed with Broadcom and built by TSMC. This Samsung-made chip is almost certainly the next step in that inference roadmap, now with a second manufacturing partner added to the mix. Spreading production across TSMC and Samsung gives OpenAI more supply flexibility and negotiating leverage as demand for AI compute keeps climbing.
Beyond the chip itself, the two companies will work together on advanced memory. OpenAI specifically called out the growing need for faster, higher-capacity memory to support more complex AI workloads. That work ties into OpenAI’s Stargate data center initiative, which already includes SK Hynix as a memory partner. So Samsung is now in the Stargate ecosystem alongside its domestic rival, which is a notable win for Samsung’s semiconductor division at a time when SK Hynix has dominated the HBM memory narrative.
The partnership also has a significant software and enterprise dimension. Samsung is described as one of the largest global deployments of ChatGPT. Samsung’s DX division, which covers smartphones, TVs, home appliances, and laptops, lets employees use ChatGPT internally alongside Google Gemini and Anthropic Claude. That’s a real-world enterprise rollout at serious scale, not a pilot program.
The South Korea numbers are striking on their own. The country is the largest paying ChatGPT user base outside the United States, and ChatGPT Enterprise adoption across South Korean institutions grew roughly 28x year over year, spreading across legal, finance, marketing, and operations teams. OpenAI clearly sees Korea as a critical market, and formalizing the Samsung relationship is part of how it plans to hold that position.
For the broader AI chip industry, this is another signal that the era of near-total TSMC and Nvidia dependence is ending, at least for the biggest AI companies. Anthropic, Google, and Meta are all building or sourcing custom silicon. OpenAI adding Samsung as a fab partner means the competitive pressure on TSMC is real, and Samsung gets a high-profile customer to help close the perception gap with its foundry business.
