ChatGPT requires significant computational power to process requests, which makes it expensive. Most companies involved in artificial intelligence currently rely on Nvidia chips and cloud providers, but this approach is costly and difficult to scale. Consequently, OpenAI and Samsung have decided to deepen their corporate ties in the chip sector by entering into a new partnership.
Harrison Kim, OpenAI's general manager in Korea, stated in Seoul that cooperation with Samsung Electronics has been strengthened. He noted that the greatest progress and recognition with Samsung Electronics have been achieved in the joint production and research of next-generation chips. However, he did not disclose technical specifications or project timelines, and Samsung also declined to comment on client information.
This collaboration continues work that was started earlier. In June, OpenAI introduced its first specialized chip named Jalapeno, developed jointly with Broadcom for AI inference tasks, which will be manufactured at TSMC fabs. Now, Samsung will also participate in the production of a new specialized chip for OpenAI, which is reported to be the next iteration after Jalapeno. In addition to chips, the partners will also focus on developing advanced memory.
OpenAI emphasized that the demand for memory will continue to grow as AI models become more complex and faster. Last year, Samsung and SK Hynix signed intentions to supply memory for OpenAI's Stargate data center project.
The partnership has two main areas. Firstly, hardware: Samsung will assist in the research, production, and supply of chips and memory for OpenAI, including specialized AI processors and high-speed memory necessary for AI. Secondly, software: Samsung is integrating ChatGPT into its business environment. According to Kim, Samsung is undertaking 'one of the largest global deployments of ChatGPT,' where employees in Korea and abroad use it for marketing, research and development, and sales.
In June, Samsung's Device eXperience division, which handles the production of phones, TVs, and home appliances, announced that employees could use ChatGPT, Google Gemini, and Anthropic Claude internally. South Korea is also a major market, holding the highest number of paying ChatGPT users outside the US; the annual growth rate of corporate users in Korean companies reached 28 times.
A key focus for OpenAI is the Stargate project, a large-scale data center designed to run AI models. Samsung and SK Hynix are ready to provide the memory for it. For Samsung, this agreement opens up the possibility of early access to specialized AI chips and integrating ChatGPT into its own products. Imagine Galaxy phones, Bixby, and smart TVs running on OpenAI models optimized for Samsung hardware.
For OpenAI, this means reducing dependence on Nvidia and gaining greater control over costs. Developing proprietary chips is critical for making ChatGPT cheaper to operate. Kim noted that cooperation with chip suppliers from South Korea is a priority for the local OpenAI office. South Korea plays a key role in the AI supply chain thanks to Samsung and SK Hynix, as both companies produce memory indispensable for AI data centers.
The deal between OpenAI and Samsung in the chip sector demonstrates OpenAI's serious intent to create its own hardware to reduce costs, as well as Samsung's aspiration to become a full-fledged AI company. Although it is unknown when the new Samsung chip will be released, the widespread adoption of ChatGPT in Korea and the growing demand for AI make this partnership timely. This could lead to faster responses from ChatGPT and the appearance of new AI features in Samsung devices.