OpenAI & Samsung chip push revealed: Why the company behind ChatGPT is betting on hardware to fuel its next AI growth

For a user, it's easy, they open ChatGPT, ask a question and wait for an answer. The hard work happens elsewhere, almost hidden. That vision is now changing as OpenAI announces it is working with Samsung Electronics on the next generation of chips it is developing.

The announcement is significant due to its focus on hardware, which is now a central part of OpenAI's story. Harrison Kim, General Manager of OpenAI Korea, said that the country was one of the areas where OpenAI had achieved the most progress and received recognition through joint production and research. He made the statements at a press conference in Seoul.


OpenAI's combined actions with Samsung are not happening in isolation. In June, the company unveiled its first custom artificial intelligence (AI) chip, which was designed with Broadcom. The chip, called Jalapeno, is designed for AI inference, which is the act of crunching data to produce answers to requests in applications such as ChatGPT.

One may think of inference as the moment when they actually request that the system do something. You type a question, request a summary or ask for help with a task. The system then has to process that request and return an answer. Jalapeno was designed to focus on that type of work, and OpenAI said at the time that Taiwan's TSMC was going to manufacture the chips.

The Samsung cooperation fits into that hardware picture. Rather than relying solely on existing technology and suppliers, OpenAI is now talking about chips it is developing itself, and working with a major technology company on their production and research. The details of that Samsung work have not been shared yet.


Samsung Electronics is already deeply involved in the semiconductor business, and OpenAI has another reason to work closely with South Korean chip suppliers. Last year, Samsung Electronics and SK Hynix signed letters of intent to supply memory chips for OpenAI's Stargate data centre project.

Memory may seem like a technical detail, but it is central to the larger computing picture. OpenAI's Kim said demand for memory chips would continue to rise due to the need for more advanced chips to allow faster and more complex computing. For OpenAI's Korea operation, that makes cooperation with South Korean semiconductor suppliers a priority.

Let's take an example of a restaurant becoming much busier every month. Hiring more chefs is useful, but it does not solve everything if the kitchen cannot handle the orders. You may need better ovens, more preparation space and a stronger supply of ingredients. In the AI world, chips and memory are part of that kitchen, and if demand grows, the hardware underneath the service has to grow with it.

There is another side to the Samsung relationship, and it has less to do with chips and more to do with people. Kim said Samsung Electronics helped with one of the largest-scale deployments of ChatGPT globally. Employees in Korea and overseas are using ChatGPT for work in fields such as research and development, marketing and sales.


AI is no longer being described purely as a tool that an individual worker opens up occasionally. OpenAI said the number of ChatGPT Enterprise users at South Korean companies and institutions had increased about 28-fold by the end of August compared with a year earlier.

Read More: OpenAI GPT-6 Astra arrives with eye-popping scores: What its 1.05-million-token context window means for the future of AI work

The relationship is going both ways. Samsung said in June that workers in its Device eXperience division (which includes mobile phones, televisions, and home appliances) would be permitted to use ChatGPT, Google's Gemini and Anthropic's Claude internally.

More business users means more AI activity. More AI activity means more computing. And more computing makes the underlying hardware increasingly important.

In the end, it can be stated that better hardware can allow AI systems to more effectively handle computing tasks. For a service such as ChatGPT, that matters because the system has to respond to users while handling increasingly complex workloads.