OpenAI says the companies are working on chip production and research. Samsung has not confirmed its exact role, so the announcement is a direction, not a finished product.
New to this? Read it in simple words
- OpenAI says it is working with Samsung on future AI chips. Harrison Kim, head of OpenAI Korea, said the work is making progress.
- OpenAI is already developing a custom inference chip with Broadcom. Reuters says TSMC is expected to make that chip.
- The news shows OpenAI wants more control over the hardware that runs its models. It also shows Samsung trying to win more AI work.
- Samsung said it could not discuss a possible customer. The companies have not explained Samsung’s exact role.
- Inference
- Using a trained AI model to answer questions or do tasks.
- Custom chip
- A chip designed for one company’s own needs instead of for many buyers.
OpenAI confirmed work, not a finished chip
Harrison Kim, the head of OpenAI Korea, said the companies are making progress on joint chip research and production. Reuters reported his comments from Seoul. Samsung said it could not discuss information about a possible customer.
The statement does not say which part Samsung may make. A modern AI chip needs a design, a factory process, memory, packaging, and testing. Different companies can handle each part.
OpenAI is already developing a custom inference chip with Broadcom. Inference means using a trained model to answer a request. Reuters says Taiwan Semiconductor Manufacturing Company is expected to make that chip.
A future AI chip needs design, production, memory, packaging, and real tests. The partners have not said which steps Samsung will handle.
Samsung already has a wider OpenAI link
OpenAI announced a larger Samsung agreement in 2025 for its Stargate data centres. Samsung Electronics and SK Hynix planned to supply advanced memory. Samsung companies also discussed data centres, cloud services, and floating power systems.
Memory matters because an AI processor must move large amounts of data very quickly. A fast main chip can still wait if its memory is too slow. Samsung is one of the few companies with skills across memory, chip factories, and packaging.
The new comments may mean that this relationship is moving beyond supply. They do not prove that Samsung will replace TSMC or make OpenAI’s first chip. That would be a larger claim than the public facts support.
Watch for the missing details
The next useful facts are the chip’s purpose, the factory, the memory type, and the planned production date. Power use and real model speed will matter more than a broad promise about “next-generation” technology.
Custom chips can reduce cost when a company uses them at very large scale. They can also take years to design and may fall behind a fast market. OpenAI still needs other suppliers during that work.
The announcement shows that OpenAI wants more control over the hardware below its models. It also shows Samsung trying to win more AI work. Readers should treat it as a direction until the companies share a product and measured results.
Sources
Every fact in this story comes from the sources below. Open them to check our work.
- 1Research · September 9, 2026OpenAI says working with Samsung on next-generation chips, deepening cooperation Reuters via Investing.com
- 2Research · September 9, 2026OpenAI says working with Samsung on next-generation chips, deepening cooperation Reuters via MarketScreener
- 3
We used two licensed Reuters pages for the September 9 comments and Samsung’s limited response. We used OpenAI’s earlier Stargate announcement for supply-chain context. We do not guess which chip Samsung will make or whether it will replace another supplier.