Samsung Electronics is preparing to supply its next-generation high-bandwidth memory, HBM4, to OpenAI for use with the company's first in-house artificial intelligence processor, the Korean Economic Daily reported on Thursday, citing unnamed industry sources.
The report said Samsung plans to provide up to 800 million gigabits (Gb) of 12-layer HBM4 chips to OpenAI in the second half of this year. Those memory modules are intended to work alongside OpenAI’s initial custom AI processor, which the report says was developed in partnership with Broadcom.
According to the story, the custom processor is expected to be manufactured by Taiwan Semiconductor Manufacturing Co starting in the third quarter, with a product launch aimed by year-end. The newspaper further recounted that last year Samsung signed a letter of intent to supply memory chips for OpenAI’s data center to help meet demand tied to its Stargate project.
OpenAI’s move to build in-house processors represents a continuing effort by the company to secure the computing capacity required for its expanding services. The report noted that last year OpenAI partnered with Broadcom to produce those processors, marking another chip collaboration for the ChatGPT maker.
Separately, the report highlighted a recent alignment between Samsung and Advanced Micro Devices. On Wednesday the two companies signed a memorandum of understanding to broaden an existing strategic partnership for memory supplies to AI infrastructure. Under that agreement, Samsung is slated to act as a key supplier of HBM4 chips for AMD’s upcoming AI GPUs.
Samsung Electronics declined to comment on the Korean Economic Daily’s report, and OpenAI was not immediately available for comment outside regular business hours, the story added.
Key points
- Samsung intends to supply up to 800 million Gb of 12-layer HBM4 memory to OpenAI in the second half of this year, according to the report.
- The HBM4 memory is planned to be paired with OpenAI’s first custom AI processor developed with Broadcom, expected to be manufactured by TSMC beginning in the third quarter and targeted for a year-end launch.
- Samsung and AMD signed a memorandum of understanding to expand HBM4 supply for AMD’s upcoming AI GPUs, underscoring demand across AI infrastructure and semiconductor supply chains.
Risks and uncertainties
- Plans reported by the newspaper rely on unnamed industry sources and have not been confirmed by Samsung or OpenAI, so the supply and timing details remain subject to change - this affects the semiconductor and AI hardware sectors.
- Production and launch timelines depend on third-party manufacturing schedules, with the report noting TSMC is expected to start manufacturing in the third quarter and a launch is targeted by year-end - schedule slippage could alter market expectations for AI hardware availability.