Samsung has reached an exclusive HBM4 supply agreement with OpenAI, with the supply volume potentially accounting for 15%

Samsung Electronics will be the first to exclusively supply its next-generation HBM4 to OpenAI, the world’s largest artificial intelligence (AI) company headquartered in the United States. According to reports, OpenAI plans to integrate Samsung’s HBM4 into its first-generation AI semiconductor, Titan. As Samsung supplies HBM4 to OpenAI in addition to NVIDIA, it is widely believed that the company is solidifying its leading position in the advanced AI chip market.

According to industry sources, Samsung Electronics has agreed to exclusively supply up to 800 million Gb of HBM4 (12-layer stacked products) to OpenAI in the second half of this year. This volume accounts for 7% of Samsung’s total annual HBM production plan (over 11 billion Gb). Based on the production capacity of its flagship HBM4 product (5.5 billion Gb), this implies that approximately 15% of that capacity is allocated to OpenAI. Reportedly, this marks the third-largest supply deal Samsung has secured with OpenAI, following agreements with NVIDIA and AMD.

The HBM4 chips supplied by Samsung will be directly integrated alongside OpenAI’s first AI semiconductor chip, Titan, which is expected to launch this year. Titan is a dedicated AI chip developed by OpenAI in collaboration with Broadcom. According to reports, TSMC is set to begin mass production in the third quarter of this year, with a launch anticipated around the end of the year.

Analysis indicates that the exclusive HBM4 supply agreement between Samsung and OpenAI is significant, as OpenAI is one of the most closely watched AI companies in the global information technology (IT) industry. OpenAI launched ChatGPT in 2022, sparking a boom in generative AI. It plays a central role in the U.S. government’s $500 billion AI infrastructure initiative, the “Stargate Project.”

OpenAI delivers generative AI services by deploying hundreds of thousands of AI semiconductor chips in its data centers. Previously, it relied primarily on NVIDIA’s general-purpose AI chips. However, OpenAI has recently recognized that to implement its inference models more efficiently, it requires custom semiconductor chips rather than NVIDIA’s general-purpose AI chips. This is because inference performance—making judgments based on training data—has become the dominant trend in the AI field.

An industry insider commented, “OpenAI is fully committed to R&D, striving to achieve mass production of its in-house developed Titan chip. Samsung Electronics met the strict requirements for the sixth-generation HBM4 set by OpenAI and successfully secured this supply agreement.”

Furthermore, some observers believe that since OpenAI has chosen Samsung as its first HBM supplier, Samsung’s HBM is highly likely to be integrated into future generations of Titan chips, such as the second and third generations.