Samsung to supply OpenAI's first AI processor with cutting-edge memory chips
Samsung Electronics is poised to become a key supplier of advanced memory chips for OpenAI's first in-house AI processor. The deal, expected to start as early as 2026, would mark a significant step in the tech giant's push to dominate the high-performance memory market. OpenAI's move to develop its own AI chips also highlights the company's efforts to reduce reliance on external suppliers.
OpenAI has been working on its own AI processor since last year in partnership with Broadcom. Production of the chip is set to begin at TSMC in the third quarter of 2024. The company's push into chip development comes as its demand for computational power continues to grow.
Samsung's potential role in supplying HBM4 memory for this processor follows earlier collaboration between the two firms. The South Korean manufacturer had previously signed a letter of intent to provide memory chips for OpenAI's data centre infrastructure. If confirmed, the new deal would strengthen Samsung's position in the AI infrastructure market, where demand for high-speed memory solutions is rising.
Beyond OpenAI, Samsung has also expanded its partnership with AMD. A Memorandum of Understanding, signed on March 18, 2026, secures Samsung's supply of HBM4 memory for AMD's upcoming Instinct MI455X AI accelerators. Each GPU in this series will integrate 432 GB of HBM4, though no deliveries have been made yet as development remains ongoing.
Samsung's involvement in both OpenAI's and AMD's AI supply chains underscores its ambition to lead the HBM4 market. The company's deals with two major players in AI infrastructure signal a strategic push to meet surging demand for high-performance memory. OpenAI, meanwhile, continues advancing its chip strategy to better control its AI hardware needs.
