OpenAI CEO Sam Altman held meetings with executives from Samsung, SK Group, TSMC, and Foxconn to discuss the production of proprietary AI chips. The discussions focused on supply chains for HBM memory, accelerator manufacturing, and infrastructure development for artificial intelligence projects.
During the visit to South Korea, Altman was accompanied by OpenAI VP Richard Ho. Reports suggest that the company's Jalapeño accelerator may utilize next-generation HBM4 memory from Samsung. Samsung is also considered a potential partner for chip manufacturing and packaging, although no formal agreements have been confirmed.
The meetings in South Korea also involved SK Group chairman Chey Tae-won, with discussions likely covering HBM supply and data center development. These talks align with the ongoing Stargate project, for which OpenAI representatives had previously negotiated semiconductor supply volumes and contract terms with local firms.
Market analysts at TrendForce project that the average price of HBM memory could rise by 121 percent in 2027 compared to 2026. This increase is attributed to limited supply and the growing market share of advanced, more expensive HBM4 components.
In Taiwan, Altman met with Foxconn chairman Young Liu, TSMC chairman C.C. Wei, and representatives from Quanta and Wistron. OpenAI is accelerating its chip development, having introduced the Jalapeño inference accelerator in June 2026, which was developed in partnership with Broadcom.
The first generation of OpenAI chips utilizes TSMC's 3-nanometer process, while the second generation may adopt the A16 technology. OpenAI aims to scale its infrastructure to the gigawatt level, with initial deployments of its proprietary hardware scheduled for completion by the end of 2026.
