The graphics processor manufacturer has commenced fabrication of its upcoming Rubin architecture while obtaining high-bandwidth memory samples from multiple vendors, according to reports from Taiwan. Chief executive Jensen Huang confirmed during a recent visit that production lines have begun processing the advanced artificial intelligence accelerators, marking rapid progress from initial laboratory prototypes received days earlier.
Manufacturing partner TSMC plans to increase three-nanometer node output by half to accommodate orders spanning current Blackwell generation products and forthcoming Rubin chips. Company president C.C. Wei declined to specify wafer quantities requested but acknowledged substantial volume requirements. The semiconductor firm has secured fourth-generation high-bandwidth memory prototypes from various suppliers ahead of anticipated mass manufacturing in the third quarter of 2026.
OpenAI has already committed to a partnership valued at 100 billion dollars that will deploy the next-generation accelerators across its computing infrastructure when production reaches commercial scale.
Manufacturing partner TSMC plans to increase three-nanometer node output by half to accommodate orders spanning current Blackwell generation products and forthcoming Rubin chips. Company president C.C. Wei declined to specify wafer quantities requested but acknowledged substantial volume requirements. The semiconductor firm has secured fourth-generation high-bandwidth memory prototypes from various suppliers ahead of anticipated mass manufacturing in the third quarter of 2026.
OpenAI has already committed to a partnership valued at 100 billion dollars that will deploy the next-generation accelerators across its computing infrastructure when production reaches commercial scale.
