NVIDIA (NVDA, Financial) is reportedly expediting the launch of its next-generation AI GPU architecture, Rubin, by six months to the latter half of 2025. This move comes amid fierce competition from AMD, Amazon, and Broadcom in the AI chip market. Despite the current Blackwell architecture facing thermal issues and yet to be mass-produced, NVIDIA aims to solidify its dominance in the data center AI chip sector, where it holds an 80%-90% market share.
Industry experts anticipate the Rubin architecture, leveraging TSMC's 3nm process, CPO technology, and HBM4, will deliver unprecedented performance. This advancement may redefine AI computing power, setting a new benchmark that competitors could take years to match. The demand from AI giants like OpenAI and Meta is driving NVIDIA to accelerate its innovation pace.
SK Hynix's potential early production of HBM4 supports the likelihood of Rubin's early release. Rubin's CPO packaging is expected to significantly enhance data transmission efficiency and energy consumption, further cementing NVIDIA's leadership in AI chips.
