Nvidia CEO Jensen Huang attended the opening ceremony of Siliconware Precision Industries Co. (SPIL)’s Tan Ke Plant in Taichung, Taiwan, on January 16, 2025. During Nvidia's annual GTC conference, the company unveiled new chips designed for artificial intelligence development, including the Blackwell Ultra chip family and the next-generation Vera Rubin GPU, which is expected to start shipping in 2026.
Nvidia has seen substantial growth in sales, more than sixfold since the launch of OpenAI's ChatGPT in late 2022, driven by demand for its advanced GPUs used in AI model training. The company’s latest offerings are being closely monitored by major cloud service providers such as Microsoft, Google, and Amazon, as they evaluate whether the new chips will provide sufficient performance and efficiency for their data centers.
Huang emphasized that the computational requirements for AI have significantly increased, necessitating annual chip releases rather than the previous biennial schedule. This shift aims to keep pace with the rapidly evolving AI landscape. The conference, hosted in San Jose, California, is expected to attract around 25,000 attendees, highlighting Nvidia's influence in the industry.
Nvidia’s upcoming Vera Rubin system will feature a custom CPU, named Vera, which is projected to be twice as fast as its predecessor, coupled with the Rubin GPU capable of managing up to 50 petaflops. The Blackwell Ultra chips will enhance content generation speed and revenue potential for cloud providers, with Nvidia reporting that these chips could yield up to 50 times more revenue than their previous generation.
Additionally, Nvidia plans to benchmark its new products against China’s DeepSeek model, which has raised competitive concerns. Huang maintains that the advancements in reasoning capabilities within AI models highlight the need for powerful hardware, positioning Nvidia favorably in the market.