Major Nvidia announcement: New chips will "surprise the world"

Nvidia CEO Jensen Huang has announced a major chip unveiling at the GPU Technology Conference – GTC – which will take place in San José from March 16 to 19, 2026. In an interview with the Korea Economic Daily, he spoke of processors that “will surprise the world.” While Huang did not provide specific details about the new hardware, it is clear that the focus will be on artificial intelligence rather than gaming or traditional consumer graphics cards.

On Reddit, speculation about Nvidia’s major reveal is already in full swing. Many users see the upcoming Vera Rubin generation as the most likely candidate. Shortly before his interview with the Korea Economic Daily, Jensen Huang met with South Korean memory manufacturer SK Hynix. He described the meeting as a “celebratory dinner with the world’s leading memory semiconductor team.” SK Hynix is one of the leading producers of high bandwidth memory – HBM – and is currently advancing the development of HBM4. Extremely high memory bandwidth and low latency are essential for AI accelerators used in data centers, making advanced HBM technology a key factor in next-generation AI hardware.

The Rubin generation is widely regarded as Nvidia’s next major architecture for this segment. Industry analysts expect Rubin to rely on HBM4 for the first time at scale, helping to eliminate bandwidth bottlenecks and improve the efficiency of large AI models. Yet HBM4 alone is not the decisive factor. What matters even more is advanced packaging and system-level integration. The real challenge often lies not in the memory chip itself but in connecting it efficiently to the processor. If Nvidia introduces new integration methods in this area, the technological impact could be just as significant as the transition to HBM4.

The timing of Huang’s interview, his meeting with SK Hynix, and the announcement of a major chip reveal makes Rubin combined with HBM4 the most plausible scenario. Some observers have also pointed to potential Rubin derivatives, such as dedicated inference variants. Others speculate about an early preview of the more distant Feynman architecture, although that is widely considered unlikely. For consumers, one point is crucial: even if Nvidia unveils the Rubin generation, it will almost certainly be aimed at data centers and large-scale AI systems rather than new gaming graphics cards. All indications suggest that Nvidia will continue to prioritize AI infrastructure.