Jensen Huang Unveils NVIDIA’s Post-Blackwell Roadmap: Annual Chip Updates
In October last year, NVIDIA introduced updates in their investor briefing, including HBM3e, PCI Express standards (6.0/7.0), and advancements in multi-GPU interconnect technology. They also unveiled a comprehensive product roadmap detailing their data center plans for 2024 to 2025. Currently, NVIDIA updates its primary GPGPU architecture every two years, and the roadmap indicates an update in 2025.
According to Wccftech, NVIDIA recently released its financial report for the first quarter of fiscal year 2025, revealing explosive revenue growth in its data center business, far surpassing its previously dominant gaming segment. Simultaneously, NVIDIA’s founder and CEO, Jensen Huang, confirmed that following the Blackwell architecture, another chip is on an annual update cycle, with new GPUs, CPUs, network NICs, switches, and various other chips forthcoming.
As Blackwell architecture products gradually hit the market, rumors about the subsequent new architecture, codenamed “Rubin,” have emerged. Named after American astronomer Vera Rubin, this next-generation product focuses on significantly reducing power consumption, a crucial improvement given that current architecture products are nearing the kilowatt range and cannot scale indefinitely. The first GPU based on the Rubin architecture, the R100, is expected to enter mass production in the fourth quarter of 2025.
Despite NVIDIA’s substantial investment in the development of the GeForce RTX 30/40 series GPUs, which have shown commendable performance improvements, their actual revenue contribution has not seen significant growth. With the surging demand for artificial intelligence (AI) hardware and the accelerated update frequency of data center GPU architectures, NVIDIA may integrate some technologies with AI PCs, potentially benefiting RTX GPUs.