Login to Continue Learning
JPMorgan analyst Harlan Sur attended an investor group meeting with Toshiya Hari, NVIDIA’s Vice President of Investor Relations and Strategic Finance. At the meeting, Sur expressed a positive outlook on the demand for NVIDIA’s current-generation GPUs and the production timeline for its next-gen Vera Rubin platform.
Sur noted that the lead times for NVIDIA’s Blackwell Ultra GPUs remain “quarters, not months,” despite a significant ramp-up in production during fiscal Q2, which now accounts for about 50 percent of the Blackwell mix. This suggests that demand is still outstripping supply more than two years into this AI spending cycle.
At the meeting, NVIDIA confirmed that its upcoming Vera Rubin platform has not experienced any delays and is on track for a C2H26 launch, with all six chips already taped out at TSMC.
Reuters reported that China’s largest tech firms, including ByteDance and Alibaba, are excited about NVIDIA’s Blackwell-based GPU. These companies are willing to pay double the price of the H20 GPU for the B30A, which is said to offer a 6x performance boost over its predecessor. This aligns with US broker research indicating a strong preference for NVIDIA’s products in China.
The success of NVIDIA’s GPUs is largely due to superior software support via the CUDA ecosystem and better performance in clusters thanks to NVLink interconnects.
NVIDIA’s RTX Pro 6000D systems, based on the B40 chip, do not require a separate license for sale in China as they do not use high-bandwidth memory (HBM) and are primarily used for inference rather than training foundational AI models. It is likely that these chips will sell well once available to Chinese companies.
The B30 is a more compact version of the Blackwell GPU, while the B40 offers a higher-end alternative to the banned H20 chip. The B30 focuses on scaling and multi-chip clusters with dynamic compression to offset lower performance per chip, whereas the B40 aims to provide a direct replacement for the H20.
