Harlan Sur, an analyst at the US Foreign Investment JPMorgan Chase (Jiangma), and Toshiya Hari, Vice President of NVIDIA Investment Relations and Strategic Finance, participated in the investor meeting, had a very happy view of the next-generation Vera Rubin platform.
Although NVIDIA has significantly increased Blackwell Ultra GPU production in the second quarter, accounting for about 50% of the Blackwell series, the delivery cycle remains "in quarters, not months." Sur believes that NVIDIA's delivery cycle in this wave of AI investment cycle has been extended by more than two years, indicating that the current "demand for GPUs continues to surpass supply".
In addition, NVIDIA revealed a key message at the conference, that is, all six Vera Rubin GPUs have entered the final production precedent and refusing to delay communication. According to Sur, NVIDIA confirmed that the upcoming Vera Rubin platform has not been delayed and will still be launched in the second half of 2026 as planned. At present, all six chips on this platform have been tape-outed on NTD.
As for the Chinese market, Chinese large technology companies such as Alibaba have shown high interest in NVIDIA's upcoming Blackwell architecture and China's special-supplied version of GPUs. It is known that these companies are willing to pay a price of up to two times the H20 to obtain the B30A because its performance is 6 times higher than the previous generation.
This also shows that the Chinese market still has a strong preference for NVIDIA products, as NVIDIA GPUs have better CUDA software and compete with competitors through NVLink interconnection technology.
Compared to B40, B30 belongs to the Blackwell simplified version, the design focus is on expansion and multi-chip clusters, and the inadequate performance of the single-component through dynamic compression technology; B40 is positioned as a higher-level product as an alternative to the banned H20.
Another RTX Pro 6000D system based on B40 chips is highly likely to be popular as it does not use HBM, does not require additional permission for sales in China, and is mainly used for AI recommendations rather than AI training.
JPMorgan’s Big NVIDIA Update: All 6 Next-Gen Vera Rubin AI Chips Enter Final Pre-Production Stages At TSMC Amidst Extremely High AI Demand! Extended reading: NVIDIA invests for the first time! Quantinuum valuation to millions of dollars Price battle begins! Transfer Samsung's HBM4 to profit from space, and only want to enter the NVIDIA supply chain