NVIDIA's next-gen Vera Rubin AI racks are expected to ship to customers by Q3 of this year, but initial production levels aren't likely to be "too high".
Suppliers Face No Issue Ramping Up Vera Rubin Production For Now, Given How NVIDIA Has Retained Several Elements From Blackwell
Team Green surprised us by announcing that Vera Rubin has entered "full production" in Q1 2026, which was seen as a massive achievement, given that NVIDIA delivered earlier than its product cadence. However, NVIDIA didn't reveal when hyperscalers would receive the first batch of the new AI racks, but according to Quanta's Executive VP Mike Yang (via Ctee), initial Rubin units could be delivered to customers by August, ensuring complete hyperscaler integration by the last quarter of this year.
Given that the Rubin architecture has shifted from chiplet design to advanced packaging and HBM integration, there were concerns about whether NVIDIA could deliver on its promised "H2 2026" timeline. However, Yang has revealed that several components within NVIDIA's Rubin infrastructure are the same as those in the Blackwell series, which means transitioning production lines doesn't pose as big a risk for now. NVIDIA is expected to follow a similar trend to what we saw with Blackwell Ultra, limiting initial shipments and then ramping up volume in Q1 of next year.
For those unaware, the Rubin lineup was unveiled extensively at this year's CES, with a key focus on "six different" chips involving networking and compute power. Here's what these chips look like:
- Rubin GPU (with 336 Billion Transistors)
- Vera CPU (with 227 Billion Transistors)
- NVLINK 6 Switch for Interconnect
- CX9 & BF4 for Networking
- Spectrum-X 102.4T CPO for silicon photonics
It is anticipated that Rubin shipments will be limited to the NVL72 configuration for now, despite NVIDIA revealing an NVL144 configuration at last year's GTC. By keeping chip count in check, the company aims to control thermal draw and ease the production process. It appears that Team Green has learned that the supply chain does take time to adjust to the company's architectural changes, as we saw significant complications when switching from Blackwell to Blackwell Ultra due to the use of Cordeila board design and similar changes.
NVIDIA's Rubin AI series is expected to be fully adopted by hyperscalers within the Q4 2026 to Q1 2027 timeline, and frontier models will benefit from the significant performance coming on board soon. Early customers include OpenAI, Microsoft, Google, and neoclouds like CoreWeave.
About the author: Muhammad Zuhair is a hardware and technology reporter for Wccftech, specializing in the semiconductor industry and the complex interplay between technology, manufacturing, and geopolitics. His coverage focuses on the corporate strategies and technological roadmaps of industry giants like TSMC, NVIDIA, Samsung, and Intel. Zuhair's expertise lies in deconstructing complex topics such as fabrication nodes (e.g., 2nm process), the economic impact of policies like the CHIPS Act, and the strategic development of AI infrastructure from NVIDIA, AMD and Intel.
Follow Wccftech on Google to get more of our news coverage in your feeds.
