NVIDIA is gearing up for a new era of AI servers, even as it begins low-volume production of its Blackwell Ultra model. This signals a strong assurance that no competitor can rival NVIDIA’s pace in the AI realm.
The company is finalizing its Rubin Design, with plans for mass production next year. NVIDIA’s rapid advancement through the AI excitement shows a staggering pace few can emulate. Currently on a six to eight-month product cycle, NVIDIA is not just producing ordinary GPUs but massive AI clusters valued at billions, marking a relentless and impressive expansion.
According to a report by the Taiwan Economic Daily, NVIDIA is preparing the design for its Vera Rubin server racks. These are expected to be introduced to mainstream suppliers by the end of this month, setting the stage for production.
Rubin architecture is anticipated as a monumental step forward in computing. Major changes from the ground up, including advancements in HBM, process nodes, and design, are expected. Similar to the pivotal Hopper generation and the leap of Ampere AI accelerators, Vera Rubin AI server racks are projected to hit the market between 2026 and 2027. This will ensure NVIDIA stays at the forefront of the AI surge.
For a glimpse into what Vera Rubin promises, NVIDIA will employ next-generation HBM4 chips to power its R100 GPUs. These represent a significant upgrade from the current HBM3E standard. Additionally, Rubin will utilize TSMC’s 3nm (N3P) process and CoWoS-L packaging, aligning with new industry standards designed to boost performance. Notably, Rubin will incorporate a chiplet design, a first for NVIDIA, along with a 4x reticle design, an advancement from Blackwell’s 3.3x.
While Rubin exemplifies ambition, questions loom about NVIDIA’s ability to launch new architectures in such short timelines, challenging supply chains to adapt quickly. Similar scenarios unfolded with the GB300 AI platform, which ultimately relied on the older Bianca board featuring the GB200 platform. The unfolding situation with Rubin is one to watch, but what’s undeniable is NVIDIA’s unparalleled speed, often dubbed “Jensen speed.”
