Login to Continue Learning
NVIDIA’s CEO, Jensen Huang, has confirmed that his company is now working on the next-generation Rubin AI architecture. He describes this as a ‘revolution’ for compute markets.
### NVIDIA Has Taped-Out Six Different Rubin Chips For TSMC, Suggesting a Revamp of the Whole Tech Stack
Jensen Huang recently visited Taiwan to focus on Rubin’s progress at TSMC. During his visit, he spoke with local media and revealed that NVIDIA has already taped out six different Rubin chips, including new CPUs and GPUs. These chips are now in TSMC’s fabrication facilities for trial production.
My main purpose is to visit TSMC as we have a next generation architecture called Rubin, which is very advanced. We have now taped out six brand new chips to TSMC, so all of these chips are now in TSMC’s fabs.
The new chips include a dedicated CPU, GPU, scale-up NVLink switch, and a new silicon photonics processor. This means that the entire tech stack will undergo a massive upgrade. NVIDIA’s Rubin architecture is expected to be the next leap in computing capabilities since it involves fundamental changes from the ground up—starting with HBM, process node, design, and more.
### The Upcoming Rubin Architecture
NVIDIA plans to use next-generation HBM4 chips to power its R100 GPUs, which are a significant upgrade from today’s HBM3E standard. They will also adopt TSMC’s 3nm (N3P) process and CoWoS-L packaging.
More importantly, Rubin will utilize a chiplet design, a first-of-a-kind NVIDIA implementation, and a 4x reticle design (versus 3.3x of Blackwell). This represents a substantial generational leap similar to what the Hopper architecture did.
### Timeline for Rubin
The debut of Rubin in the market is expected between 2026 and 2027, depending on when trial production is completed. However, the anticipation around NVIDIA’s Rubin makes it clear that this will be a major launch by Team Green.
