NVIDIA's Jensen Huang confirms your new Blackwell AI chip is already obsolete

Jensen Huang announced that NVIDIA has completed tape-out procedures for six distinct Rubin architecture processors at Taiwan Semiconductor Manufacturing Company facilities. The chip manufacturer designed these components to encompass central processing units, graphics processing units, NVLink switching hardware, and silicon photonics technology. NVIDIA plans to implement HBM4 memory specifications alongside TSMC's advanced 3-nanometer manufacturing process for the R100 graphics cards. The Rubin platform represents the company's first chiplet-based design approach with a 4x reticle configuration that surpasses current Blackwell architecture specifications. Market introduction will occur between 2026 and 2027 following successful trial production completion at the foundry partner's Taiwan manufacturing centers.