#Nvidia announced the #RubinCPX, a specialised accelerator optimised for the prefill phase of inference, emphasising compute FLOPS over memory bandwidth. This development, alongside the Oberon rack-scale form factor, widens the gap between Nvidia and its competitors, forcing them to invest in prefill chips and delaying their ability to catch up. https://semianalysis.com/2025/09/10/another-giant-leap-the-rubin-cpx-specialized-accelerator-rack/?eicker.news #tech #media #news