NVIDIA news — AI chips, GPUs, platforms and the company powering the AI boom, from data centres to RTX Spark laptops.
Berlin will put a three-digit million-euro sum into industrial AI with Nvidia, renting compute on Deutsche Telekom's Munich cloud. Siemens and Audi join.
Modal will use Nvidia's Vera CPU for hosting AI agents after tests doubled agent work, Nvidia's first CPU win at a cloud built on Intel and AMD.
NVIDIA released Nemotron 3.5 Super VL, a 120B open model for always-on agents, plus Bolt edge models that run local agent work in 24GB and 4GB of memory.
Santander AI lab evaluates open-weight models in a Model Arena and specialises them with NVIDIA NeMo Customizer, cutting some AI serving costs by over 50%.
Samsung will use TSMC's 3nm process for base dies in Nvidia's custom NVHBM memory, Reuters and Seoul Economic Daily reported on October 18, 2026.
NVIDIA makes money from AI mainly through data centre GPUs and systems, plus networking, software, gaming, and automotive and robotics platforms worldwide.
Blackwell is NVIDIA's data centre GPU generation after Hopper, designed for larger AI models, faster inference, and efficient rack-scale computing worldwide.
NVIDIA powers the AI boom through GPUs built for parallel computing, the CUDA software platform, and complete data centre systems sold worldwide today.
NVIDIA's RTX Spark laptops are on sale from $2,599, running AI agents and a 74B model on the device with a Grace and Blackwell superchip.