Desk
NVIDIA hardware, software, and developer news.
Nvidia's European cloud update is mostly a capacity roll call, but the agent-security piece and a power-efficiency result are the parts worth taking seriously.
NVIDIA's Berlin roll call of European manufacturers is a sales pitch, but the claimed gains in simulation and inspection are the useful kind of AI worth tracking.
The least flashy NVIDIA news from Berlin may have the longest tail: a GPU library for the sparse math that throttles chip, aircraft and materials simulations is now generally available.
NVIDIA wants frontier labs to rent GPUs anywhere without trusting the landlord, and Confidential Bare Metal is its bid to make that possible at rack scale.
NVIDIA wants to own the layer where enterprise agents meet company data, and the open-source Agent Knowledge Foundation is its bid to make that layer feel neutral.
Nvidia keeps widening its open-model lineup, from business forecasting to protein design to quantum calibration. The strategy is generous on weights and very clear about whose hardware runs them.
NVIDIA wants to move agent security off the server and into a separate chip the agent cannot touch. That is the right instinct, even if it arrives next year.
NVIDIA's pitch for agent safety starts from a blunt premise we agree with: an agent cannot police itself. The catch is that the fullest version of the fix runs on NVIDIA silicon.
NVIDIA's walkthrough of agents helping design AI data centers is most useful for what it admits: the agent drafts, and engineers still have to check every dimension.
NVIDIA wants AI data centers to arrive like manufactured goods, built and tested in a factory and plugged in on site. It is a sensible fix for a real bottleneck.
Jensen Huang's Berlin keynote sold Europe a full AI factory, from containerized data centers to sandboxed agents, and the loudest line was that agent security is an engineering job.
With TorchTPU, Google joins Amazon in going after Nvidia's quietest advantage, which was never just the chip but the default assumptions baked into PyTorch.