Desk
NVIDIA hardware, software, and developer news.
The New York Times found New Matrix on the H200 approved list despite a hollow listed address and an ownership trail pointing to a US-blacklisted private equity firm. It reports a pattern, not a one-off, as a revenue-sharing deal turned restricted chips into a taxable export.
The Times reports MSS enlisted Alice Huang to seek intelligence on US AI chips and potentially influence Nvidia’s China access. Trump was reportedly briefed last year and didn’t act, and three House committees want a closed-door briefing. Influence operations are the chip war waged through relationships.
Black Forest Labs’ FLUX 3 Image Klein is getting a real Blackwell stack — quantization, sparse attention, caching, and multi-GPU — that turns high-res diffusion from a wait into a workflow.
Six pallets left Oakland on August 29 on a truck disguised as a food hauler and never reached Fort Worth. Investigators cite more than $100 million in similar thefts since June, with GPS spoofing and staged collisions. AI silicon now moves like bullion.
NVIDIA is turning RL weight refit into a true DTensor-to-DTensor transfer — and on big open models, that cuts the wasted shuttle of parameters that was starting to own the training step.
A year after open-sourcing Aerial alongside Sionna, NVIDIA says its 5G/6G stack has crossed a million downloads — and researchers are already using it to cut robot latency, boost receivers, and sleep idle base stations.
House China, foreign affairs, and intelligence chairs want Jensen Huang in a closed briefing after reports he was targeted by a Chinese spy operation tied to a major chip customer.
NVIDIA pitches Vera as the agent-era CPU that is both fast and dense per watt — backed by partner numbers from Perplexity, DeepInfra, Daytona, and others.
NVIDIA released four 16-billion-parameter Cosmos 3 Nano world models for multi-view driving, streaming sensors, simulator-to-video and two-armed robots under the OpenMDW 1.1 license. Open weights help the field compare; they don’t replace real-world testing.
PC enthusiasts are piping PS5 footage through capture cards and open-source Veyra so Nvidia DLSS 5 can upscale and frame-gen GTA VI before the PC port arrives.
NVIDIA opens DSX OS — AICR, NVCRE, and NVSentinel — so GPU Kubernetes clusters get reproducible recipes, workload certification, and continuous health loops instead of hope-based handovers.
NVIDIA profile of Chris Alexiuk — YouTube's llm_wizard and self-styled No. 1 Nemotron fan — argues open weights, data, and recipes are the precondition for useful AI, not a nice-to-have.