NVIDIA
NVIDIA designs the GPUs that train and run virtually every major AI model, from the H100 to the Blackwell generation. Its CUDA software stack and data-center chips made it one of the most valuable companies on Earth and the bottleneck of the AI boom. All our NVIDIA coverage: chips, earnings and ecosystem news.
Avito Built Aviflow: An ML Platform for Distributed LLM Training on HGX
An engineer on Avito's ML platform explains how the team moved from DevBox machines with SSH access to Aviflow, a cloud-native platform for distributed LLM training…
I Ran Qwen on a $50 FPGA: How a Homemade AI Accelerator Works
An engineer ran the Qwen LLM on a Xilinx Kintex-7 FPGA with 4GB of memory — no GPU required — using openTPU. Qwen3-0.6B hits 31 tokens per second. Instead of…
NVIDIA Physis-Lang Describes Physics in Words and Takes First Place on Physics-IQ Verified
NVIDIA's Physis-Lang uses text descriptions of physical processes, such as melting butter, to help a video model reproduce physics more accurately. NVIDIA says it…
Principal Global Investors Strategist Weighs In on the AI Capex Cycle and Market Sentiment
NVIDIA Scales Bit-for-Bit Determinism in Megatron Core to 2,432 GPUs with 2% Overhead
Trump Hands Out Awards to Tech Billionaires a Month Before Elections Instead of Helping Ordinary Americans
NVIDIA Outlines 5 Steps for Preparing SimReady Assets for Robots Using AI Agents
Bloomberg: AI Boom Has Pushed Global Trade Imbalances to Their Widest Since 2008
Bloomberg Originals links the concentration of capital around AI to trade imbalances the world hasn't seen since the 2008 crisis. We break down the mechanics and…
NVIDIA Dynamo: Session-Based Infrastructure Management for Agentic Inference
NVIDIA Dynamo turns request-aware infrastructure into program-aware infrastructure using a session-level identifier. This lets it optimize KV cache, routing, and…
How Humanoids Are Moving from Prototypes to Production: Jabil's View
Black Forest Labs Releases FLUX 3 Action, an Open 7B Model for Robots
MSI Prestige N16 Flip AI+ Heads to Russian Pre-Order: The First RTX Spark Convertible
miniTE speeds up RL in NVFP4 on B300: 1.39× over BF16 on Qwen3-30B-A3B
Britain invests billions in supercomputer for independence from the USA
The British government is launching a project to create a supercomputer worth over one billion pounds to help local startups develop their own semiconductors and…
SpaceX leases entire Colossus 1 data center: engineers unable to handle operations
SpaceX failed to effectively operate its Colossus 1 data center in Tennessee and decided to fully lease it out — despite acute shortage of computational power in…