Updated daily77 verified stories
AI chips, accelerators, GPUs, foundries, data centers, export controls, and the infrastructure behind model training and inference.
Every story links to the strongest available evidence: original sources when available, otherwise clearly attributed reporting.
What happened, why it matters, and what to watch — no jargon tax.
When the signal is thin, we publish nothing rather than padding the feed.
A growing stream of verified perspectives for people who need to understand AI without chasing hype.
Nvidia denied a Wall Street Journal report that it had paused some transactions under its new AI Compute Partnership, while Tom’s Hardware reported that cloud providers objected to Nvidia’s alleged limits on which customers could rent its GPUs.
TIME reports that Nvidia has agreed to buy Hugging Face, an online hub for open AI models and datasets, for $12.9 billion. The report says the deal could help Nvidia hedge against major AI companies developing their own chips, though neither company confirmed it.
Engadget reports that a partly built New Jersey AI data center tied to Microsoft was operating dozens of gas-powered generators without permits, though regulators have not completed a compliance determination.
CRN reports that Nutanix plans to support AMD GPUs in its enterprise AI inference stack by the end of the calendar year, following AMD’s $250 million investment in the company. CEO Rajiv Ramaswami described the move as a choice for customers seeking alternatives to Nvidia and more predictable costs for running AI…
AWS says Salesforce used a new SageMaker Inference Components placement capability to balance Agentforce model copies across Availability Zones while preserving GPU-sharing savings.
Crypto Briefing reports that Nvidia is accelerating releases of its open-weight Nemotron AI models from every six to eight months to roughly every four to six weeks, based on an interview with Nvidia executive Bryan Catanzaro.
Deepgram says new CloudWatch, Prometheus and OpenTelemetry metrics give customers more visibility into speech-AI usage, AWS Marketplace billing, engine capacity and per-GPU utilization when running its models on Amazon SageMaker AI.
Databricks says distributed, asynchronous checkpointing and cached data loading can reduce recovery time and GPU idle time during large-scale PyTorch training, while warning that incomplete data-pipeline checkpoints can silently distort resumed jobs.
A new benchmark evaluates whether language models can generate and improve GPU code for database-style queries, with the paper reporting speedups of up to 2.11× on one H100 GPU and 2.54× across four H100 GPUs.
A new preprint reports that BanglaMamba, a state space model for Bangla fake-news detection, uses less GPU memory and processes data faster than tested BERT-based systems, while trailing BanglaBERT on accuracy and external-dataset performance.
Nvidia told investors it has $18 billion committed to equity investments for the rest of its fiscal year, Business Insider reports. The company’s private-company holdings reached $47.9 billion in late July, more than double the prior fiscal year-end figure.
AWS and NVIDIA say they plan to deploy 2 million additional NVIDIA GPUs across AWS infrastructure in 2027 and 2028 while expanding cooperation on CPUs, networking, open models, data processing and robotics.
One useful briefing each week
Get the week’s verified AI news, original data, useful tools, learning picks, and fresh AI jobs.
Hiring an AI professional or launching a useful AI product? Put it in front of people who came here to learn and act.
Post an AI job Submit an AI tool