#Hardware
Curated coverage on «Hardware»: releases, research, benchmarks, open-source.
US eases restrictions: UAE to gain access to AI chips
The Trump administration has authorized shipments of advanced semiconductors and satellites to the United Arab Emirates, opening the market for high-tech components…
SambaNova raises $1B at $11B valuation, rejecting Intel offer
AI chipmaker SambaNova closed a $1B Series F round at an $11B valuation, rejecting Intel's lower $1.6B acquisition offer.
Meta to begin production of its own AI chips in September 2026
Meta announced the launch of production of its own AI chips in September 2026, using a modular architecture for rapid adaptation to the development of artificial…
Former OpenAI Product Director Kevin Weil Joins Stoke Space Board
NVIDIA opens Nemotron datasets: 10 trillion tokens for training AI agents
NVIDIA Nemotron 3 Ultra sets an accuracy record among open models in LangChain test
The US closed access to Claude Mythos, China restricts GLM and Qwen: neural networks have become a strategic weapon
habrGPT: training a 0.5B LLM from scratch on Habr articles — Karpathy’s nanochat and an fp8 vs bf16 comparison
An experiment in training a 500-million-parameter language model from scratch on the Habr corpus: Karpathy’s nanochat, a home GPU instead of an 8×H100 cluster, and…
Meta contractor discharged bacteria-contaminated water into Wyoming's sewer system while building an AI data center
A Meta construction contractor discharged bacteria-contaminated wastewater into Cheyenne's municipal sewer system — the incident forced Wyoming authorities to…
Nvidia pairs GPUs with startup d-Matrix's inference chips instead of competing with it
NVIDIA released Audex-30B — a unified audio-text MoE model based on Nemotron-Cascade-2
Hugging Face: transformers backend in vLLM now runs at native speed
Microsoft bets on in-house AI models to cut costs
Perplexity among the first AI companies to adopt Nvidia's new Vera processor
Perplexity AI said it plans to run agent workloads on Nvidia's new Vera CPU — the company's first in-house processor, opening a path beyond GPU dominance.
Neural networks are bad calculators: why LLMs fail at arithmetic and what to do about it
Language models predict tokens rather than calculate, which is why they make confident mistakes in the middle of long numbers. We break down the mechanism and…