AI Chips news and analysis on Web Pulse: 6915 curated articles tracking the latest AI Chips developments, tools, and research, updated continuously from vetted sources.
NVIDIA added the RTX PRO 5500 "Blackwell" workstation GPU to its pro-viz lineup, featuring 84 GB of GDDR7 memory on a 416-bit bus delivering about 1,398 GB/s of bandwidth. The card uses a GB202 die with 21,760 CUDA cores…
ChemSec warns that PFAS producers led by Chemours are planning a "tidal wave" of production expansion to supply the AI boom's semiconductor and data center cooling needs, with Chemours adding capacity in Fayetteville, No…
A developer explains the difference between CUDA cores and Tensor cores for machine learning workloads, noting that Tensor cores are specialized for matrix multiplication and accumulation in mixed precision and can deliv…
Nvidia signed deals to build AI factories in Australia, according to Stockhead. The report states the world's most valuable company chose Australia for the AI infrastructure push, while noting concerns about the national…
A Chinese state-run newspaper accused Anthropic CEO Dario Amodei of invoking a "Cold War playbook" after Amodei called for the US to impose stronger controls on exports of AI-related chips to China and to take action aga…
An analysis of AI assistant infrastructure argues that differences in training and serving infrastructure between OpenAI's ChatGPT, Anthropic's Claude, and Google's Gemini matter more to small businesses than model size …
India's Semiconductor Mission, backed by ₹76,000 crore (~$10 billion) in government incentives, is turning to agentic AI tools such as Synopsys.ai Copilot and NVIDIA infrastructure to accelerate chip design and verificat…
Ulanqab prefecture in Inner Mongolia is building more than 5 million data center racks, roughly 1,000 times the scale of Elon Musk's 200,000-chip Colossus supercomputer in Memphis, Tennessee, according to figures publish…
A from-scratch autoregressive transformer trained in 1.5 hours on a single Nvidia RTX 5090 GPU reached 44% on the ARC-1 benchmark, matching specialized architectures TRM and HRM, according to a writeup covered by Mango D…
A research note argues that the AI release slowdown debate conflates release cadence with training cadence, contending that coding, math and cyber are being remade where RL data and feedback loops exist while most other …
Over the past nine months, low Earth orbit has become a contested layer of the AI infrastructure war, with SpaceX filing for a million-satellite AI constellation, signing roughly $26 billion in annualized AI compute agre…
Perplexity made its local-first Portable Computer agent generally available on Windows PCs with NVIDIA RTX graphics cards on 14 September, following its late-August Linux launch, according to NVIDIA. Perplexity VP of eng…
AMD's Helios GPU, the Instinct MI455X, packs 432 GB of HBM4, 23 TB/s of HBM bandwidth per GPU, and 40 PFLOPs of FP4 compute, according to AMD's product brochure. A new educational blog post builds a ladder of BF16 genera…
A GitHub project called "CUDA for AMD on Windows," built by a developer going by "speed," packages the ZLUDA translation layer with AMD's HIP SDK and ROCm to run unmodified CUDA applications on AMD GPUs through an automa…
Anthropic CEO Dario Amodei called for a coordinated slowdown in frontier AI development in a September 12 essay titled "We Must Pace the Frontier" and a CBS Sunday Morning appearance, warning that advanced AI systems cap…
ApexSovereign Infrastructure Protocol launched ApexSovereign, a bare-metal GPU compute service it says cuts model fine-tuning and inference overhead by up to 40% by scanning global spot-market pricing maps every 1,500 mi…
ApexSovereign launched a real-time GPU compute arbitrage protocol that it says scans global spot-market pricing maps every 1,500 milliseconds to secure underutilized bare-metal hardware and cut model fine-tuning and infe…
Anthropic CEO Dario Amodei's call to "pace the frontier" on AI development is drawing pushback from President Trump and House Speaker Mike Johnson, who view any slowdown as a strategic surrender, according to an analysis…
Anthropic plans to embed outside safety evaluators with employee-like access, matching CEO Dario Amodei's call for a slower, coordinated pace among frontier developers, while Sam Altman signaled openness to pacing OpenAI…
Nvidia's next-generation Rubin Ultra accelerator will ship with 192GB of HBM per GPU, down from 288GB in standard Rubin and B300, according to SemiAnalysis, which first reported the change. SemiAnalysis attributes the re…