AI Chips news and analysis on Web Pulse: 6901 curated articles tracking the latest AI Chips developments, tools, and research, updated continuously from vetted sources.
Astera Labs announced the Leo X-Series fabric-attached memory controller plus next-generation Leo 2 E-Series and P-Series controllers, which support up to 4 TB of DDR5 capacity per controller and up to 768 GB of repurpos…
Nvidia Corp. is introducing a new infrastructure category it calls "scale-in," extending its BlueField data processing unit, DOCA software and the OpenShell 0.1.0 agent runtime to enforce security boundaries across AI fa…
Running an open-weight model at the level of Opus 5.5 requires roughly 610 GB of VRAM plus RAM for a maximally quantized Kimi K3, while the full-precision model would need 1600 GB, according to a Level1Techs forum discus…
Researchers submitted a paper on 24 Sep 2026 presenting Loom, a tuning-free symbolic compiler framework for tile-based SPMD programs on spatial dataflow architectures. Loom formulates one CP-SAT problem per schedule cand…
AT&T now routes roughly 40% of its internal AI requests through open-weight models, cutting AI coding costs by 56% while absorbing only a 2% decline in output quality, the company said. The telecom giant processes approx…
NetApp announced Novus at its annual Insight event, claiming the file system delivers 100 TB/sec of scale-out performance for AI factories and supports hundreds of thousands of GPUs, according to Chief Product Officer Sy…
TensorFold 0.3.6.2 delivered decode speeds over 62 tokens per second on a single stream and 119 tokens per second across five concurrent streams running Qwen3.8-Flash-Next on a single Nvidia DGX Spark, according to bench…
A developer testing an AMD Ryzen AI NPU (~50 TOPS) on a Beelink SER9 found the hardware, driver, and runtime all work — including running passthrough kernels, conv2d, and a ResNet bottleneck block directly on the NPU til…
A wave of open-weight model releases this summer — including Zhipu's GLM-5.3, Moonshot AI's Kimi K3, DeepSeek's V4 Pro and V4.1 Flash, and Xiaomi's MiMo-V2.6 — narrowed the gap with top proprietary models, with DeepSeek'…
Two trailers belonging to autonomous trucking company PlusAI — one bearing Nvidia branding — were stolen overnight Thursday and recovered the same afternoon by Newark Police after a PlusAI employee's friend spotted them …
Nvidia announced its Open Agent Safety Platform on Monday, a two-part security system it says can stop autonomous AI agents from misbehaving, built around the open-source OpenShell sandbox and a hardware-level watchdog c…
Synopsys certified its digital and analog EDA flows on TSMC's A14 process and completed 64G UCIe IP tape-outs on TSMC's 2nm and 3nm nodes, part of a broader joint announcement around TSMC's Open Innovation Platform ecosy…
Los Alamos National Laboratory is co-designing AI infrastructure with HPE and NVIDIA, including the Venado supercomputer built on NVIDIA Grace Hopper Superchips that fuse an ARM-based CPU and Hopper GPU on a single modul…
The University of Utah deployed a full-stack sovereign AI factory built with HPE and NVIDIA, a public-private-philanthropic co-investment championed by the university, the State of Utah, and the Huntsman Family Foundatio…
SkyPilot and Cast AI address GPU scarcity from opposite ends, with SkyPilot running workloads across 20+ clouds and selecting the cheapest available GPU while Cast AI's OMNI Compute extends existing EKS, GKE, and AKS clu…
Nvidia is combining its OpenShell agent software with Sentry, a new hardware watchdog, to create the Open Agent Safety Platform, which is designed to isolate AI agents that break out within milliseconds. Nvidia's watchdo…
China has expanded its foreign-travel restrictions on top AI and chip executives to cover their spouses and children, who now need government approval before going abroad even for short trips, Bloomberg News reported, ci…
Google's Project Suncatcher will launch its first physical test payload, a refrigerator-sized Planet Labs satellite carrying four Google Trillium Tensor Processing Units, on a SpaceX Falcon 9 rideshare mission from Vande…
Inferact, a startup whose team includes vLLM maintainers, released open-source "megakernels" for Google's TPU v7 (Ironwood) that delivered 709 tokens per second for a single user on Moonshot AI's Kimi K3 across 16 chips,…
China has extended overseas travel restrictions to the family members of leading AI and chip executives at private companies, requiring government pre-approval before relatives can leave the country, according to a Bloom…