AI Infrastructure news and analysis on Web Pulse: 35178 curated articles tracking the latest AI Infrastructure developments, tools, and research, updated continuously from vetted sources.
Google is preparing to make Search more agentic, with persistent AI agents that can monitor information, evaluate options, and take actions on a user's behalf. The company announced that information agents will launch fo…
Seventy-five researchers, including four from USC Viterbi School of Engineering, attended the Argonne Training Program on Extreme-Scale Computing (ATPESC), a two-week intensive course covering GPU architectures, parallel…
Cerebras Systems announced that the Qwen 3.8 27B model is now available on its platform at 1500 tokens per second, with all public models served unpruned and using selective weight-only quantization for storage. The comp…
Shrijith Venkatramana, developer of LiveReview, explains how autoregressive LLM decoding can be accelerated using speculative decoding and Medusa-style multi-token prediction, which break the sequential token-by-token bo…
A routing error starting at 7:43 am PT on Thursday, September 3 made OpenAI's ChatGPT and Codex unavailable for some users, with a solution implemented by 8:17 am PT, according to an OpenAI spokesperson. Anthropic's Clau…
ADATA Technology Co., Ltd. announced on Thursday from Taipei a reorganization that groups its TRUSTA enterprise brand and ADATA Industrial into five application areas under a new Business Applications hub, with TRUSTA en…
A widespread network failure disrupted OpenAI's ChatGPT, Anthropic's Claude, and SpaceXAI's Grok on Thursday morning, with Downdetector recording a sharp surge in complaints and OpenAI acknowledging elevated error rates …
MCP Studio, a new product from an unnamed company, enables users to conversationally create task-specific, governed MCP servers for any business process, integrating over 1,000 enterprise data sources via REST or SOAP AP…
Four major AI models operated by OpenAI, Anthropic, xAI, and Google suffered rare overlapping service interruptions Thursday morning. Anthropic reported a partial outage affecting Claude Mythos 5.1, Claude Fable 5.1, and…
An engineering team's stress-test of AI-native business operating systems like NubirOS reveals that the unified-OS model struggles with conditional workflows and opaque debugging, despite vendor claims of seamless integr…
NVIDIA confirmed at IFA that its upcoming N1X SoC for the RTX Spark PC platform will launch in two configurations: a top-end model with a 20-core Grace CPU and a Blackwell GPU with 6,144 CUDA cores, and a lower-end model…
Mireye, a company focused on infrastructure for physical world AI agents, has a website at mireye.com that uses Geist, Geist Mono, and -apple-system fonts, with a color palette of #1c1c1e, #ffffff, #f0efe9, and #666666, …
Roboflow published a guide detailing how manufacturers should run computer vision inference at the edge rather than in the cloud, citing a standard cloud round trip of 100-300 ms, roughly 1.49 Gbit/s of raw video per cam…
Creative Strategies projects that 2027 will be the peak year of constraint for AI infrastructure, with demand running 115-120% above annual capacity and critical manufacturing capacity not fully contributing until 2028. …
The House Energy and Commerce Committee held a September 3 hearing on the draft Water Cost Accountability Act, which would bar public water systems from passing data center water infrastructure costs to regular customers…
Three Kansas cities—Emporia, Edgerton, and El Dorado—have filed lawsuits to block citizen petitions that would restrict large data centers, with courts already ruling in favor of the petitions in Emporia and El Dorado. T…
Ably, a realtime messaging platform, argues that raw WebSockets leave AI streaming developers to build and maintain four pieces of infrastructure—reconnection, cancellation, multi-device delivery, and crash detection—eac…
NVIDIA researchers Nandish Mehta and Ben Lee published a paper on March 20, 2025, in Proc. SPIE 13372, Optical Interconnects and Packaging 2025, detailing methods to overcome challenges in co-integrated optics for accele…
Microsoft released MAI-Transcribe-2, a speech-recognition model it claims is faster, more accurate, and cheaper than offerings from OpenAI, Google, or ElevenLabs, priced at 10 cents per hour of audio, a 72% cut from the …
A GPU lock bug that had been quietly degrading an internal RAG sweep was exposed during a pipeline validation run on 2026-06-19, according to a technical post from the engineering team. The team fixed the issue by treati…