AI Infrastructure news and analysis on Web Pulse: 34982 curated articles tracking the latest AI Infrastructure developments, tools, and research, updated continuously from vetted sources.
Ray 2.55 adds official TPU support with atomic gang scheduling for TPU slices, aiming to cut GPU costs by enabling Ray-based workloads to run on Google's TPUs. Google's motive is to boost external TPU demand, following A…
Nearly 1 in 10 publicly accessible LiteLLM instances accept a default master key or require no authentication, exposing cloud AI infrastructure to root-level remote code execution and IAM theft, according to research pre…
CrowdStrike announced Agentic Identity Provider (Agentic IdP) in Falcon Next-Gen Identity Security at Fal.Con 2026, giving AI agents cryptographically verifiable identities and brokering short-lived, least-privilege acce…
Traditional load balancing fails for large language model inference because requests vary by over 100x in compute and memory cost, can last from seconds to minutes, and servers are not interchangeable due to KV cache loc…
Vertebrae, a privacy-first AI notetaker from the team behind OpenTools, launched in beta on iOS, offering end-to-end encrypted voice calls and outbound phone calls with local transcription on each participant's device. T…
OpenAI's GPT-6 Astra, its latest and most capable model, is now available in private preview on Snowflake Cortex AI, a launch partnership announced by Snowflake. The model sets new frontiers in benchmarks such as Termina…
Google Cloud announced that Spanner, its operational database, now removes cumulative mutation limits for DML transactions, allowing a single transaction to contain any number of DML statements as long as each individual…
OPAQUE, a San Francisco confidential-AI company, published Weight Custody Manifest (WCM) on September 9, an open specification and developer SDK that lets AI developers place encrypted model weights on customer-controlle…
Zayo introduced Agentic Networking for DynamicLink, an AI agent service that embeds agents into networking workflows to automate tasks such as cloud-to-cloud connectivity and network performance investigation. The agents…
AWS has launched Ray Serve Deep Learning Containers (DLCs), pre-built Docker images for model inference that bundle PyTorch, Ray Serve, and the GPU stack, addressing the maintenance gap left by TorchServe's discontinuati…
CoreWeave CEO Michael Intrator said at the Goldman Sachs Communacopia & Tech Conference on September 9 that public opposition to AI data centers is a reaction to the rapid pace of technological change, and that demand wi…
System76 has launched Thelio Mira AI, a Linux workstation starting at $3,299, designed for local AI workloads with up to two NVIDIA RTX PRO 6000 Blackwell GPUs providing 192 GB of GPU memory. The system, available now, t…
Arduino's new Nesso N1 board powers a smart boxing band created by Manivannan, which uses AI to quantify training data for athletes and enthusiasts. The project, highlighted on the Arduino Blog, demonstrates how the N1 c…
Meta and Panmnesia have jointly proposed a next-generation AI datacenter architecture that operates like a single chip using CXL interconnect technology, as published in Nature Reviews Electrical Engineering. The design …
ByteByteGo's sponsored webinar and article explain that smart model routing can cut LLM costs by up to 10 times by sending simple requests to smaller, cheaper models and reserving larger models for complex tasks. The app…
Omarchy, the keyboard-first Linux desktop built on Arch and Hyprland by David Heinemeier Hansson, has moved its production pipeline, packaging builds, PR review agents, and QA testing to DigitalOcean. DigitalOcean is joi…
AAEON (Stock Code: 6579) released two new microATX industrial motherboards, the MAX-Q870A and MAX-H810A, featuring Intel Core Ultra 200S Series Processors (formerly Arrow Lake-S). The MAX-Q870A supports up to 125W SKUs f…
Vercel has released an experimental beta of DeepSeek V4.1 Flash, available through its AI Gateway, with pricing set at $0.22 per 1 million input tokens and $0.66 per 1 million output tokens, and the beta expires on Septe…
Parallel Works and CoreWeave have partnered to deliver a managed AI cloud environment for DARPA biological research, combining Parallel Works' ACTIVATE platform with CoreWeave's GPU infrastructure to accelerate genomics …
Meta launched Muse, an AI agent that operates a Linux virtual machine in Meta's cloud to autonomously handle tasks like emailing, booking travel, and making purchases. The US-only service offers a free tier with 100 mill…