AI Infrastructure news and analysis on Web Pulse: 35157 curated articles tracking the latest AI Infrastructure developments, tools, and research, updated continuously from vetted sources.
Shinsegae Group Chairman Chung Yong-jin joined U.S. political and technology leaders in Washington on Thursday for the launch of Foundry School, a workforce initiative aimed at strengthening advanced manufacturing and te…
LG will host LG SPARK 2026, a two-week AI festival from Sept. 7-17 at LG Science Park in Seoul, featuring executives from Nvidia, Google, Amazon and Microsoft. The event, which drew about 60,000 attendees last year, will…
Warp launched Factory Benchmarks on September 3rd, enabling engineering teams to compare coding models against their own historical tasks and route future work based on results. The tool, part of Warp's agent infrastruct…
Cango Inc. announced on September 3 that its EcoHash subsidiary has begun commercial GPU computing at its 50-megawatt Bitcoin mining facility in Georgia, with a dedicated AI compute section supporting up to 3 megawatts o…
Researchers introduced AdaptiveSpec, a training-free per-step speculative decoding method that adapts both draft verification and tree shape from internal decoding signals, improving throughput over EAGLE-3 by up to 56% …
Researchers introduced GrowPage, an on-demand KV budgeting framework for efficient LLM reasoning serving, which dynamically adjusts key-value cache capacity per request based on attention demand. GrowPage uses dual-times…
Researchers introduced LeanStream, a speculate-and-refine streaming framework for on-device large language model (LLM) inference that reduces memory usage by 4.8x to 7.5x and improves token generation throughput by 1.6x …
A developer has created an open-source bridge that connects ChatGPT Web to local IDEs via the Model Context Protocol (MCP), allowing developers to offload heavy code-planning tasks to their ChatGPT Plus subscription whil…
A developer outlines strategies for managing AI token costs, emphasizing that usage-based pricing requires a shift from fixed-cost budgeting to measuring cost per outcome. The article introduces the 'Elephant and Goldfis…
A contributed article argues that AI infrastructure planning must begin earlier as AI workloads become more distributed across cloud, data center, and edge environments, warning that delays can slow deployment and postpo…
OpenAI's GPT-6 Astra is now generally available for all customers in Microsoft Foundry, designed to help organizations make decisions for complex work and execute across applications and systems. The model can take open-…
Kong AI Gateway 2.0, now available on Google Cloud, provides unified security and governance for AI traffic across GKE, Cloud Run, and Vertex AI. The gateway treats AI models, agents, and MCP servers as first-class entit…
Nvidia Corp. announced the Personal AI Router (PAIR) at IFA 2026 in Berlin, a beta tool that lets users cluster idle Macs and PCs in a home to run small language models and accelerate agentic AI workloads by distributing…
NVIDIA's GH200 Grace Hopper Superchip modules and baseboards are appearing on eBay at prices below conventional PCIe H200 variants, prompting hobbyists to explore building custom HPC workstations from them. A forum user …
LLMRix Inc. has released LLMRix Model Router, an open-source multi-model routing and orchestration framework for Java that manages provider differences, model selection, failover, quotas, costs, and observability. The fr…
A developer behind the open-source project MeshCtx detailed how running AI agents in production revealed that failures often occur silently at 2 AM, going unnoticed until the next morning. The project, which runs on a th…
Figure has committed an initial $3.5 billion in compute from AI infrastructure provider Nscale, securing a path to as many as 100,000 NVIDIA Vera Rubin GPUs to scale its Helix AI models for humanoid robots. The first sys…
MemWeaver, a temporal vector database with predictable memory usage, calculates memory cost as a function of vectors, dimensions, and M before deployment, with 1M vectors at dim=128 and M=16 requiring approximately 640 M…
The UK government has launched the first procurement competitions under its £100m Sovereign AI R&D Procurement Scheme, inviting UK-based firms to develop AI solutions for public sector challenges including NHS productivi…
An AWS Builder series concluded with a hands-on test of Amazon FSx for ONTAP S3 access points, finding that features like qtrees, quotas, and FlexClone work on this path despite NetApp's documentation listing them as uns…