AI Infrastructure news and analysis on Web Pulse: 34647 curated articles tracking the latest AI Infrastructure developments, tools, and research, updated continuously from vetted sources.
At the AI Infra Summit in Santa Clara on Sept. 16, 2026, Google distinguished engineer Dave Patterson said the push to make AI agents complete tasks autonomously is increasing demand for CPUs, while Nvidia vice president…
LF Energy expanded on September 15th with two new members, European grid coordination center Coreso and smart-meter manufacturer Landis+Gyr, and four open source projects — AssetLife, CityLearn, EnerGNN and Smart HEMS Be…
A prefill/decode disaggregation setup running DeepSeek-V4-Flash — a 284B total / 13B active model with 256 routed experts — bridged a 700,630-token cold prompt end to end in 11 minutes 26 seconds over plain 10 gigabit Et…
Hong Kong chief executive John Lee announced on Sep 16 the city's first five-year development plan, targeting research and development spending at 3 per cent of GDP after 2030, up from 1.63 per cent in 2024, and naming a…
Uber and ServiceNow exhausted their entire annual allocations for Anthropic's tools within the first months of 2026, with Uber capping spending at USD 1,500 per employee per tool each month, according to The Information.…
A Level1Techs forum user posting as boatman reported that USB RDMA on Macs is "just good software" and not real RDMA hardware offload, after being unable to obtain ib_read_lat, ib_write_lat or ib_send_lat latency numbers…
A research paper published at arxiv.org/abs/2608.09867 and hosted at stolen-thoughts.com demonstrated attacks that decrypt the encrypted chain-of-thought reasoning blocks returned by OpenAI, Anthropic, and Google APIs, e…
Apple's iOS 27 introduces Region-Based Isolation, a Swift compiler feature that enforces data-race safety at compile time by tracking the memory regions of mutable values across async execution boundaries, eliminating th…
Qualcomm is redesigning its next Snapdragon flagship system-on-chip around High Bandwidth Compute (HBC), a near-memory computing architecture intended to address the "memory wall" that limits generative AI decode perform…
A developer built Korea Business Verify (KBV), a remote MCP server that lets AI agents query South Korea's National Tax Service public API for business registration status and KYB identity matching, returning English JSO…
Researchers at the University of Manchester trained NVIDIA's Earth-2 CorrDiff generative downscaling model to forecast air pollution across the UK, completing the training run in two days on a single eight-GPU node of th…
SGLang and Miles added day-0 support for DeepSeek-V4.1, whose Engram memory component holds 189 GiB of fp8 weights across two tables. In paired tests on 4x GB300 (TP4/EP4), host offload of the Engram tables increased KV …
A developer outlined five rules for creating and reusing image transformation presets in game catalogs, recommending that teams define immutable, reusable transformation definitions once and generate responsive thumbnail…
Salesforce and NVIDIA announced Koa on September 15, 2026, Salesforce's first CRM reasoning model, built by post-training NVIDIA Nemotron 3 Super on a proprietary synthetic dataset drawn from 27 years of Salesforce CRM d…
A new arXiv paper (2609.15992v1) introduces "inference networks," a graph-based framework for routing queries among multiple expert large language models to minimize expected inference cost under a target performance con…
A schema-adaptive action-conditioned Joint-Embedding Predictive Architecture (SAAC-JEPA) for CNC dynamics achieved zero-shot RMSE of 0.546, R² of 0.012, and NLL of 0.52 on a sealed target machine sharing only 10 of the s…
A new arXiv paper (2609.16161v1) presents an end-to-end integer-only quantization approach and a sparse dictionary-based KV cache compression strategy that enables LLM inference on Flash compute-in-memory devices, reduci…
A new arXiv paper (2609.16338v1) introduces BITCOS, a distribution-adaptive storage layout for ternary LLMs that costs 2 − z bits per weight given a zero density z, after measuring 29 ternary models and finding zeros acc…
A developer argues that agent memory systems need a consolidation stage beyond the standard record-and-recall pattern, addressing duplicates, silent contradictions, and unbounded growth that degrade retrieval quality ove…
OpenAI and Anthropic's remaining competitive moat is subsidized inference, and losing it would cost them 50% or more of individual developer customers and small businesses, according to an analysis published on vroni.com…