AI Infrastructure news and analysis on Web Pulse: 35616 curated articles tracking the latest AI Infrastructure developments, tools, and research, updated continuously from vetted sources.
Experiential Labs, a San Francisco startup in Y Combinator's Summer 2026 batch, open-sourced World Model Optimizer, a model router that learns from an AI agent's production traces to route requests to cheaper models, aim…
OpenAI revealed first benchmarks for its Jalapeño AI accelerator ASIC, developed with Broadcom and Celestica, showing 1.5 to 1.9× more AI work per watt at peak throughput, 1.7 to 3.6× lower end-to-end latency, and 2.1 to…
A developer explains how batch LLM traffic can degrade interactive request latency when both share the same infrastructure, and proposes admission control with protected capacity and token-aware reservations to prevent s…
Google has updated its developer documentation to list the first devices running Gemini Nano 4, including the Pixel 11 series and Samsung's Galaxy Z Flip8, Z Fold8, and Z Fold8 Ultra, while restoring the 'Nano v3 or grea…
Meta's AI-optimized data centers use closed-loop liquid cooling, which recirculates a water-glycol mixture through heat exchangers, to efficiently cool GPU servers, reducing water usage to less than that of a couple of f…
Four reproducible parser failures in vLLM 0.26.0, 0.27.1, and 0.28.0 can drop or garble tool calls and reasoning content while still returning HTTP 200, according to the Ingot team's tests on CPU without GPU or weights. …
IQ Routing, a new AI gateway, claims to cut LLM call costs by 40 to 80 percent by routing each request to the cheapest model that meets a quality bar, with a five-step LangChain loop example dropping from $1.84 to $0.77 …
Sopro V2, a CPU-optimized voice cloning model, enables high-quality text-to-speech synthesis on standard hardware without GPUs, using small audio samples of a few minutes for training. This development removes hardware b…
Chroma's new memory layer, Foundation, uses an agent swarm to cooperatively build a wiki, but the company found that standard database transactions fail because agent retries are expensive, costing minutes and tokens. Ch…
OpenAI is building a dormant "Subscription sharing" feature, internally called ChatPass, discovered in the Codex desktop client by RuntimeWire, which could let third-party AI applications draw from a user's subscription …
Experiential Labs released Experiential, an open-source gateway and router for AI agent workflows that unifies hosted, BYOK, and local models behind an OpenAI-compatible API, with features for controlling model access, s…
Singapore's NUS Medicine, DayOne, and Cortical Labs switched on a data center rack running on sixteen million living human neurons this month, using twenty CL1 units that draw about one kilowatt total versus up to one hu…
The AI boom has reignited Australia's nuclear energy debate, with proponents arguing that nuclear power could provide reliable, low-carbon electricity to meet the surging demand from data centers and AI infrastructure. T…
Anthropic released a research preview of the Model Hardware Standard (MHS), an extension of the Model Context Protocol (MCP) that lets AI agents control physical devices such as robotic arms and lab instruments. Early pa…
At Hot Chips 2026, semiconductor vendors presented divergent designs for AI inference, signaling an era of ferment where no dominant architecture has emerged, according to analyst Ben Bajarin. Companies disagree on wheth…
A developer has shared a technique for reducing latency in LLM API calls by implementing prompt caching at the edge using AWS CloudFront Functions and Lambda. The approach stores prompt-response pairs in a fast lookup ta…
A 24-hour census of a single MCP endpoint recorded 60 distinct named crawler agents making 6,309 requests, including roughly 600 that hit a /v1/buy/* route and received a valid HTTP 402 payment challenge, yet not one att…
Apple's Accelerate framework outperforms GitHub Copilot-generated BLAS Level 3 dtrsm routines on Apple M1 and M2 chips, with performance jumps at matrix sizes that are multiples of 8 attributed to the AMX matrix coproces…
Nvidia's $673 billion forecast reveals that AI compute demand remains robust, with inference revenue now growing faster than training capex, according to the company's Q1 earnings call. The supply constraint has shifted …
Alberta Premier Danielle Smith joined a virtual town hall on AI data centre development Thursday, as the Opposition NDP called for a moratorium on new approvals until stricter rules are in place. The government has faced…