AI Infrastructure news and analysis on Web Pulse: 14051 curated articles tracking the latest AI Infrastructure developments, tools, and research, updated continuously from vetted sources.
Mercor, a startup that supplies domain experts to train AI models, is in talks for a $20 billion valuation, doubling its worth from eight months ago. The company, founded in 2022 by three college dropouts, projects annua…
Wire, a provider of context containers for AI agents, is moving off Cloudflare Durable Objects due to four structural limits: the vector index lived outside the object, compute could not be placed next to data, placement…
RAG (Retrieval-Augmented Generation) has not died but evolved to higher levels of the AI stack, according to Forbes Councils Member Ben Gutkovich. The article argues that despite claims of its demise, RAG remains relevan…
Anthropic is following the same playbook as AWS by building first-party versions of successful products built on its platform, exemplified by the launch of Claude Design, a direct competitor to Figma. The pattern mirrors…
NVIDIA launched a program offering revenue sharing and credit support to help AI cloud providers procure its infrastructure, naming Sharon AI and Firmus Technologies as inaugural partners. Sharon AI will deploy up to 40,…
Ollama raised $65 million in Series B funding led by Theory Ventures, bringing its total funding to $88 million, as the open-source AI infrastructure company reported nearly 8.9 million monthly developers and 85% Fortune…
Europe still has a chance to achieve sovereign AI, but it must move beyond treating AI as a procurement and regulation problem and instead build a complete stack of compute, capital, models, applications, procurement, di…
South Korean chipmaker SK Hynix is listing on the Nasdaq with a $1 trillion valuation, targeting $28 billion in one of the largest foreign equity raises ever. The company's high-bandwidth memory chips are essential for A…
OpenAI released GPT-5.6 Sol, along with Terra and Luna, after a limited launch. CEO Sam Altman told CNBC the model is 54% more token efficient on agentic coding tasks and performs as well or better than competitors. The …
Federal Reserve Bank of New York president John Williams warned that surging demand for artificial intelligence could drive US inflation higher, potentially forcing the Fed to raise interest rates. Williams projected 2.5…
Multiple critical vulnerabilities in Langflow and Dify AI agent platforms have enabled remote code execution and data exposure, with one Langflow flaw exploited within 20 hours of disclosure. Security experts warn that c…
Meta is testing a tiered subscription model called Meta One with Plus and Premium tiers starting at $7.99 per month, while also launching AI business agents on WhatsApp with consumption-based pricing. The company is bett…
Saudi Arabia's Humain, operating under the Public Investment Fund, partnered with Toronto-based Cohere to develop sovereign AI models and expand compute capacity in the Middle East. Humain committed at least 50 megawatts…
BareMetalRT launches TensorRT-LLM natively on Windows without WSL, enabling heterogeneous tensor parallelism across consumer GPUs over standard networking. Users can run HuggingFace models locally via PyTorch backend, su…
Mahesh Rajasekharan argues that context, not just data, is the critical resource for the AI economy, enabling intelligent systems of action. The Forbes Councils Member emphasizes that businesses must focus on contextual …
China is allowing select AI firms, including Alibaba and ByteDance, to buy limited numbers of Nvidia's H200 chips, temporarily easing a training bottleneck while maintaining a focus on tech self-sufficiency, analysts say…
DeepSeek released V4 Flash, a 284B-parameter MoE model with 13B active parameters per token, featuring FP4+FP8 hybrid attention and 1M native context. The open-weight model, available under MIT license, can self-host on …
NVIDIA released Nemotron 3 Ultra, a 550-billion-parameter open-weight MoE model with hybrid Mamba-2 and Transformer architecture, under the Linux Foundation's OpenMDW-1.1 license. The model requires server-class hardware…
Open-weights models are dramatically cheaper per token than closed models, with the five cheapest models on the Artificial Analysis pricing index all being open-weights and the five most expensive all closed. The cheapes…
Independent benchmarks of NVIDIA's DGX Spark desktop through mid-2026 show Qwen 3.5 27B as the most consistent all-rounder on the easy Ollama path, while GPT-OSS 120B pushes nearly 4x the throughput but loses on reasonin…