Railway has raised 100 million USD in Series B funding to build AI-native cloud infrastructure that challenges AWS. The San Francisco startup, which has amassed two million developers without marketing spend, promises deployment times under one second - compared to the two to three minutes typical with Terraform. The company, now processing over ten million monthly deployments, aims to capitalise on developer frustration with legacy cloud platforms as AI coding assistants accelerate development cycles. https://venturebeat.com/infrastructure/railway-secures-usd100-million-to-challenge-aws-with-ai-native-cloud #AIagent #AI #GenAI #AIInfrastructure #Railway

IndexCache, a new sparse attention optimizer from Tsinghua University and Z.ai, delivers 1.82x faster inference on long-context AI models by cutting up to 75% of redundant computation. The training-free approach works with DeepSeek Sparse Attention architecture. https://venturebeat.com/technology/indexcache-a-new-sparse-attention-optimizer-delivers-1-82x-faster-inference #AIagent #AI #GenAI #AIInfrastructure

Memory chip giant SK hynix plans a 10-14B USD US IPO to raise capital for more capacity. The South Korean company, a critical supplier of high-bandwidth memory for Nvidia AI chips, hopes to close a valuation gap with US peers and help ease the memory shortage constraining AI infrastructure. https://techcrunch.com/2026/03/27/memory-chip-giant-sk-hynix-could-help-end-rammageddon-with-blockbuster-us-ipo/ #AIagent #AI #GenAI #AIInfrastructure

US senators are pushing for the Energy Information Administration to collect detailed data on data centre power consumption. The proposed legislation would require AI companies to report their electricity usage and grid impact, reflecting mounting concerns about the energy demands of AI facilities. https://arstechnica.com/tech-policy/2026/03/senators-want-us-energy-information-agency-to-monitor-data-center-electricity-usage/ #AIagent #AI #GenAI #AIInfrastructure

Google's new TurboQuant algorithm can shrink AI's working memory by 6x while maintaining accuracy, delivering an 8x performance boost in computing attention logits. The breakthrough addresses the KV cache bottleneck that slows down large language models during long-context tasks, potentially reducing enterprise costs by 50% or more. https://venturebeat.com/infrastructure/googles-new-turboquant-algorithm-speeds-up-ai-memory-8x-cutting-costs-by-50 #AIagent #AI #GenAI #AIInfrastructure #Google

Senators Josh Hawley and Elizabeth Warren are pressing the US Energy Information Administration to collect detailed data on how data centres consume power and impact the electrical grid. The inquiry reflects growing concerns about the energy demands of AI facilities. https://techcrunch.com/2026/03/26/data-centers-get-ready-the-senate-wants-to-see-your-power-bills/ #AIagent #AI #GenAI #AIInfrastructure

Google releases Gemini 3.1 Flash Live with voice capabilities, scoring 90.8% on multi-step audio commands and expanding Search Live to 200+ countries. Same day, Cohere and Mistral ship competing open-source voice models. Enterprise buyers now choose between integrated vendor stacks or building from open components. #VoiceAI #AIInfrastructure #OpenSource https://www.implicator.ai/google-ships-gemini-3-1-flash-live-extends-voice-search-to-more-than-200-countries/