GKE stories
The service aims to cut storage overheads for firms running thousands of AI agents, while keeping workspaces isolated and quick to spin up.
Teams running AI agents at scale could save idle compute and tighten isolation as Google opens Agent Substrate on Kubernetes Engine.
Enterprises seeking AI-ready infrastructure may see Google Cloud's higher score as validation of its database, analytics and Kubernetes services.
Under heavy generation load, the smaller model kept scaling on TPU v6e while the 27B version hit a throughput plateau above 64 users.
The recognition bolsters Google's pitch to enterprises seeking AI-ready cloud systems with tighter data control as rivals vie for big workloads.
Enterprises facing AI demand spikes are being told to secure spare capacity and automate backups to avoid outages and overspending.
The experimental setup lets AI teams run untrusted model code in isolated Ray clusters on Google Kubernetes Engine without exposing host containers.
Customers face a 2029 deadline to test post-quantum controls as Google Cloud begins rolling out quantum-safe encryption across its services.
The preview could help GKE users enforce cluster-wide security without clashing with namespace rules, especially in multi-tenant environments.
The upgrades are aimed at easing bottlenecks for large AI and machine learning deployments, with clusters now scaling to 15,000 nodes.
Tests on an n2-standard-48 virtual machine showed Google Cloud's sandbox and orchestration tweaks could cut per-agent costs by up to 75%.
The framework targets CISOs and platform teams as they move AI systems into production, amid rising risks from prompts, models and outputs.
The update gives enterprises managed access to Anthropic's models with built-in failover, data residency and compliance controls for production AI use.
Carmakers could cut warranty costs and breakdowns as the open-source system unifies telemetry, maintenance and service data in one stack.
The ranking boosts Google Cloud's bid to win more AI infrastructure spend as firms look for cheaper, scalable systems for training and inference.
Carmakers may gain cheaper, more secure software-defined vehicle tools as the open-source Nexus SDV core targets 100 million devices.
Benchmark tests show Managed Lustre can cut AI inference costs by more than half, easing GPU demand for long-context model serving.
The self-hosted gateway gives firms central control over sign-ins, access and costs as Claude Code roll-outs move beyond small developer teams.
The update aims to simplify security operations as enterprises grapple with unmanaged devices, partners and multi-cloud workloads across AI projects.
Retailers using Malachyte's Google Cloud system are seeing sales double as it personalises search and recommendations within 100 milliseconds.
