Google Cloud Platform (GCP) stories
Most enterprises will need costly infrastructure changes before agentic AI can move from pilots to production, Google Cloud research says.
Large language model post-training could get cheaper as idle accelerator time is cut from about 60% to 30% in early tests.
The switch has cut Voicify's model costs by up to 30% while helping the platform keep restaurant and healthcare calls flowing at peak times.
Partners will be able to sell higher-margin advisory and managed services as organisations brace for AI-driven identity sprawl and PQC risk.
By shifting orchestration to Google Cloud's managed service, the payments firm cut monthly data platform costs by about 30% and improved reliability.
Cloud operators can now use IAM conditions to curb privilege escalation and narrow access to MCP tools, BigQuery and logs.
More than 1,000 engineers and executives are expected in Warsaw as digital firms face mounting pressure to keep systems fast and reliable at scale.
Downtime risk is lower for apps running across multiple regions, as traffic can now shift to a healthy Cloud Run region within seconds.
Cloud-based testing could cut prototype costs as Panasonic Automotive brings cockpit software validation closer to production hardware.
Developers can now use Google Cloud's examples to build and govern Gemini-powered agents for approvals, compliance and data workflows.
The preview could ease multi-region compliance headaches by letting organisations apply column-level security through global tags and regional policies.
Developers are being told to trim AI context and automate routine checks as token-heavy workflows raise latency, errors and cloud costs.
The new workflow could cut model migration from months to hours, easing the testing burden for businesses running multiple AI features.
Defenders could gain a faster edge against AI-driven attacks as Google Cloud ties Gemini, Wiz, CodeMender and Mandiant into one platform.
The move could help organisations curb cloud admin account abuse by revoking elevated access automatically after approved sessions end.
Developers can now improve Chinese, Japanese and Korean search in AlloyDB without external tokenisers, cutting preprocessing overhead for PostgreSQL users.
Security and staffing gaps are slowing enterprise rollouts, with networking now emerging as a key bottleneck for agentic AI projects.
The move could help IT teams track staff use, audit access and compliance risks across large Gemini Enterprise roll-outs without bespoke tools.
Fraud teams could cut investigation time by 90% as the new stack combines real-time scoring, case assembly and human review in milliseconds.
That could cut cloud migration risk and planning time for firms moving workloads across Amazon Web Services, Azure and Google Cloud Platform.