Flexible GPU Commitments Now Cover G2 and G4 Machine Series

Google Cloud has made spend-based Compute flexible committed use discounts generally available for G2 and G4 GPU machine series.

The commitments cover vCPUs, memory, Local SSD, and GPUs. They can apply across Compute Engine, Google Kubernetes Engine, and Cloud Run, and eligible workloads can shift between supported machine series and regions.

This flexibility can help teams align commitments with changing GPU workload placement. It also gives FinOps practitioners a broader way to manage committed spend across supported services, machine series, and regions.

Google Cloud Updates Carbon Footprint Allocation With Granular Energy Certificates

Google Cloud is incorporating granular energy certificates into hourly load matching and will delay its July 2026 Carbon Footprint methodology refresh by one month.

The change affects sustainability reporting and the allocation of carbon-intensity data to cloud customers. Teams tracking cloud-related environmental information will need to account for the updated allocation methodology and revised timing.

This provides more specific information about how carbon-intensity data is assigned to customers, supporting sustainability reporting based on the methodology described by Google Cloud.

Cloud SQL for MySQL 9.7 Is Generally Available

Cloud SQL for MySQL 9.7 is now generally available with a hypergraph optimizer, vector search, JSON Duality Views, and in-place upgrade support from MySQL 8.4.

The hypergraph optimizer and additional database capabilities may improve workload efficiency. In-place upgrades can reduce migration effort and the need for duplicate infrastructure during an upgrade.

For teams managing database costs, the upgrade path can simplify the transition from MySQL 8.4, while improved workload efficiency may support more effective use of existing resources.

Google Cloud Expands Trace Visibility for Applications and Agent Workloads

Google Cloud has added higher Cloud Trace API limits and more trace details, including trace spans for remote Model Context Protocol tool calls and trace exemplars for SQL-based custom dashboard charts.

These capabilities improve visibility into agent and application behavior. They can help teams diagnose performance issues and identify inefficient workloads.

That additional trace data gives engineering and FinOps teams more context when investigating workload behavior that may be contributing to unnecessary resource use or higher costs.