Make AI and HPC images easier to standardize

Google Cloud previewed Advanced Compute Images for AI and HPC workloads. These are preconfigured, performance-tuned OS images that include things like drivers, fabrics, and Slurm agents.

That can cut operational overhead and may improve performance and cost by reducing variation across environments. For AI/ML and HPC teams, it’s a practical way to simplify provisioning.

Keep BigQuery slot usage under tighter control

BigQuery added project caps for reservations in preview. These scheduling policies let you limit the maximum slots and concurrency per project inside a reservation.

If you manage shared BigQuery reservations, this gives you a clearer way to enforce boundaries. It should help with governance, predictability, and more consistent cost allocation.

Tune network-heavy workloads with a new machine series

Google Cloud also previewed the C4N machine series. It’s network-optimized and offers large local Titanium SSD attachments, with machine types tuned for high network and storage I/O.

For FinOps teams, the new series gives you another option to evaluate when a workload feels expensive but still needs high throughput. It’s especially relevant for I/O-bound services where the instance shape drives efficiency.