GitHub

GCP — Services & Pricing

Google Cloud’s design center is data, analytics, AI, and cloud-native infrastructure. It’s the smallest catalog of the three majors but the most opinionated and, in its strongholds, the strongest: BigQuery (serverless warehousing) and Vertex AI are genuine differentiators, GKE reflects that Google originated Kubernetes, and Google’s global premium network underpins low-latency, data-heavy workloads. Its pricing model is also the cleanest — per-second billing with automatic sustained-use discounts and committed-use options, so you capture savings without heavy reservation management.

This page localizes the reference architecture to GCP: Cloud CDN → API Gateway / Cloud Load Balancing → Compute Engine · GKE · Cloud Run → Cloud SQL / Firestore / Memorystore / Cloud Storage → BigQuery · Dataflow · Vertex AI, with Cloud IAM, Cloud KMS, Cloud Operations, and Cloud Build cross-cutting.

CapabilityServiceWhat it isUnit priceSource
Object StorageCloud Storage StandardObject storage standard class for frequently accessed data, regional.$0.02/GB-monthlive
Block Storage (SSD volume)Persistent Disk (Balanced)Balanced SSD persistent disk for general workloads.$0.1/GB-monthlive
VM Compute (general purpose, ~2 vCPU / 8 GB)e2-standard-2 (2 vCPU)Cost-optimized general-purpose VM, 2 vCPU / 8 GB; price normalized per vCPU-hour.$0.0335/vCPU-hourlive
Serverless Functions / Edge ComputeCloud Run / FunctionsRequest-driven containers/functions; first 2M requests/mo free, then per-request.$0.4/1M requestslist
Managed Relational DB (PostgreSQL, ~2 vCPU / 8 GB)Cloud SQL PostgreSQL (2 vCPU)Managed PostgreSQL, ~2 vCPU / 8 GB custom instance, on-demand (compute only).$0.1/instance-hourlist
Internet Data EgressNetwork egressInternet data transfer out to worldwide destinations, first tier.$0.12/GBlist
CDN (content delivery egress)Cloud CDNCDN cache egress (NA/EU first tier).$0.08/GBlist
Managed Kubernetes (control plane)GKEManaged Kubernetes control plane (per-cluster fee after the free tier).$0.1/cluster-hourlist
  • ComputeCompute Engine (cost-optimized E2 / general N-series families; the table’s e2-standard-2 is the value tier), GKE for best-in-class managed Kubernetes, Cloud Run for serverless containers, Cloud Functions for event-driven code.
  • StorageCloud Storage (regional/multi-region classes), Persistent Disk (balanced / SSD), Filestore.
  • DataBigQuery is the flagship: serverless, separates storage from compute, and scales to petabytes without cluster management — often the reason teams pick GCP. Cloud SQL / AlloyDB (managed Postgres), Spanner (globally consistent relational), Firestore / Bigtable (NoSQL).
  • AI/MLVertex AI unifies training, tuning, and serving, with strong access to Google’s Gemini models.
  • Analytics pipelineDataflow (Apache Beam), Dataproc (managed Spark), Pub/Sub for streaming — a cohesive, managed data stack.
  • Data- and analytics-heavy workloads — BigQuery alone is often the deciding factor; the end-to-end data/streaming/ML stack is cohesive and low-ops.
  • Kubernetes-native / cloud-native — GKE is the reference managed-Kubernetes experience.
  • AI/ML and access to Gemini — Vertex AI for a managed, integrated ML platform.
  • Simpler economics — per-second billing + automatic sustained-use discounts mean you save without managing reservations (note GCP compute is the cheapest of the three in the comparison).
  • Lean toward elsewhere when: you need the broadest service catalog (AWS) or the deepest enterprise-Microsoft integration (Azure) — GCP has the narrowest breadth and a smaller enterprise/marketplace footprint.

Cost note: prices above are first-tier on-demand list/live. GCP applies sustained-use discounts automatically and offers committed-use discounts — effective compute cost is often below the on-demand figure without any reservation work.


Pricing generated from the live cost catalog via scripts/gen_cloud_docs.py.