Cloud Run is Google’s fully managed serverless container platform that enables running containerized applications without infrastructure management. With automatic scaling from zero to thousands of instances, billing on a 100ms basis, and support for any programming language, Cloud Run is ideal for modern web applications, APIs, microservices, and AI workloads.
What is Cloud Run?
Cloud Run is a serverless compute platform that combines the flexibility of containers with the simplicity of fully managed hosting. Unlike traditional cloud platforms, you don’t need to worry about servers, clusters, or scaling: Cloud Run handles this automatically.
The platform uses Google’s global infrastructure with over 20 regions worldwide, including several in the EU. Cloud Run offers two billing models: with request-based billing you only pay for actually used resources, calculated to the nearest 100 milliseconds, and the service automatically scales to zero instances when there is no traffic. With instance-based billing you pay for the entire instance lifecycle, which suits steady traffic and background processing.
Cloud Run comes in four flavors: Services for HTTP and gRPC workloads, Jobs for batch processing up to 24 hours, Worker Pools for non-request background workloads, and Functions for lightweight event handlers.
Core Features
- Any containerized application with HTTP, gRPC, or WebSocket endpoints
- Automatic scaling from zero; default maximum of 100 instances, higher via quota
- Two billing models: request-based (100ms granularity) and instance-based
- GPU support for AI inference (NVIDIA L4 with 24 GB, RTX PRO 6000 Blackwell with 96 GB)
- Worker Pools for queue-based and streaming workloads, Jobs for batch up to 24 hours
- Direct VPC egress, multi-region services, and Secret Manager integration
Typical Use Cases
Web Applications and APIs: Modern web applications with dynamic traffic. Automatic scaling handles traffic spikes effortlessly while scale-to-zero saves costs during low traffic.
AI Inference with GPU: NVIDIA L4 and RTX PRO 6000 Blackwell GPUs for real-time AI inference, including LLMs, computer vision, and speech models. GPU instances start in about 5 seconds and scale to zero when not in use. GPUs are available in EU regions such as europe-west1 and europe-west4.
Event-Driven Microservices: Seamless integration with Pub/Sub and Eventarc for asynchronous, event-driven architectures.
Background and Streaming Workers: Worker Pools consume from pull queues (Pub/Sub, Kafka) and run continuous streaming or inference pipelines without an HTTP endpoint.
Scheduled Batch Jobs: Combine Cloud Run Jobs with Cloud Scheduler for periodic tasks like ETL pipelines, data aggregation, or backups.
Benefits
- Zero infrastructure management
- Scale to zero for cost efficiency
- Any language or framework in containers
- Fast deployments under 1 minute
- Multi-region services for high availability and low latency
- Native integration with the Google Cloud ecosystem (Cloud SQL, Pub/Sub, Gemini Enterprise Agent Platform)
Integration with innFactory
As a certified Google Cloud Partner, innFactory supports you across the full Cloud Run lifecycle:
- Microservices and event-driven architecture with Services, Jobs, and Worker Pools
- Migration from App Engine, GKE, or on-premises to Cloud Run
- CI/CD pipelines with Cloud Build, GitHub Actions, and Terraform
- AI inference with GPU acceleration and integration with the Gemini Enterprise Agent Platform (formerly Vertex AI)
- Cost optimization, observability, and security hardening
Contact us for a free consultation on Cloud Run and serverless architectures on Google Cloud.
Available Tiers & Options
Cloud Run Services
- Zero infrastructure management
- Automatic scaling to zero
- Fast deployments under 1 minute
- GPU support for AI inference
- Stateless applications
- Request timeout up to 60 minutes
Cloud Run Jobs
- Batch processing with parallel execution
- Runs up to 24 hours
- Task-based scaling
- No HTTP endpoints
- Asynchronous execution
Cloud Run Worker Pools
- For non-request workloads (pull queues, streaming, background jobs)
- Persistent, long-lived instances
- Manual or custom scaling
- GPU support for inference pipelines
- No HTTP endpoint
- No automatic scaling
Cloud Run Functions
- Simple event handlers
- Full Cloud Run Service control
- Source code deployment
- Limited configuration options
Typical Use Cases
Technical Specifications
Frequently Asked Questions
What is Cloud Run?
Cloud Run is a fully managed serverless container platform from Google Cloud. It enables running containers without infrastructure management, scales automatically based on traffic, and only charges for actually used resources (100ms granularity). Cloud Run supports any programming language that runs in containers and offers GPU acceleration for AI workloads.
What is the difference between Cloud Run Services, Jobs, Worker Pools, and Functions?
Cloud Run Services are for HTTP-based workloads (APIs, web apps) with automatic scaling. Cloud Run Jobs are for batch processing and long-running tasks (up to 24h) without HTTP endpoints. Cloud Run Worker Pools, generally available since 2026, target non-request workloads such as pull queues, streaming, or background jobs: they use persistent instances and manual or custom scaling instead of request-based autoscaling. Cloud Run Functions offer a simplified experience for event handlers with fewer configuration options but are based on Cloud Run infrastructure.
When should I use Cloud Run instead of Cloud Functions or GKE?
Use Cloud Run when you need container-based applications with HTTP/gRPC endpoints, want more control than Cloud Functions, but less complexity than GKE. Cloud Run is suitable for most web and API workloads. GKE is better for complex Kubernetes-native applications with advanced networking requirements. Cloud Functions is ideal for simple event handlers without container management.
Does Cloud Run support GPU acceleration for AI models?
Yes, Cloud Run offers GPU support for services, jobs, and worker pools. Available options are NVIDIA L4 (24 GB VRAM) and, since 2025/2026, NVIDIA RTX PRO 6000 Blackwell (96 GB VRAM) for larger models. One GPU per instance is supported, instances start in about 5 seconds and scale to zero when not in use. This is ideal for hosting Large Language Models (LLMs like Llama, Mistral, Gemma), computer vision, video transcoding, and other GPU-intensive applications. GPUs are available in EU regions such as europe-west1 (Belgium) and europe-west4 (Netherlands).
How does scaling work in Cloud Run?
Cloud Run automatically scales services based on incoming requests. With no traffic, the service scales to zero containers (no costs). During traffic spikes, new instances start within seconds. The default maximum is 100 instances per revision and can be increased significantly via regional quota. You can configure minimum and maximum instances as well as concurrency per instance (up to 1000). Worker pools use manual or custom scaling instead.
What are Cloud Run Worker Pools and what do I use them for?
Cloud Run Worker Pools are a resource generally available since 2026 for non-request workloads. Unlike services, they have no HTTP endpoint and do not autoscale based on incoming requests: they use persistent, long-lived instances with manual or custom scaling. Typical use cases are pull-queue consumers (for example Pub/Sub or Kafka), streaming pipelines, and continuous background processing. Worker pools also support GPUs for inference pipelines.
How does Cloud Run billing work?
Cloud Run offers two billing models. With request-based billing (default), you only pay when instances process requests, start, or shut down, billed at 100ms granularity. The instance-based model charges for the entire instance lifecycle and suits steady traffic or background processing. In both cases you pay for CPU, memory, requests, and outbound traffic. There is a free monthly tier, and with request-based billing scale-to-zero means no costs when there is no traffic. See the Cloud Run pricing page for current values.
Can I deploy Cloud Run across multiple regions?
Yes, since 2025 Cloud Run supports deploying multi-region services from a single gcloud command or via YAML or Terraform. This distributes a service across multiple regions for higher availability and lower latency without deploying to each region separately. Combined with a global load balancer, this creates a robust, globally distributed architecture.
Is Cloud Run GDPR compliant and which EU regions are available?
Yes, Cloud Run is GDPR compliant. Available EU regions are europe-west1 (Belgium), europe-west3 (Frankfurt), europe-west4 (Netherlands), europe-west6 (Zurich), europe-west9 (Paris), europe-north1 (Finland). Google Cloud offers comprehensive data protection controls, compliance certifications, and data residency guarantees.
Note: All product information on this page has been compiled with care, but is provided without guarantee and may be outdated or incomplete. Cloud services evolve rapidly — features, pricing, SLAs, and availability change frequently. Authoritative and up-to-date information can only be found on the official product page of Google Cloud (official documentation). This page does not represent an offer by Google Cloud.
