Compute Engine
Virtual machines for any workload
Run VMs on high-performance and reliable cloud infrastructure. Choose from preset or custom machine types for web servers, databases, and applications that fuel your agents.
Features
Preset and custom configurations
Deploy an application in minutes with prebuilt samples called Jump Start Solutions. Create a dynamic website, load-balanced VM, three-tier web app, or ecommerce web app.
Choose from predefined machine types, sizes, and configurations for any workload, from large enterprise applications, to modern workloads (like containers) or AI/ML projects that require GPUs and TPUs.
For more flexibility, create a custom machine type between 1 and 96 vCPUs with up to 8.0 GB of memory per core. And leverage one of many block storage options, from flexible Persistent Disk to high performance and low-latency Local SSD.
Industry-leading reliability
Compute Engine offers the best single instance compute availability SLA of any cloud provider: 99.95% availability for memory-optimized VMs and 99.9% for all other VM families.
Is downtime keeping you up at night? Maintain workload continuity during planned and unplanned events with live migration. When a VM goes down, Compute Engine performs a live migration to another host in the same zone.
Automations and recommendations for resource efficiency
Automatically add VMs to handle peak load and replace underperforming instances with managed instance groups.
Manually adjust your resources using historical data with rightsizing recommendations, or guarantee capacity for planned demand spikes with future reservations.
All of our latest compute instances (including C4A, C4, C4D, N4, C3D, X4, and Z3) run on Titanium, a system of purpose-built microcontrollers and tiered scale-out offloads to improve your infrastructure performance, life cycle management, and security.
Transparent pricing and discounting
Review detailed pricing guidance for any VM type or configuration, or use our pricing calculator to get a personalized estimate.
To save on batch jobs and fault-tolerant workloads, use Spot VMs to reduce your bill from 60-91%.
Receive automatic discounts for sustained use, or up to 70% off when you sign up for committed use discounts.
Security controls and configurations
Encrypt data-in-use and while it’s being processed with Confidential VMs.
Defend against rootkits and bootkits with Shielded VMs.
Meet stringent compliance standards for data residency, sovereignty, access, and encryption with Assured Workloads.
Workload manager
Now available for SAP workloads, Workload Manager evaluates your application workloads by detecting deviations from documented standards and best practices to proactively prevent issues, continuously analyze workloads, and simplify system troubleshooting.
VM manager
VM Manager is a suite of tools that can be used to manage operating systems for large virtual machine (VM) fleets running Windows and Linux on Compute Engine.
Sole-tenant nodes
Sole-tenant nodes are physical Compute Engine servers dedicated exclusively for your use. Sole-tenant nodes simplify deployment for bring-your-own-license (BYOL) applications. Sole-tenant nodes give you access to the same machine types and VM configuration options as regular compute instances.
Autonomous infrastructure management
Empower AI agents to securely manage your VMs with the new Google Compute Engine MCP server. Agents can discover and execute tools to provision, inspect, and resize resources, enabling you to automate workflows from day-1 builds to day-2 operations, like dynamically adapting to load or hunting down orphaned resources to eliminate waste.
TPU accelerators
Cloud TPUs can be added to accelerate machine learning and artificial intelligence applications. Cloud TPUs can be reserved, used on-demand, or available as preemptible VMs.
Linux and windows support
Run your choice of OS, including Debian, CentOS Stream, Fedora CoreOS, SUSE, Ubuntu, Red Hat Enterprise Linux, FreeBSD, or Windows Server 2008 R2, 2012 R2, and 2016. You can also use a shared image from the Google Cloud community or bring your own.
Container support
Run, manage, and orchestrate Docker containers on Compute Engine VMs with Google Kubernetes Engine.
Placement policy
Use placement policy to specify the location of your underlying hardware instances. Spread placement policy provides higher reliability by placing instances on distinct hardware, reducing the impact of underlying hardware failures. Compact placement policy provides lower latency between nodes by placing instances close together within the same network infrastructure.
Choose the right VM
| Optimization | Workloads | Our recommendation |
|---|---|---|
Efficient Lowest cost per core. |
| General purpose E-Series |
Flexible Best price-performance for balanced and flexible workloads. |
| |
Performance Best performance with advanced capabilities. |
| |
Compute Highest compute per core. |
| |
Memory Highest memory per core. |
| |
Storage Highest storage per core. |
| Specialized Z-Series Z3, Z3H Bare Metal (preview) |
Training, inference, and HPC with GPUs and TPUs Highest performing accelerators. |
| |
Graphics and inference with GPUs Balanced performance and efficiency GPUs. |
| |
Network and I/O High performance I/O and lower TCO for data bound workloads |
| C4N (preview) and M4N (preview) |
Bare Metal Direct hardware control and native architectural parity. |
|
Documentation: Machine families resource and comparison guide
Efficient
Lowest cost per core.
- Web and app servers (low traffic)
- Dev and test environments
- Containerized microservices
- Virtual desktops
General purpose E-Series
Flexible
Best price-performance for balanced and flexible workloads.
- Web and app servers (low to medium traffic)
- Containerized microservices
- Virtual desktops
- Back-office, CRM, or BI applications
- Data pipelines
- Databases (small to medium sized)
- Agentic planning, reasoning, and orchestration
- Secure agent sandboxes for untrusted code execution
Performance
Best performance with advanced capabilities.
- Web and app servers (high traffic)
- Ad servers
- Game servers
- Data analytics
- Databases (any size)
- In-memory caches
- Media streaming and transcoding
- Agentic planning, reasoning, and orchestration
- Secure agent sandboxes for untrusted code execution
- Reinforcement learning (RL) loops and simulations
- Small language model (SLM) inference
Compute
Highest compute per core.
- Web and app servers
- Game servers
- Media streaming and transcoding
- Compute-bound workloads
- High performance computing (HPC)
- CPU-based AI/ML
Memory
Highest memory per core.
- Databases (large)
- In-memory caches
- Electronic design automation
- Modeling and simulation
- High-performance vector databases
- Retrieval-augmented generation (RAG) data layers
- Massive in-memory context caching
- Real-time semantic search
Storage
Highest storage per core.
- Data analytics
- Databases (large horizontal scale-out, flash-optimized, data warehouses, and more)
- Hypervisors
Specialized Z-Series
Z3, Z3H Bare Metal (preview)
Training, inference, and HPC with GPUs and TPUs
Highest performing accelerators.
- AI model training and fine-tuning including large language models (LLM), Mixture of Experts (MoE), deep learning, computer vision
- High-performance AI inference including real-time LLM, generative AI, recommendation systems, conversational AI, natural language processing (NLP)
- HPC including climate modeling, molecular dynamics (drug discovery), and scientific visualization
Graphics and inference with GPUs
Balanced performance and efficiency GPUs.
- AI inference including computer vision and BERT NLP
- Video streaming and analytics
- Video encoding, decoding, and transcoding
- Graphics rendering, and visualization
- Virtual workstations
Network and I/O
High performance I/O and lower TCO for data bound workloads
- Latency-critical AI applications
- Massive-scale data ingestion and AI preprocessing pipelines
- High-throughput data sanitization and secure AI ingress
- Demanding network and security appliances
- High performance computing
- Latency-sensitive databases (such as Oracle)
- Distributed parallel file systems
C4N (preview) and M4N (preview)
Bare Metal
Direct hardware control and native architectural parity.
- Full and direct control over CPU scheduling
- Custom hypervisors and private cloud platforms
- Android and arm device emulation
- Workloads requiring integrated CPU accelerators (QAT, DSA)
- Workloads that are sensitive to CPU performance
- Strict per-core software licensing environments
Documentation: Machine families resource and comparison guide
How It Works
Compute Engine is a computing and hosting service that lets you create and run virtual machines on Google infrastructure, comparable to Amazon EC2 and Azure Virtual Machines. Compute Engine also offers scale, performance, and value to easily launch large compute clusters with no up-front investment.
Compute Engine is a computing and hosting service that lets you create and run virtual machines on Google infrastructure, comparable to Amazon EC2 and Azure Virtual Machines. Compute Engine also offers scale, performance, and value to easily launch large compute clusters with no up-front investment.
Create your first VM
Three ways to get started
Three ways to get started
- Complete a tutorial. Learn how to deploy a Linux VM, Windows Server VM, load balanced VM, Java app, custom website, LAMP stack, and much more.
- Deploy a pre-configured sample application—Jump Start Solution—in just a few clicks.
- Create a VM from scratch using the Google Cloud console, CLI, API, or Client Libraries like C#, Go, and Java. Use our documentation for step-by-step guidance.
How to choose the right VM
Tutorials, quickstarts, & labs
Three ways to get started
Three ways to get started
- Complete a tutorial. Learn how to deploy a Linux VM, Windows Server VM, load balanced VM, Java app, custom website, LAMP stack, and much more.
- Deploy a pre-configured sample application—Jump Start Solution—in just a few clicks.
- Create a VM from scratch using the Google Cloud console, CLI, API, or Client Libraries like C#, Go, and Java. Use our documentation for step-by-step guidance.