ArcGIS Pro on AWS: Bringing GPU-Accelerated GIS to the Cloud

Organizations across industries are modernizing how they deliver ArcGIS. As ArcGIS Pro continues to evolve, many teams are turning to Amazon Web Services (AWS) to achieve workstation-class performance, elasticity, and simplified operations.

Running ArcGIS Pro in AWS unlocks GPU-accelerated performance while maintaining flexibility in how desktops are delivered—whether as managed cloud desktops, browser-based streamed sessions, or fully customizable GPU servers. Choosing the right AWS service can reduce costs, streamline onboarding, and improve collaboration across distributed GIS teams.

Choosing the Right AWS Platform

Selecting the most suitable AWS service for ArcGIS Pro depends on three key factors:

• Deployment model: Do users need a dedicated GPU desktop, a streaming session, or full administrative control?
• Scalability: How quickly must environments grow or shrink to meet demand?
• Operational ownership: Will IT manage updates and drivers, or should AWS handle the stack?

Amazon WorkSpaces Graphics Bundle: Managed GPU Desktops

For individual GIS professionals who need a ready-to-use ArcGIS Pro environment without the overhead of infrastructure management, Amazon WorkSpaces Graphics Bundles offer Windows 11 desktops powered by dedicated NVIDIA GPUs.

These desktops deliver consistent performance for 2D and moderate 3D workflows—ideal for analysts and project-based teams needing predictable, secure environments.

Recommended ArcGIS Pro Workflows

• Light: Viewing and querying 2D maps
• Medium: Editing, moderate geoprocessing, small 3D scenes
• Not recommended: Dense 3D city models, LiDAR classification, or deep learning

Deployment Steps

1. In the AWS Console, create a WorkSpace and select a Graphics Bundle with Windows 11.
2. Assign the user profile and launch the instance.
3. Install ArcGIS Pro 3.x and activate the license (Named User, Single Use, or Concurrent Use).
4. Validate rendering and frame rate using the Performance Assessment Tool (PAT).
5. Connect through the Amazon WorkSpaces Client for everyday use.

WorkSpaces provide a turnkey GPU experience, users can focus on maps, not management.

Amazon AppStream 2.0: Streamed ArcGIS Pro Sessions

When GIS administrators need to serve multiple ArcGIS Pro users from a single GPU-backed fleet, Amazon AppStream 2.0 delivers ArcGIS Pro as a streamed application. Users launch Pro directly in a browser or client, with GPU acceleration running behind the scenes.

This model is ideal for large organizations, universities, and training environments that require scalability and simplified access.

Recommended ArcGIS Pro Workflows

• Light: Map reviewers, simple 2D projects
• Medium: Analysts editing in 2D/3D, moderate raster analysis
• Heavy: Complex visualization or analysis tasks with GPU acceleration

Design Highlights for ArcGIS Pro

• GPU-Backed Fleets: Graphics.g4dn (NVIDIA T4), Graphics.g5 (NVIDIA A10), and Graphics.g6/g6f (NVIDIA L4)
• Scaling: Fleets automatically expand or contract with user demand

Deployment Steps

1. In the AWS Console, create an AppStream 2.0 Fleet using Graphics.g4dn, g5, g6, or g6f instances.
2. Build an image by installing ArcGIS Pro 3.x, GPU drivers, and required extensions.
3. Publish ArcGIS Pro as an app in an AppStream Stack.
4. Assign Stack access to users or groups.
5. Users connect via browser or AppStream client—ArcGIS Pro streams directly.
6. Validate session performance using PAT.

AppStream 2.0 enables browser-based GIS anywhere, with elastic scaling that adapts to real-time demand.

Amazon EC2 GPU Instances: Full Control and Maximum Performance

For power users and advanced teams running complex 3D visualization, LiDAR editing, or deep learning, Amazon EC2 GPU Instances (G4, G5, G6, and G6f) provide full Windows desktops with administrative control.

This option transforms AWS into a powerful virtual GIS workstation. It provides a complete, customizable environment that’s ideal for development, high-end spatial analysis, and integration testing of custom Python and ArcPy scripts. With Amazon EC2, users can choose from a broad range of compute and GPU-backed instance types to fit their specific workload needs. This flexibility allows GIS administrators to scale performance and cost efficiently, ensuring each project, whether it’s 2D mapping, 3D visualization, or deep learning—runs at its optimal level.

Instance Scalability for ArcGIS Pro Workloads

AWS provides several GPU-backed instance families that can be tuned to the scale and complexity of ArcGIS Pro tasks:

• G4dn instances – Powered by NVIDIA T4 GPUs, G4dn offers a versatile and cost-effective option for ArcGIS Pro users focused on 2D mapping, cartography, and moderate 3D visualization. These instances are ideal for lightweight design, editing, and visualization workflows where GPU acceleration enhances responsiveness without requiring the full performance of higher-end GPUs.
• G5 instances – Equipped with NVIDIA A10 GPUs, G5 provides a strong balance of GPU power and system memory for demanding ArcGIS Pro workloads. It excels in large 3D scene rendering, complex raster processing, and terrain analysis. G5 instances are a solid choice for teams that need higher GPU throughput and performance consistency across multi-layered, data-intensive projects.
• G6 instances – Powered by NVIDIA L4 GPUs paired with Intel Xeon Sapphire Rapids CPUs, these instances deliver excellent balance between graphics performance and compute efficiency. G6 is ideal for high-performance 3D visualization, ArcGIS Pro rendering, and GPU-accelerated geoprocessing where consistent frame rates and lower power draw are priorities.
• G6f instances – Built on the same NVIDIA L4 GPU but paired with AMD EPYC Genoa processors, G6f instances introduce fractional GPU configurations that allow smaller, cost-effective allocations of GPU resources per virtual machine. This provides additional vCPU capacity and pricing efficiency for multi-threaded ArcGIS Pro workflows such as LiDAR classification, deep learning with ArcGIS Pro extensions, or large geoprocessing batch jobs.

Deployment Steps

1. Launch an EC2 instance with GPU support (G4dn, G5, G6, or G6f).
2. Attach an SSD-based EBS volume for project data and cache.
3. Install NVIDIA drivers from the AWS Marketplace.
4. Install ArcGIS Pro 3.x and configure licensing.
5. Validate frame rate and rendering using PAT.
6. Save the configuration as a golden AMI for reuse.

EC2 delivers complete control and repeatable builds—the closest cloud equivalent to a physical workstation, with the elasticity to scale performance for any ArcGIS Pro workload.

Key Considerations for ArcGIS Pro on AWS

• GPU Memory: At least 4 GB VRAM; G5, G6, and G6f instances offer larger buffers for 3D and raster workflows.
• System Memory: 32 GB minimum; 64 GB recommended for LiDAR or raster-heavy analysis.
• Network Latency: ≤ 200 ms RTT for responsive interaction—ArcGIS Pro is latency-sensitive.
• Bandwidth: 1.5 Mbps (light), 3 Mbps (medium), 5+ Mbps (heavy 3D).

Together, these ensure consistent, high-performance experiences for ArcGIS Pro users in the cloud.

Bringing It All Together

Whether you’re enabling a single analyst or scaling ArcGIS Pro across hundreds of users, AWS offers multiple paths to GPU-accelerated performance:

• WorkSpaces Graphics Bundle → Best for individuals needing managed, always-ready GPU desktops
• AppStream 2.0 → Best for groups needing scalable, browser-based streaming
• EC2 GPU Instances (G4/G5/G6/G6f) → Best for advanced users requiring full control and workstation-class power

By combining these AWS services, organizations can deliver ArcGIS Pro securely, flexibly, and at scale—meeting the diverse needs of modern GIS teams.

ArcGIS Pro on AWS brings GPU-accelerated GIS to the cloud, empowering professionals to visualize, analyze, and collaborate from anywhere—without compromising performance.