AWS Snow Family
A family of physical devices - Snowcone, Snowball Edge, and Snowmobile - that has handled large-scale data transfer and edge computing where network transfer is impractical (support in commercial regions is scheduled to end at the end of 2026)
Overview
AWS Snow Family is a suite of services that has provided offline data transfer and edge computing using physical devices. Its three device types - Snowcone (8-14 TB), Snowball Edge (210 TB of data-transfer storage), and Snowmobile (100 PB) - have served environments with limited network bandwidth and petabyte-scale data migrations. The devices also support edge computing by running EC2 instances and Lambda functions directly on the hardware, enabling data processing in locations without internet connectivity. Note that after the lineup was progressively consolidated around Snowball Edge, the family closed to new customers on November 7, 2025, and support in commercial regions is scheduled to end on December 31, 2026 (as of August 2026).
Three Device Types and Selection Criteria (as Offered)
Snowcone is the smallest device, weighing just 2.1 kg and capable of battery-powered operation. It comes in HDD (8 TB) and SSD (14 TB) models, making it ideal for field data collection via drones or vehicles. A pre-installed DataSync agent automatically transfers data once network connectivity is restored. Snowcone stopped accepting orders on November 12, 2024, and support for existing users ended on November 12, 2025. Snowball Edge offers two models: Storage Optimized and Compute Optimized. Storage Optimized provides 104 vCPUs and 416 GB of memory plus 210 TB of NVMe storage to hold the data being transferred, and its primary use case is migrating tens to hundreds of terabytes. Compute Optimized provides 104 vCPUs and 512 GB of memory (up to 416 GB of which can be assigned to instances running on the device) along with 28 TB of NVMe SSD for those instances. The two capacity figures refer to different things: the 210 TB is the area for the data being shipped, while the 28 TB is the area for instances running on the device. Multiple units can also be clustered together, with cluster sizes ranging from 3 to 16 devices. Snowmobile is a 45-foot shipping container trailer that transfers up to 100 PB in a single trip, used for full data center migrations and exabyte-scale projects. The general selection guideline was: Snowcone for under 10 TB, Snowball Edge for 10-500 TB, and multiple Snowball Edge units or Snowmobile for 500 TB and above. Snowmobile and Snowcone were dropped from the lineup first, and Snowball Edge, the last device remaining, has also closed to new customers (as of August 2026).
Data Transfer Workflow and Encryption
The Snow device workflow consists of five steps: create a job in the console, receive the shipped device, copy data to it, ship it back, and AWS imports the data into S3. All data on the device is protected with 256-bit encryption, with keys managed by KMS. Upon arrival at AWS, data is imported into S3 at a secure facility, and the device is then wiped clean using NIST 800-88 compliant methods. Data can be copied using AWS OpsHub (a GUI tool) or the Snowball Edge Client (CLI). The device can also be exposed as an NFS mount point, allowing existing backup tools to copy files directly. An S3-compatible API is also available, enabling data transfer via the aws s3 cp command. Transfer speeds depend on the device's network interface (10 GbE RJ45, 25 GbE SFP28, or 100 GbE QSFP28) and local network bandwidth. All of these ports are present, but only one can be used at a time, so the ceiling changes with the port you choose. AWS states data transfer performance of up to 1.5 GB/s for Storage Optimized (210 TB), and the actual elapsed time depends on the distribution of file sizes, the degree of copy parallelism, and the read performance of the source storage.
Edge Computing and Operational Patterns
Snowball Edge Compute Optimized can run EC2-compatible instances (sbe-c), enabling compute processing in environments without internet connectivity. Real-world use cases include visual inspection on factory production lines, sensor data analysis aboard ships, and geological data processing at mining sites - all scenarios requiring machine learning inference and data preprocessing where cloud connectivity is unavailable. Lambda functions can also run on the device, and when combined with IoT Greengrass, you can build event-driven processing pipelines at the edge. For container-based designs, EKS Anywhere on Snowball is another option. Compute Optimized models equipped with a GPU (NVIDIA V100) could accelerate inference on sbe-g instances, but they reached end of availability as a previous-generation model on November 12, 2024, and the Compute Optimized model offered since then (2nd generation AMD EPYC) does not include a GPU. Multiple Snowball Edge units can be clustered in groups of 3 to 16 devices, serving as S3-compatible object storage while also acting as distributed processing worker nodes. Because the effective capacity of a cluster is determined by the number of devices and the model, you work backward from the capacity you need to the number of devices. Pricing consists of a per-job service fee, a daily rate for on-site use beyond the days included at no charge, and return shipping. Monthly plans and prepaid 1-year commitment plans are also offered, and applying the prepaid daily-rate discount requires contacting an AWS sales representative. Because the amounts differ by region and job size, check the figures listed on the AWS Snowball pricing page.
References (Official AWS Resources)
The primary sources for this page are the official AWS website and documentation. Check the official pages below for the latest specifications and pricing.
- Amazon S3 official page
- Amazon S3 Glacier official page
- AWS DataSync official page
- AWS Documentation (official)
If this page and the official documentation disagree, treat the official documentation as authoritative.