AWS Knowledge

Understanding Amazon Sagemaker Pricing

Stuart Lundberg

Aug 31, 2023

    Table of contents will appear here.
    Table of contents will appear here.
    Table of contents will appear here.

Understanding Amazon SageMaker Pricing

As someone part of a tech startup, you may find yourself constantly gawking at the rising costs of AWS month after month. As your company grows, chances are that you may find your company using more types of Amazon Web Services, including Amazon SageMaker, a fully managed machine learning (ML) service.

What is Amazon SageMaker

Amazon SageMaker is a significant player in the machine learning management business. SageMaker is a fully managed service that provides every developer and data scientist with the ability to build, train, and deploy machine learning (ML) models quickly. SageMaker removes the heavy lifting from each step of the machine learning process to make it easier to develop high-quality models. SageMaker can start as low as $0.05 per hour for Studio Classic.

Amazon SageMaker operates on a pay-as-you-go model, meaning you pay only for what you use, billed by the second, with no minimum fees. This flexible pricing is designed to scale with your needs, whether you're experimenting with new models or deploying machine learning solutions at scale. For newcomers to the AWS ecosystem, SageMaker even provides a free tier to help you get started without any upfront investment.

How does Amazon SageMaker work?

Amazon SageMaker streamlines the machine learning workflow, enabling users to prepare, build, train, and deploy ML models. It integrates various components that are essential for machine learning, such as Jupyter notebook instances for data exploration and preprocessing, built-in algorithms and frameworks, automatic model tuning, and one-click deployment.

SageMaker greatly simplifies the process of machine learning by providing a suite of modular components that can be used independently or together. There are some functions for Amazon SageMaker that can be used to achieve different goals:

  • SageMaker Ground Truth: helps you build highly accurate training datasets for machine learning quickly by applying human feedback across the ML lifecycle. It offers easy access to human labelers through Amazon Mechanical Turk, third-party vendors, or your own employees, to create labeled data required for training. Ground Truth also uses machine learning to offer pre-labeling, which can help reduce the time and cost of manual labeling by suggesting labels for annotators to review and approve.

  • SageMaker Studio: an integrated development environment (IDE) for machine learning that provides a single, web-based visual interface where you can perform all machine learning development steps. This includes writing code, tracking experiments, visualizing data, and debugging and monitoring models. With SageMaker Studio, you can easily build, train, debug, deploy, and monitor your machine learning models.

  • SageMaker Model Monitor: regularly tracks the quality of your machine learning models in production. It detects deviations in data quality, model performance, and operational metrics, and provides alerts so you can take corrective action. This ensures that your deployed models continue to operate as expected and maintain high accuracy over time. By utilizing SageMaker Model Monitor, you can automate the monitoring of your models, saving time and reducing the risk of human error. This tool is particularly useful for maintaining the integrity of your models in cases where data patterns change over time, which can lead to model drift.

Deep Dive into SageMaker Pricing Structure

With SageMaker, you pay only for what you use. You have three choices options: an On-Demand Pricing that offers no minimum fees and no upfront commitments, and the SageMaker Savings Plans that offer a flexible, usage-based pricing model in exchange for a commitment to a consistent amount of usage, and an AWS free tier for your first two months.

  • Amazon SageMaker Free Tier: For two months, you can try AWS SageMaker free of charge. As part of the AWS Free Tier, your free tier starts from the first month when you create your first SageMaker resource. The details of the free tier usage limits for Amazon SageMaker can be found in the table below or on AWS’s website. Whether you're utilizing the service for data preparation, building models, or deploying and managing them, each stage comes with its own set of costs that need to be factored into your AWS machine learning expenses.

  • Amazon SageMaker capability

  • Free Tier usage per month for the first 2 months

  • Studio notebooks, and notebook instances

  • 250 hours of ml.t3.medium instance on Studio notebooks OR 250 hours of ml.t2 medium instance or ml.t3.medium instance on notebook instances

  • RStudio on SageMaker

  • 250 hours of ml.t3.medium instance on RSession app AND free ml.t3.medium instance for RStudioServerPro app

  • Data Wrangler

  • 25 hours of ml.m5.4xlarge instance

  • Feature Store

  • 10 million write units, 10 million read units, 25 GB storage (standard online store)

  • Training

  • 50 hours of m4.xlarge or m5.xlarge instances

  • Amazon SageMaker with TensorBoard

  • 300 hours of ml.r5.large instance

  • Real-Time Inference

  • 125 hours of m4.xlarge or m5.xlarge instances

  • Serverless Inference

  • 150,000 seconds of on-demand inference duration

  • Canvas

  • 160 hours/month for session time, and up to 10 model creation requests/month, each with up to 1 million cells/model creation request

  • HyperPod

  • 50 hours of m5.xlarge instance

  • Free Tier usage per month for the first 6 months

  • Experiments

  • 100,000 metric records ingested per month, 1 million metric records retrieved per month, and 100,000 metric records stored per month

  • On-Demand Pricing: On-Demand Pricing for Amazon SageMaker is a flexible pricing option where you are charged based on the actual consumption of resources without any upfront fees or long-term commitments. Charges are calculated by the second for the instances and services you use, making it easy to start and scale with your machine learning projects. You are billed separately for each of the different components of SageMaker that you use, such as notebook instances, training jobs, real-time inference, batch transform jobs, and storage. This granular billing approach allows you to optimize costs according to your usage patterns and needs.

  • For instance, if you are using the SageMaker notebook instances, you will be charged for the duration that the instance is running. Similarly, training jobs are billed for the compute resources. You can learn more about SageMaker on-demand pricing on SageMaker AWS.

  • Amazon SageMaker Savings Plans: Amazon SageMaker Savings Plans offer a way to save on your SageMaker costs by committing to a consistent amount of usage for a one or three-year period. You can select from compute savings plans which apply to the compute usage across SageMaker services, or you can choose SageMaker Instance Savings Plans which apply to a specific instance family within SageMaker. By committing to a certain level of usage, you receive a significant discount compared to on-demand pricing, with the savings percentage depending on the length of the commitment and the amount of usage to which you commit. The savings plans provide flexibility as your usage grows or changes and can lead to substantial cost reductions over time, especially for workloads with predictable and consistent usage patterns. More details about the savings plans can be found here.

Additional Pricing Benefits

Beyond the two pricing options – on-demand and savings plan – of SageMaker (which we'll delve into shortly), AWS has structured this payment model around paying solely for the amount used. The pay-as-you-go approach might help with cost efficiency in some ways, but be sure to read the “How to Optimize Amazon SageMaker costs” section of this post, so you can cut down costs as much as 60% on your current AWS bill. Whether you're engaged in model training, hosting, or other machine learning workflows, this flexible payment structure aligns with your specific utilization, allowing you to optimize costs and allocate resources judiciously based on your evolving needs.

Case Study: A Startup's Journey to Cost Optimization with SageMaker

Imagine an electric boat startup at the cusp of significant expansion into a new region. The company is grappling with escalating data analysis demands. As they look for a solution that works for them - they integrated SageMaker into their operations to address the surging data complexities. Initially opting for the most robust SageMaker configuration, they believed it would fortify their operations for future growth. However, they soon confronted the reality of the substantial monthly expenses incurred, which were unsustainable in the long run.

The CTO realized they needed a more budget-friendly SageMaker setup. They adjusted their deployment by optimizing node types, using reserved instances, and refining data queries to reduce reliance on expensive features like SageMaker Spectrum.

It can be different to minimize costs without compromising on machine learning capabilities. Here are three ways the company was able to cut AWS costs for SageMaker:

1. Choosing the Right Pricing Model:

They chose the correct pricing model. Evaluate whether on-demand pricing or reserved instances offer the best value for current and projected usage. For startups with predictable workloads, reserved instances can yield substantial savings over time. Since the boat company, knows it will continue to use at least a certain amount of AWS per month, it makes sense for purchasing reserved instances.

2. Effective Data Management:

Efficient data management plays a pivotal role in controlling costs related to storage. By regularly archiving obsolete data and streamlining unnecessary records can significantly reduce storage footprints and lower monthly bills.

3. Optimizing SageMaker Spectrum Usage:

Given that SageMaker Spectrum charges based on data scanned, it's imperative to structure queries for maximum efficiency. Implementing data partitioning and utilizing columnar formats can curtail the volume of data scanned, thereby optimizing costs.

By implementing these strategies, this startup can manage SageMaker expenses.

Tools and Tips for Cutting SafeMaker Costs

Utilizing SageMaker's built-in tools can provide insights into your spending patterns and help you identify areas where savings can be made. Below are three ways you can optimize your costs for SageMaker.

  1. Clean Up Unused Data: Regularly review and delete any unnecessary data stored in SageMaker to avoid incurring storage costs for data that is no longer needed. SageMaker's built-in tools can help you identify datasets that are not being used and can be removed.

  2. Leverage Spot Instances: Take advantage of Managed Spot Training in SageMaker for non-time-sensitive training jobs. Spot Instances can be significantly cheaper than on-demand instances and can lead to substantial savings, especially for larger or ongoing

  3. Use Amazon S3 for Efficient Data Storage: Store your datasets in Amazon S3 and manage them efficiently to take advantage of the lower storage costs compared to storing data directly in SageMaker. Utilize S3 lifecycle policies to move data to even more cost-effective storage classes, such as S3 Glacier for long-term archival, which can further reduce your expenses. By integrating S3 with SageMaker, you can streamline your data input and output for training and inference jobs while keeping costs under control.

  4. Set up Data Lifecycle Policies to Automate Data Management: Implementing data lifecycle policies within Amazon S3 can automate the process of data management, ensuring that your data is transitioned to the most cost-effective storage classes or even deleted when it is no longer required. S3 lifecycle policies can be set up to automatically move datasets to lower-cost storage tiers after a certain period of inactivity, or archive data to S3 Glacier for long-term storage at a reduced rate. By automating these transitions, you can save on storage costs without manual intervention, allowing for a more efficient and cost-effective data management strategy within your SageMaker environment.

  5. Use Pump to Cut Your AWS Bill by 60%: Consider utilizing Pump to save up to 60% off AWS for early stage startups, and for completely free (yes, that's right). Pump is the only company with AI automated savings on 12 AWS services built in, including SageMaker. Have questions? Go to Pump’s website to learn more.


With a clear understanding of the SageMaker pricing model, and by employing strategies for cost optimization, you can make SageMaker a cost-effective part of your machine learning workflow.

Remember to start with SageMaker's free tier to gain familiarity with the service without initial investment. As you scale, keep a close eye on your SageMaker pricing analysis to ensure that you're getting the most out of your AWS budget. Whether it's by selecting the appropriate instance type or by carefully planning your SageMaker pricing strategies, the goal is to maximize value while minimizing costs.

Similar Blog Posts

Join Pump for Free

If you found this post interesting, consider checking out Pump, which saves up to 60% off AWS for early stage startups, and for free (yes, that's right). Are you ready to take control of your Amazon Redshift costs and all other AWS spending? Learn more here.

1550 Mission Street, San Francisco, CA 94103

Made with

in San Francisco, CA

© All rights reserved. Pump Billing, Inc.

1550 Mission Street, San Francisco, CA 94103

Made with

in San Francisco, CA

© All rights reserved. Pump Billing, Inc.

1550 Mission Street, San Francisco, CA 94103

Made with

in San Francisco, CA

© All rights reserved. Pump Billing, Inc.