What Is Amazon Bedrock?

Amazon Bedrock is a fully managed AWS service for building generative AI applications. It offers foundation models from leading AI companies through an API, allowing users to access them without having to manage underlying infrastructure.

Written by Ellen Glover
A closeup photo of a smartphone resting on a computer keyboard, and on the smartphone screen it reads the Amazon Bedrock homepage
Image: Shutterstock
UPDATED BY
Brennan Whitfield | Sep 24, 2025
REVIEWED BY
Ellen Glover | Sep 24, 2025
Summary: Amazon Bedrock is an AWS service that provides a choice of foundation models from leading AI companies, as well as Amazon’s own models, to help developers build and scale generative AI applications. It’s a managed service that aims to simplify access to large language models.

Amazon Bedrock is a fully managed service for building generative AI applications, offered on cloud computing platform Amazon Web Services (AWS). The Amazon Bedrock platform acts as a kind of marketplace, allowing businesses to pick which AI models they would like to build their generative AI apps on top of and customize with their own proprietary data.

What Is Amazon Bedrock?

Amazon Bedrock is an AWS service for building generative AI applications. The platform allows companies to choose which AI models they would like to build their apps on top of, and then customize them with their own proprietary data through Amazon Web Services (AWS).

“Amazon Bedrock is leading the industry to a new place,” Sherry Marcus, director of Applied Science at AWS, told Built In. “A place where customers can easily choose from different high-performing [models] from leading AI companies, get access to a broad range of capabilities they need to build generative AI applications, and simplify the development while maintaining privacy and security.”

Related5 Ways Generative AI is Changing the Job Market

 

What Is Amazon Bedrock?

Amazon Bedrock — also known as AWS Bedrock — is an AWS service that enables developers to create their own generative AI applications by giving them access to the foundation models of Amazon and AI startups. Foundation models are large language models that are trained on massive volumes of data and can serve as the cornerstones of new applications. Through AWS Bedrock, developers can choose between foundation models from companies like AI21 Labs, Anthropic, DeepSeek, Meta and Stability.ai. 

The idea is to provide developers with the AWS cloud as an environment for building and hosting their generative AI applications. This speeds up the application development process and makes generative AI technology more accessible to a wider audience. Developers simply combine their data with the model of their choice to customize their application. As a result, any user can quickly craft chatbots, text generators and image generators, among other tools.

Companies big and small are eager to implement generative AI into their own business operations as a way to enhance productivity and cut costs. While Microsoft-backed OpenAI and Google have raced out to a fast start, Amazon Bedrock is Amazon’s biggest move yet into this rapidly emerging sector and promises to reshape the generative AI landscape.

RelatedJeff Bezos: My Secret to Amazon’s Success

 

What Makes Amazon Bedrock Different From Competitors?

Choice and Flexibility

The element of choice makes Amazon Bedrock different from pretty much any other generative AI tool today. While other companies like Microsoft and Google offer access to only their own models, AWS Bedrock allows companies to pick from a variety of foundation models made by AI companies like AI21 Labs, Anthropic, Cohere, DeepSeek, Meta, Stability AI and, of course, Amazon itself. As Marcus put it, this democratizes access to generative AI technology. 

“What we like about Amazon Bedrock is they don’t restrict to one single model, or a set of specific models from a specific vendor,” Jeff Reihl, CTO of legal tech company and Amazon Bedrock customer LexisNexis, told Built In. “It gives us, as a user, a lot of options because we’re going to use the best model to solve the customer’s problem at the best price and the best performance.” 

Security

Because Amazon works with all of these foundation model vendors directly, they actually manage the model on behalf of those companies within their own ecosystem. So if a business uses one of the models, all of the proprietary data they plug in stays within AWS, and it is managed and protected to Amazon’s standards. This can give companies extra peace of mind when it comes to data governance and compliance with any existing privacy legislation, according to Travis Rehl, senior vice president of product and services at cloud company and AWS partner Innovative Solutions.  

AWS Bedrock also offers encryption features and HIPAA eligibility, making it useful for a variety of contexts. In contrast, neither ChatGPT nor Gemini are inherently HIPAA compliant. Both have sparked security concerns as well. 

Personalization

Because companies can feel more comfortable using their own data in these models, they can create more personalized generative AI products and services that are helpful to their unique needs. Leveraging these models and their own data, businesses can create AI agents that automatically execute tasks, like processing insurance claims or creating ad campaigns, without compromising the security of their sensitive data.

While OpenAI has given users the ability to customize ChatGPT, both ChatGPT and Gemini already come with their own data and limitations. There may be some room for flexibility, but it’s not the same as having different models to choose from.

 

Benefits of Amazon Bedrock 

Customization 

AWS Bedrock places the foundation models of Amazon and a range of startups at the fingertips of developers, who can then select the models that suit their needs and equip them with their own company data for a more personalized experience.    

Accuracy and Performance

Because developers can integrate their company data with AWS Bedrock’s foundation models, they can equip AI generative tools with knowledge related to their companies and industries, which can then produce more accurate and relevant responses.    

Data Security 

Amazon Bedrock follows data security best practices like encrypting all data, refraining from training foundation models on user input and preventing third parties from accessing this data. Developers can further bolster their security with features like HIPAA eligibility.  

 

Drawbacks of Amazon Bedrock

Data Privacy 

Despite data security features offered by AWS Bedrock, the creation and use of generative AI apps that store user data can be a cause for concern. Users may not feel comfortable with their data being collected, especially if any sensitive information is involved.  

Ethical Dilemmas

Amazon Bedrock’s foundation models are created by humans and further trained by humans, leaving room for biases to creep in. Developers and companies need to take extra steps to ensure unconscious biases don’t affect the development and behavior of generative AI tools.    

Potential Monopoly Concerns

The nature of AWS Bedrock as a foundation model marketplace means this service can expand to include other companies. If competition isn’t stiff enough, Amazon’s generative AI presence may grow too large and monopolize resources in the field.

 

What Can You Build With Amazon Bedrock? 

The inherent flexibility of foundation models and services provided by Amazon Bedrock makes it easy to build generative AI products for a wide variety of tasks, powering everything from search to content creation.

Chatbots

AWS Bedrock can help companies build their own chatbots, or virtual assistants that understand user requests, break down tasks, engage in conversations to collect information and take actions to fulfill those requests.

For instance, data analytics and risk assessment firm Verisk uses Amazon Bedrock to power its AskMax chatbot, which gives users quick contextual answers to support their questions, potentially eliminating hours of manual search. 

Search Engines

Generative AI-powered search engines can find and synthesize relevant information to answer questions based on a large corpus of data.

LexisNexis created its own search engine, called Lexis+ AI, which allows users to easily search for legal and regulatory documents. It also has a conversational AI component, which means it can carry on conversations with users and refine their queries, Reihl said. 

Text Summarization Tools

Amazon Bedrock helps companies create tools for text summaries, providing abridged versions of articles, books, blog posts and any other writing. 

LexisNexis also incorporated this capability into its Lexis+ AI so that users don’t have to read hundreds of pages of legal documents on their own. 

Text Generation Tools

Generative AI is perhaps best known for its ability to create new pieces of original content, such as blog posts, social media posts and marketing emails.

Web hosting company GoDaddy uses AWS Bedrock to help its customers accelerate their content creation, helping customers fill out their online business pages and more efficiently connect with relevant suppliers, consumers and funding opportunities.

Image Generation Tools

Another popular application of generative AI is image generation, which automatically creates artistic (and sometimes photorealistic) images or animations. These can be helpful in creating presentations, websites and ad campaigns. That’s how online travel booking site Booking.com uses Amazon Bedrock.

 

Amazon’s Growing Investment in AI

Looking ahead, the generative AI market is expected to eclipse $1 trillion by 2028, according to an analysis from Morgan Stanley. And skyrocketing demand for new generative AI products — powered by new products and assistants that accelerate their creation — gives cloud providers like AWS, Microsoft and Google an opportunity to claim a substantial piece of that pie, as businesses shift more workloads to the public cloud.

In fact, generative AI could increase the ROI of cloud computing platforms by as much as 110 percent. As the largest cloud provider in the world by market share, AWS can combine its cloud computing prowess with AWS Bedrock to gain a stronger foothold in the generative AI space. 

Amazon has also developed its own chips made specifically for training AI models with less computing power, a move that will likely help the company compete with Nvidia, Microsoft and Google. Since March 2024, Amazon’s chips have been used by Anthropic — a top competitor to OpenAI — to build, train and deploy its foundation models, opening a major new front in the ongoing AI arms race.

Still, generative AI is a bit of a “Wild West,” as Reihl put it. It can be a threat to data privacy, produce biased results and infringe on copyright and other laws. Amazon has been a vocal supporter of responsible AI, making voluntary commitments toward AI safety at the behest of the White House and speaking at the United Nations. But it remains to be seen how any future AI regulation will affect Amazon and other industry giants.

 

Notable Amazon Bedrock Developments

Below are a few key milestones and features of Amazon Bedrock, highlighting its evolution from an initial preview to a robust, fully managed service for building generative AI applications.

Prompt Caching for Amazon Bedrock General Availability Launch (April 2025)

Prompt caching for Amazon Bedrock became generally available in April 2025. The prompt caching capability can reduce costs by up to 90 percent and latency by up to 85 percent by caching frequently used prompts across multiple API calls. This development was significant for enterprises seeking to process requests faster, improve efficiency and manage operational costs as they scale their use of large language models. 

Multi-Agent Collaboration for Amazon Bedrock General Availability Launch (March 2025)

Multi-agent collaboration on Amazon Bedrock was made generally available in March 2025. This feature, pioneered by Amazon Bedrock, allows organizations to orchestrate specialized agents to handle complex, multi-step tasks across their systems and data, driving greater efficiency and accuracy. By enabling specialized agents to work together, this capability moves beyond simple single-task agents and into more complex, goal-driven automation, which is crucial for enterprise-level adoption.

Guardrails for Amazon Bedrock General Availability Launch (April 2024)

Guardrails for Amazon Bedrock became generally available in April 2024, enabling customers to implement safeguards across large language models based on their specific use cases and responsible AI policies. Guardrails is an AWS feature that helps to filter out harmful content and ensure that AI applications remain secure and aligned with a company’s safety standards. This capability is critical for promoting responsible AI development and increasing enterprise trust and adoption of generative AI technologies.

Amazon Bedrock General Availability Launch (September 2023)

Amazon Bedrock officially became generally available to all AWS customers on September 28, 2023. This announcement marked Amazon Bedrocks’s transition from a limited preview to a fully managed, production-ready offering. Its launch was also a major step in democratizing access to generative AI, allowing customers to easily build and scale AI applications without needing to manage the underlying infrastructure.

Frequently Asked Questions

Amazon Bedrock is a fully managed AWS service for building generative AI applications. The service allows companies to choose which AI models they would like to build their apps on top of, and then customize them with their own proprietary data through Amazon Web Services.

Yes, Amazon Bedrock became generally available in September 2023.

Amazon Bedrock was first launched in preview in April 2023 as part of a set of new tools for building generative AI on AWS. It was then made generally available in September 2023.

Matthew Urwin contributed reporting to this story.

Explore Job Matches.