Amazon Web Services Inc. and Databricks Inc. say they have signed a five-year deal to help companies cut costs when building and running their artificial intelligence applications.
As part of the deal, Databricks says it will use AWS’s Trainium AI chips to power its Mosaic AI service, which helps businesses to customize existing large language models or build their own from scratch. Amazon says the main benefit for customers is that they’ll pay a lot less to use its chips, as opposed to Nvidia Corp’s graphics processing units, which are the most popular hardware for AI workloads today.
Databricks’ AI service stems from its $1.3 billion acquisition of a company called MosaicML Inc. last year. The Mosaic AI service includes tools for model serving, which support a variety of third-party LLMs, including a selection of models available through Amazon Bedrock.
The companies explained that Databricks’ customers will be able to scale model training on Mosaic AI at lower costs when using the AWS Trainium chips. In addition, they’ll provide other capabilities to help their joint customers build, deploy and monitor custom AI applications, without sacrificing control over their data or intellectual property.
In addition, the deal will see Databricks and AWS work together to make it easier for customers to run Databricks’ big data services on AWS via new integrations on the AWS Marketplace, including simplified onboarding, configuration and serverless compute from AWS.
Databricks said it will work with its system integration partners to create additional technical solutions and implementation resources designed to help customers identify generative AI use cases and on-premises workloads that can be optimized with Databricks on AWS.
According to Databricks, the partnership intends to make it faster and cheaper for businesses to design and build AI applications, primarily from the savings derived from using Amazon’s AI chips.
“Strengthening our collaboration with AWS allows us to provide customers with unmatched scale and price-performance so they can bring their own generative AI applications to market more rapidly,” said Databricks co-founder and Chief Executive Ali Ghodsi.
The deal should help both companies in their efforts to compete with rivals such as Microsoft Corp., Snowflake Inc. and Salesforce Inc., which are all vying with them to court businesses’ AI dollars. The most successful AI initiatives tend to be those that rely on businesses’ internal data to customize AI applications. For instance, a company might create a bespoke AI chatbot that’s fueled with knowledge from its internal databases.
Amazon’s strategy is to position itself as a neutral provider of AI technology, providing customers with access with a variety of AI models and the infrastructure to run them.
AWS CEO Matt Garman said the collaboration with Databricks will enable companies to build AI applications powered by valuable insights from their own data. “We’re helping customers innovate faster by focusing on what truly matters most for their business,” he said.
Databricks’ derives its revenue from its data analytics services, AI tools and other software that can tap into AI-ready data, helping companies to build various kinds of AI applications.
The two companies have an existing partnership, where customers can run Databricks’ data tools and services on AWS.