Amazon Partners with Databricks to Challenge NVIDIA with Cost-Saving AI Chips
Amazon.com Inc. has entered into a significant five-year agreement with Databricks, a key competitor to Snowflake, to provide access to its Trainium AI chips. This strategic partnership offers an alternative to the dominant NVIDIA GPUs, presenting a more cost-effective solution for companies developing AI models.
Under the terms of the agreement, although financial specifics were not disclosed, Amazon aims to reduce AI development costs. Both Amazon and Databricks see this collaboration as a way to accelerate and economize AI integration within enterprises, offering a competitive edge.
Naveen Rao, a Databricks executive, emphasized that the partnership would allow Databricks to pass on the savings from using Amazon's AI chips directly to their customers. This strategy could challenge NVIDIA’s stronghold in the AI accelerator market, potentially disrupting the current dynamics.
Despite this new venture, Databricks will continue to utilize NVIDIA GPUs rented through Amazon Web Services (AWS), which remains a significant growth sector for the company, reportedly generating over $1 billion in revenue.
AWS representative Dave Brown highlighted that the Trainium chips, specifically designed for AI applications, could lower companies' AI development expenses by up to 40%. This reduction could drive more businesses to adopt Amazon’s AI solutions.
In November 2023, Amazon released the second version of its Trainium chips, continuing its push into AI infrastructure with a range of Inferencia chips designed to build and deploy AI models efficiently.