Amazon.com said it will invest up to $4 billion in cash in the high-profile startup Anthropic, in its effort to compete with growing cloud rivals on artificial intelligence.
Amazon's employees and cloud customers will gain early access to technology from Anthropic as part of the deal, which they can infuse into their businesses. The San Francisco-based startup also committed to rely primarily on Amazon's cloud services, including training its future AI models on large quantities of proprietary chips it would buy from the online retailing and computing giant.
In a joint interview, the CEOs of Amazon's cloud division and Anthropic said the immediate investment will be $1.25 billion, with either party having the authority to trigger another $2.75 billion in funding by Amazon.
They declined to state how much Amazon now would own of Anthropic or the startup's updated valuation, last estimated at more than $4 billion. Amazon said it would not gain a board seat and that its stake amounted to a minority position.
The news represents perhaps Amazon's biggest answer yet to challenges from Microsoft and Alphabet's Google, smaller cloud rivals that have marketed or developed powerful AI this year. The deal also shows ongoing maneuvering by the cloud companies to secure ties with AI startups reshaping their industry.
Since 2019, Microsoft has put billions of dollars into its partnership with ChatGPT's creator OpenAI, giving its customers special access to the startup's prose-writing, image-generating technology.
Google meanwhile helped pioneer this branch of AI and in May invested in Anthropic's $450-million fundraise, in a relationship the startup said is not going away.
For Amazon, Monday's deal may spell an uptick in demand, including for chips powering AI. Anthropic agreed to work on developing technology for Amazon's in-house Trainium and Inferentia chips, for instance.
Adam Selipsky, Amazon Web Services's CEO, said the pact "will help make Anthropic's models better, will help make our chip technology and our AI infrastructure better".
Dario Amodei, Anthropic's CEO, said his company "has obtained the money in a way that allows it to prioritise safety" and "allows us to continue to scale up our models".
STILL TIED TO GOOGLE
Anthropic, founded by former OpenAI executives including Amodei, is one of a series of companies building so-called generative AI, systems that can draft content as if a human created it. Anthropic has aimed to distinguish its work by training AI to adhere to moral values.
Amodei declined to state if its latest financing was competitive.
Anthropic is continuing an agreement it announced with Google in February, he said. Anthropic is using Google's custom chips and has planned to make its technology available via Google Cloud and elsewhere.
Yet with Monday's deal, Anthropic is giving a boost to Amazon Bedrock, a service that has attracted thousands of users to start building AI applications.
Amazon's customers will gain features from Anthropic early, such as the ability to customise their AI. Selipsky said, "both companies are committing that, for many years to come, future versions of Claude will be available on Amazon Bedrock, and that's important."
Anthropic's Claude 2 is an AI model that is able to respond to particularly large prompts, setting it up to analyze long business or legal documents.
Amodei said the deal "allows us to work more closely to drive enterprise usage for Claude," which he said represented much of the demand on Bedrock so far.
LexisNexis, a data analytics company, is working with Anthropic and Amazon to make its own legal search capabilities more "intelligent," Amodei said. Other customers include Bridgewater Associates and Lonely Planet, he said.
Anthropic has yet to gain the name recognition or usage of OpenAI, the startup behind the GPT-4 model and ChatGPT, which is one of the fastest-growing software applications in history.
Amazon has aimed to give customers a broad range of AI models so they have little reason to look afield for cloud services.
Asked if Amazon would invest in additional AI startups beyond Anthropic, Selipsky said, "I honestly don't know what the future will hold."