Amazon.com Inc. will invest as much as $4 billion in Anthropic, bagging a crucial partner in its effort to become a major player in generative artificial intelligence and offering a vote of confidence in the hot startup.
As part of the deal, Anthropic will move most of its software to Amazon Web Services data centres, and use the cloud computing company’s homegrown chips to train the models it uses to power chatbots and other applications. Besides getting access to Amazon’s computing power, Anthropic will gain a financial infusion that will help it pay the huge costs required to train and run massive AI models. Amazon will have a minority position in Anthropic, the companies said in a statement on Monday.
The e-commerce and cloud computing giant has long taken stakes in partners in areas it deems a priority, including cargo airlines, a grocery distributor and an electric truck manufacturer. But if the investment in Anthropic lands anywhere near $4 billion, it would represent the largest known piece of corporate dealmaking directly related to AWS. The unit has tended to build its own products rather than relying on technology or businesses purchased from others, making relatively modest acquisitions in a market of hype-fueled, billion-dollar valuations. Amazon says its engineers, including those who work outside of AWS, will have access to Anthropic’s models.
AWS is the world’s largest seller of on-demand computing power and data storage. But it has been widely seen as a laggard in the emerging field of computer models trained to generate text, images and other content because it has lacked both a hit product and a high-profile exclusive partner in the arena. ChatGPT creator OpenAI runs its software on Microsoft Corp.’s data centers, a partnership based on a $13 billion investment that has sparked a renewal for Amazon’s Seattle-area neighbour.
Amazon executives have said that it’s early days for so-called generative AI, and that more than 100,000 customers have used the company’s machine-learning tools to date. Anthropic’s foundational model, called Claude, was already available as part of an Amazon service called Bedrock, still in early release, that makes both Amazon and third-party models available to its customers.
“We have tremendous respect for Anthropic’s team and foundation models, and believe we can help improve many customer experiences, short- and long-term, through our deeper collaboration,” Amazon CEO Andy Jassy said in a statement.
The deal also represents a milestone moment for Amazon’s in-house chipmaking effort, which includes processors called Trainium and Inferentia that are designed to power machine-learning applications. Most cutting-edge artificial intelligence applications rely on pricey chips built by Nvidia Corp. that can be hard to come by. Anthropic will use AWS chips to build and train future foundational models, the companies said.
Founded by OpenAI veterans, Anthropic has raised more than $1 billion to date with a pitch to make a safer kind of chatbot for such tasks as summarising, searching, answering questions and coding. The company’s backers include Alphabet Inc.’s Google, which invested almost $400 million in Anthropic, Bloomberg reported in February. Google is also an AWS cloud rival and builder of models used in generative AI.
“By significantly expanding our partnership, we can unlock new possibilities for organisations of all sizes, as they deploy Anthropic’s safe, state-of-the-art AI systems together with AWS’s leading cloud technology,” Anthropic co-founder and CEO Dario Amodei said.
