Claude, the developer of a chatbot, is looking for specialists to join its AI chip development team, as reported by Business Insider. “This position is intended for someone who has already brought chips to mass production, realistically assesses timelines, and is ready to make independent decisions with serious consequences while working in a small team,” the job description states.

As a spokesperson for Anthropic explained to the publication, the startup intends to design hardware and models so that Claude can operate faster and more efficiently “at the scale required by customers.” At the same time, he emphasized that Anthropic will continue to adhere to “an approach that involves using multiple types of chips”: hardware from Amazon Web Services, Google, Nvidia, and AMD will remain a key part of the AI lab’s scaling plans.

Context

Anthropic currently leases computing power from various providers. The startup has a 10-year contract worth more than $100 billion with Amazon Web Services, including up to 5 gigawatts of capacity on Amazon’s proprietary chips—Trainium and Graviton. The AI startup will also spend about $200 billion over five years to lease servers from Alphabet running on Google’s TPU chips, which are developed by Broadcom. Through this arrangement, Anthropic plans to lease up to 3.5 gigawatts of computing power.

In addition, the startup has committed to purchasing $30 billion worth of cloud services from Microsoft Azure and reserving up to 1 gigawatt of infrastructure on Nvidia systems. The company is also leasing all available capacity of SpaceXAI’s Colossus 1 supercomputer—more than 300 megawatts—through May 2029. The potential value of the contract is approximately $45 billion.

Anthropic also has a multi-year agreement with CoreWeave, the financial terms of which have not been disclosed. Under the agreement, the company has leased servers equipped with Nvidia AI chips.

On August 4, Bloomberg sources reported that Anthropic had signed a $10 billion deal to lease computing power from the infrastructure startup Volta Infra Holdings, which was founded with Nvidia’s support just a few months ago.

Sources told Reuters in April that Anthropic was exploring the possibility of developing its own chips. And in July, the industry publication The Information reported that the company was in talks with Samsung Electronics, considering it as a potential manufacturing partner.

Developing a chip in-house is a costly and complex process, but it could help AI companies reduce their dependence on Nvidia, which dominates the graphics processing unit market, notes Business Insider. In addition, having their own product allows AI labs to tailor their computing hardware to the needs of specific models.

In June, an OpenAI competitor unveiled the Jalapeño specialized chip, developed in collaboration with Broadcom. Meta also plans to begin production of its next-generation AI chip in September, according to an internal memo reviewed by Reuters. The CEO of the French AI startup Mistral said in May that the company is considering developing its own processors.

This article was AI-translated and verified by a human editor

#Anthropic

#Open AI

#Broadcom

#Google