CoreWeave inks Anthropic deal
CoreWeave signed a multi‑year agreement to run Anthropic’s Claude at production scale, and the company says nine of the ten largest AI model providers are now on its GPU platform — a sign specialist GPU clouds are cementing market roles alongside hyperscalers. That deal suggests frontier model developers are diversifying infrastructure to get faster GPU access and avoid being wholly dependent on the biggest public clouds. (thenextweb.com)
CoreWeave just added Anthropic as a customer, with a multi-year deal to run Claude at production scale and bring the first compute online later in 2026. That means one of the companies behind the most widely used artificial intelligence assistants is buying capacity from a specialist cloud instead of relying only on the biggest public-cloud vendors. (coreweave.com) CoreWeave is not Amazon Web Services, Microsoft Azure, or Google Cloud. It is a cloud company built around graphics processing units, which are the chips artificial intelligence systems use the way a video editor uses a render farm: lots of parallel machines chewing through huge jobs at once. (coreweave.com) Anthropic makes the Claude family of models, and those models need two kinds of computing power. One kind trains new systems on giant piles of data, and the other kind serves live answers to users, which is called inference and is the part people feel when they open a chatbot and wait for a response. (anthropic.com) (coreweave.com) The squeeze in this market has never really been software first. It has been access to enough graphics processing units, in enough racks, with enough electricity and networking, to keep models available when millions of people ask questions at the same time. (coreweave.com) (openai.com) That is why this deal landed two days after another giant CoreWeave announcement. On April 9, 2026, CoreWeave said Meta expanded its agreement to about $21 billion through December 2032, with capacity spread across multiple locations and some early deployments of Nvidia’s Vera Rubin platform. (coreweave.com) CoreWeave is also already deep inside OpenAI’s supply chain. In March 2025, it announced an OpenAI infrastructure contract worth up to $11.9 billion, and in October 2025 it said the total value of its OpenAI agreements had risen to about $22.4 billion. (coreweave.com 1) (coreweave.com 2) Put those names together and the pattern gets clearer. CoreWeave said on April 10, 2026, that nine of the ten largest artificial intelligence model providers now run on its platform, which suggests the market is settling into a split where giant model labs use hyperscalers and specialist graphics processing unit clouds at the same time. (coreweave.com) Anthropic has been moving the same way on its side. On March 31, 2026, it said it was expanding work with Google and Broadcom for multiple gigawatts of next-generation compute, so the new CoreWeave deal fits a broader strategy of lining up several infrastructure partners instead of betting everything on one pipe. (anthropic.com) CoreWeave has been growing fast enough to make that pitch credible. In results released on February 26, 2026, the company said it reached $5 billion in annual revenue in 2025 and reported a record backlog, which is the pile of contracted business it expects to deliver later. (coreweave.com) The simplest way to read this is that artificial intelligence labs are shopping for compute the way airlines shop for gates at busy airports. If one cloud is full, too expensive, or too strategically important to depend on alone, the lab wants another runway ready before the traffic spike hits. (coreweave.com) (anthropic.com)
