CoreWeave has signed a new multi-year agreement with Anthropic to supply cloud computing capacity for the AI startup’s Claude models, in another sign of how fiercely companies are competing for the infrastructure needed to build and run advanced artificial intelligence systems. The deal will bring new computing capacity online later in 2026 and initially focus on a phased rollout, with room to expand over time. Investors reacted strongly to the announcement, sending CoreWeave shares up more than 13%.
The agreement matters because access to computing power has become one of the main bottlenecks in the AI industry. Training and operating large language models requires enormous amounts of specialized hardware, especially high-end chips and cloud infrastructure. The Anthropic deal is the latest in a fast-growing list of major contracts CoreWeave has signed as demand for AI compute continues to accelerate across the sector.
For CoreWeave, the new partnership is important not only because it adds another high-profile customer, but also because it helps diversify the company’s revenue base. Seemingly, Microsoft accounted for about 67% of CoreWeave’s revenue last year, making the company highly dependent on a single customer. Meta has recently become another major client, and the Anthropic agreement gives CoreWeave another large name in its customer mix at a time when investors are watching closely to see whether it can broaden its business beyond a few dominant partners.
Also CoreWeave is on a recent growth spree. The company signed an $11.9 billion agreement with OpenAI last year, a $6.3 billion initial order with Nvidia in September, and an expanded $21 billion deal with Meta just one day before the Anthropic announcement. Those numbers suggest CoreWeave is rapidly becoming one of the most important specialized cloud providers in the AI economy, especially for companies that need access to advanced computing resources without building the full infrastructure stack themselves.
Anthropic’s side of the deal is just as revealing. The startup has been rushing to secure more capacity from several directions. Earlier in the same week, Anthropic signed a separate arrangement involving Broadcom and Google for AI compute, and Anthropic was exploring the possibility of designing its own chips. Together, those moves show that Anthropic is trying to avoid depending too heavily on any one infrastructure source while preparing for continued growth in demand for Claude.
That broader pattern says something important about the current phase of the AI race. Competition is no longer centered only on who has the best model or the most visible chatbot. It is increasingly about who can secure the underlying hardware, power, and cloud access needed to keep scaling. CoreWeave, the “neocloud,” has benefited from this shift by positioning itself as a supplier of hardware and cloud capacity to other technology firms. Its close relationship with Nvidia has helped make it a key channel for prized AI chips that remain in extremely high demand.
In practical terms, the Anthropic agreement strengthens both companies. It gives Anthropic more room to run and expand its Claude family of models, and it gives CoreWeave greater credibility as a major infrastructure provider in a crowded but fast-growing market. While the financial terms were not disclosed, the announcement still sent a clear message to investors: AI demand remains strong, and companies that can provide the compute behind that boom are becoming more strategically valuable.
Overall, the deal reflects a deeper reality in artificial intelligence today. Success depends not just on software breakthroughs, but on securing the expensive physical systems that make those breakthroughs possible. CoreWeave’s latest agreement with Anthropic shows how cloud infrastructure providers are becoming some of the most important power brokers in the AI economy.
