Keypoints
- CoreWeave (CRWV) shares closed at $102.00, up roughly 10.8%, following the announcement of a multiyear partnership to provide “production scale” AI cloud services to Anthropic.
- Anthropic recently secured 3.5 gigawatts (GW) of Google’s Broadcom-made TPUs and is reportedly exploring the design of its own custom semiconductors.
- Meta (META) expanded its own relationship with CoreWeave on Thursday, signing a $21 billion agreement for AI capacity through 2032, which will include the first deployments of Nvidia’s Vera Rubin system.
- The competitive “AI chip crunch” has led OpenAI to partner with Broadcom for 10GW of custom chips, while Microsoft recently debuted its Maia 200 processor in January 2026.
Main Story
The race for artificial intelligence supremacy has shifted from model development to infrastructure security. On Friday, April 10, 2026, CoreWeave, the AI-focused cloud provider that went public in 2025 announced a strategic agreement to power Anthropic’s next generation of Claude models.
The deal validates CoreWeave’s “specialized cloud” model as an alternative to hyperscalers, sending its stock to a daily high of $105.90 before settling at $102.00.
This partnership follows a massive week for the AI hardware sector. Just 24 hours prior, Meta committed an additional $21 billion to CoreWeave, bringing their total partnership value to roughly $35 billion. As the demand for compute remains insatiable, companies are diversifying their hardware stacks.
Anthropic now utilizes a blend of AWS Trainium, Google TPUs, and Nvidia GPUs via CoreWeave, while simultaneously working with Broadcom to develop the “multi-gigawatt” infrastructure necessary to keep pace with a revenue run-rate that has reportedly surpassed $30 billion.
The Issues
The primary challenge for the industry is the escalating cost of compute and the “gigawatt gap.” AI firms must solve the problem of hardware dependency on Nvidia; hence the aggressive pivot toward custom silicon (OpenAI’s 10GW Broadcom deal and Meta’s MTIA 400). However, building custom chips is a multi-year endeavor, leaving a supply-demand mismatch in the interim. Furthermore, CoreWeave faces the challenge of financing its rapid expansion, having recently increased its convertible note offering to $3.5 billion to fund the construction of the 32 data centers required to fulfill its $66 billion contract backlog.
What’s Being Said
- “AI is no longer just about infrastructure, it’s about the platforms that turn models into real-world impact,” stated Michael Intrator, CEO of CoreWeave.
- Anthropic’s CFO Krishna Rao noted that the company is making its “most significant compute commitment to date” to serve a customer base that has doubled in just two months.
- Scott Guthrie, Microsoft’s EVP of Cloud + AI, described the new Maia 200 chip as “the most performant first-party silicon from any hyperscaler,” directly challenging Nvidia’s dominance.
- Market analysts at Dow Jones suggest that CoreWeave’s high debt levels remain a concern for some investors, despite the company’s massive revenue projections of $12 billion for 2026.
What’s Next
- Anthropic is expected to begin a phased rollout of its CoreWeave-powered infrastructure later in 2026, with the option to scale as Claude model demand increases.
- The Nvidia Vera Rubin platform will see its first major deployments via the Meta-CoreWeave partnership starting in early 2027.
- Amazon is tipped to begin selling its proprietary AI chips in large server racks to third-party customers, following CEO Andy Jassy’s latest shareholder signal.
- OpenAI and Broadcom are scheduled to begin deploying their first racks of custom-designed 10GW accelerators in late 2026 or early 2027.
Bottom Line
CoreWeave has positioned itself as the “essential utility” of the AI era. By locking in both Anthropic and Meta within a single week, it has consolidated a significant portion of the world’s most valuable AI workloads, even as those giants move toward building their own custom processors.
