NVIDIA invests $2B in CoreWeave, targets 5 GW AI data centre capacity by 2030

Tech giant NVIDIA announced a $2 billion equity investment in AI cloud-computing firm CoreWeave, marking another deal in the world of artificial intelligence (AI) infrastructure partnerships. 

The deal sees NVIDIA buy Class A common stock and deepen a technical and commercial alignment intended to accelerate the construction of specialised data centres or “AI factories”, with CoreWeave targeting more than 5 gigawatts of capacity by 2030. 

“AI is entering its next frontier and driving the largest infrastructure buildout in human history,” said Jensen Huang, Founder and CEO of NVIDIA. 

He added, “CoreWeave’s deep AI factory expertise, platform software, and unmatched execution velocity are recognised across the industry. Together, we are racing to meet extraordinary demand for NVIDIA AI factories—the foundation of the AI industrial revolution.”

The arrangement combines direct capital, closer product integration, and mutually exposed demand for hardware and cloud services, signalling a move beyond simple supplier-customer relationships.

The companies will deepen coordination across infrastructure and software, where NVIDIA will assist CoreWeave in securing land, power, and data centre shell, and CoreWeave’s software will be tested and validated for closer interoperability and possible inclusion in NVIDIA reference architectures.

CoreWeave will deploy multiple generations of NVIDIA hardware across its infrastructure stack. The collaboration also aims to extend the availability of CoreWeave’s software to global cloud service providers and enterprise customers.

This transaction is part of an industry concept increasingly common in the field: the capital-plus-compute partnership. In these structures, a technology vendor supplies funds or takes equity in a cloud operator or model developer, while securing long-term product placement, integration and, in some cases, guaranteed offtake for compute capacity.

Several recent deals follow the same pattern. Microsoft and NVIDIA pledged substantial investment tranches into Anthropic, which committed to multi-year, high-value cloud purchases, a reciprocal arrangement tying capital to compute and distribution. 

Around the same period, Anthropic expanded its use of Google Cloud TPUs on a very large scale, highlighting that model developers will often stitch together multiple deep partnerships to secure capacity and price performance.

Meta offers a parallel, but internally focused approach. Rather than relying primarily on external cloud partners, Meta has accelerated a vast internal infrastructure programme to support generative AI research and products, committing to large GPU inventories and multi-gigawatt data centre projects. This is another industry response: vertical integration through in-house build-out when firms view control of compute as a strategic necessity.

Earlier this month, ChatGPT creator OpenAI announced a multi-year partnership—the deal, worth over $10 billion—with AI chipmaker Cerebras to enhance its computing infrastructure.