NVIDIA Expands CoreWeave Alliance With $2B Investment and 5 GW AI Factory Target

Core announcement

In a January 26, 2026 press release, NVIDIA and CoreWeave said they are expanding their partnership to accelerate construction of more than 5 gigawatts of AI factories by 2030. The announcement combines infrastructure scale targets, software integration plans, and direct balance-sheet support from NVIDIA. For enterprises and cloud buyers, this is notable because it links hardware roadmap access, data center development, and operational tooling in one coordinated package rather than separate vendor contracts.

NVIDIA also said it invested $2 billion in CoreWeave Class A common stock at a purchase price of $87.20 per share. In the release, both companies framed the investment as a signal of long-term alignment around AI cloud capacity expansion. The document states that the relationship will deepen across infrastructure, software, and platform layers, with the stated goal of meeting rapidly growing demand for AI compute.

What the companies plan to do

• Build AI factories operated by CoreWeave using NVIDIA accelerated computing technology.
• Use NVIDIA’s financial strength to speed procurement of land, power, and shell capacity.
• Test and validate CoreWeave software components, including SUNK and Mission Control, for deeper interoperability and possible inclusion in NVIDIA reference architectures.
• Deploy multiple NVIDIA generations through early use of Rubin platform systems, Vera CPUs, and BlueField storage systems.

This matters because AI infrastructure constraints are increasingly physical and operational, not only model-related. If land acquisition, interconnection, and power delivery slip, compute expansion slips. The release explicitly addresses those bottlenecks by pairing capital with pre-integrated technical stacks. It also suggests that future enterprise procurement may favor providers that can present verified reference architectures plus predictable deployment schedules, not only peak benchmark claims.

Important caveat

The same release includes extensive forward-looking statement language. Targets such as more than 5 gigawatts by 2030 and broader software inclusion are described as expectations subject to risk and uncertainty. That is standard for public-company disclosures, but it is still a key interpretation point: announced capacity trajectories are directional until construction milestones, power delivery, and commercial workloads are independently observed over time.

Even with that caveat, the announcement is a concrete signal that AI cloud competition is moving from short-cycle GPU procurement to multi-year industrial planning with integrated financing, software validation, and early access to successive compute platforms.