NVIDIA has significantly deepened its strategic bet on CoreWeave, committing about $2 billion in new equity to support what executives are calling the largest AI infrastructure buildout in history. The investment, disclosed in an SEC Form 8-K, nearly doubles Nvidia’s ownership stake to about 11.5%, following the purchase of approximately 23 million additional Class A shares. NVIDIA Corporation (NVDA) closed at $186.47, down $1.21 (-0.64%) at 4:00 PM EST, before edging up to $187.15, a gain of 0.36%, in overnight trading. CoreWeave, Inc. (CRWV) closed at $98.31, surging 5.73%.
The shares were acquired at an average price of $87.20 per share, representing an estimated 6% discount to recent market levels. For CoreWeave, which markets itself as the “Essential Cloud for AI,” the infusion provides critical balance sheet relief by reducing reliance on debt while accelerating the construction of next-generation data centers purpose-built for artificial intelligence workloads.
Commenting on the development, Jensen Huang, founder and CEO of NVIDIA, stated, “CoreWeave’s deep AI factory expertise, platform software, and unmatched execution velocity are recognized across the industry. Together, we’re racing to meet extraordinary demand for NVIDIA AI factories, the foundation of the AI industrial revolution.”
Michael Intrator, co-founder, chairman, and CEO of CoreWeave, stated, “From the very beginning, our collaboration has been guided by a simple conviction: AI succeeds when software, infrastructure, and operations are designed together.”
CoreWeave went public in March 2025 and has rapidly emerged as a leading “neocloud” provider, positioning itself as a focused alternative to traditional hyperscalers such as Amazon Web Services, Microsoft Azure, and Google Cloud. Its operational headquarters and several flagship facilities are based in Livingston, New Jersey, where much of the new capacity is expected to come online.
The 5-Gigawatt AI Factory Vision
At the center of the deal is an ambitious AI Factory roadmap. NVIDIA and CoreWeave have jointly set a target of 5 gigawatts (GW) of installed power capacity by 2030, a scale comparable to the electricity usage of roughly 3.75 million U.S. homes. If realized, the footprint would place CoreWeave among the largest AI-focused compute operators globally.
The announcement reinforces the growing “circular financing” narrative, where Nvidia invests directly in its own customers to guarantee demand for its hardware. Alongside the equity purchase, Nvidia maintains a reported $6.3 billion service commitment with CoreWeave, effectively backstopping capacity expansion while locking in long-term deployment for its platforms.
From a technology standpoint, CoreWeave will serve as a primary launch partner for Nvidia’s Vera CPU, the company’s first standalone central processing unit. The platform will be integrated with Nvidia GPUs, including the upcoming Rubin architecture, which is expected to succeed Blackwell when it enters deployment in 2026.
High-speed networking and storage across the new AI factories will rely on BlueField-4 DPUs, reinforcing Nvidia’s strategy of controlling the full data center stack.
On the software layer, CoreWeave plans to standardize its SUNK (Slurm on Kubernetes) orchestration system for managing large-scale AI clusters. In parallel, its Mission Control observability platform will be embedded into Nvidia Reference Architectures, enabling tighter performance tuning and faster training cycles for advanced AI models.
The deal builds on CoreWeave’s expanding customer base, which includes Meta Platforms, tied to a previously disclosed $14.2 billion compute agreement, and OpenAI, with a reported $22.4 billion total contract value. Together, these commitments highlight surging demand for dedicated AI infrastructure and Nvidia’s increasingly direct role in shaping it.
