SANTA CLARA, Calif., Jan 26, 2026, 07:09 (PST)
- Nvidia will buy about 23 million CoreWeave shares at $87.20, nearly doubling its stake.
- CoreWeave says the cash will speed land and power purchases for new U.S. AI data centers.
- CoreWeave shares jumped in early U.S. trading after the announcement.
Nvidia invested $2 billion in CoreWeave on Monday, becoming the AI infrastructure provider’s second-largest shareholder as the companies deepen a partnership aimed at adding U.S. data center capacity. Nvidia agreed to buy about 23 million shares at $87.20 apiece, nearly doubling a stake that had stood at 6.3%, Reuters calculations based on LSEG data showed. CoreWeave said the cash would go toward securing land and power for new sites and would not be used to buy Nvidia processors. (Reuters)
The deal lands as companies rush to roll out AI tools that need far more computing power than traditional software. That has pushed demand toward cloud services built around accelerators and fast networking, and away from general-purpose servers.
CoreWeave is part of a so-called neocloud wave — specialist providers that rent out graphics processing units, or GPUs, the chips used to train and run AI models. In practice, the bottleneck is often not the hardware alone, but the ability to power and cool it in the right place.
CoreWeave shares were up about 15% in early U.S. trading, while Nvidia shares were little changed, according to LSEG data.
In a joint statement, the companies said CoreWeave plans to accelerate the buildout of more than 5 gigawatts of “AI factories” — large data centers designed for AI workloads — by 2030. Nvidia CEO Jensen Huang said “AI is entering its next frontier,” while CoreWeave chief Michael Intrator said “AI succeeds when software, infrastructure, and operations are designed together.” CoreWeave said it will test its Mission Control and SUNK software for possible inclusion in Nvidia “reference architectures,” or pre-validated design blueprints used by cloud partners, and will deploy newer Nvidia platforms including Rubin and Vera along with BlueField storage systems. (CoreWeave)
CoreWeave started out mining cryptocurrency and later repurposed that infrastructure to lease Nvidia chips to AI developers and technology firms. It has marketed itself as a faster route to large pools of GPU capacity as enterprise spending shifts toward AI.
The funding will be used to speed up the less glamorous work: locking up sites, building shells, and getting power connections in place. Those steps can take years, and delays can ripple through schedules even when chips are available.
But the buildout is exposed to two old problems in a new wrapper: will customers keep buying enough capacity, and can CoreWeave get the power it needs on time. “The circular economy continues,” said Steven Dickens, chief executive at HyperFrame Research, adding that he wants proof demand holds as AI spending moves toward “buildout into demand.” Permits, grid upgrades and equipment lead times could still slow projects that look straightforward on a slide deck. (DataCenterKnowledge)
CoreWeave sits between hyperscale cloud groups and smaller AI startups, competing for workloads that want dedicated GPU clusters without broader cloud lock-in. For Nvidia, the tighter relationship offers another way to keep its AI chips embedded in the infrastructure layer as rivals and customers hunt for alternatives.
