SANTA CLARA, Calif., Jan 26, 2026, 07:09 (PST)
- Nvidia plans to acquire roughly 23 million CoreWeave shares at $87.20 each, almost doubling its current stake.
- CoreWeave says the funds will accelerate land and power acquisitions for its new AI data centers across the U.S.
- After the announcement, CoreWeave shares surged in early U.S. trading.
On Monday, Nvidia poured $2 billion into CoreWeave, becoming the AI infrastructure firm’s second-largest shareholder as they ramp up efforts to expand U.S. data center capacity. The chipmaker agreed to purchase roughly 23 million shares at $87.20 each, nearly doubling its previous 6.3% stake, according to Reuters calculations using LSEG data. CoreWeave confirmed the funds will be used to acquire land and power for new facilities, explicitly not for buying Nvidia processors. Reuters
As companies race to launch AI tools requiring far more computing power than traditional software, the demand has shifted sharply. Cloud services leveraging accelerators and high-speed networking are now favored, edging out general-purpose servers.
CoreWeave belongs to the neocloud trend—specialized providers leasing graphics processing units, or GPUs, the crucial chips behind AI model training and operation. The real challenge isn’t just the hardware itself; it’s having the right setup to supply power and cooling where it matters most.
CoreWeave’s shares jumped roughly 15% in early U.S. trading, while Nvidia’s stock remained mostly flat, according to LSEG data.
The companies announced that CoreWeave aims to ramp up construction of over 5 gigawatts of “AI factories” — massive data centers tailored for AI tasks — by 2030. Nvidia CEO Jensen Huang described this as “AI entering its next frontier.” CoreWeave’s Michael Intrator added, “AI succeeds when software, infrastructure, and operations are designed together.” CoreWeave plans to test its Mission Control and SUNK software for potential inclusion in Nvidia’s “reference architectures,” which are pre-validated design templates used by cloud partners. The company will also roll out newer Nvidia platforms like Rubin and Vera alongside BlueField storage systems. CoreWeave
CoreWeave began with cryptocurrency mining before pivoting to lease Nvidia chips to AI developers and tech companies. It positions itself as a quicker way to access massive GPU capacity, tapping into the surge in enterprise AI spending.
The funding is aimed at accelerating the less visible tasks: securing sites, constructing building shells, and establishing power connections. These stages often drag on for years, causing schedule delays even when chip supplies are ready.
But this expansion faces two familiar challenges wrapped in a fresh package: will customers actually keep buying enough capacity, and can CoreWeave secure the power it needs when it counts? Steven Dickens, CEO of HyperFrame Research, put it bluntly: “The circular economy continues,” and he’s looking for solid proof that demand will hold steady as AI spending shifts toward “buildout into demand.” Potential bottlenecks like permits, grid upgrades, and equipment lead times could still trip up projects that look simple on paper. DataCenterKnowledge
CoreWeave occupies a niche between massive cloud providers and leaner AI startups, targeting workloads that need dedicated GPU clusters but want to avoid getting tied into larger cloud ecosystems. For Nvidia, this closer partnership provides an additional avenue to embed its AI chips deep into infrastructure, just as competitors and clients explore other options.
