In a move that underscores the growing demand for AI infrastructure, Nvidia expanded its partnership with neocloud vendor CoreWeave.
The vendors revealed on Monday that Nvidia will help CoreWeave build AI factories to produce 5 gigawatts of electric power capacity by 2030. Nvidia also invested $2 billion in the neocloud vendor by buying CoreWeave Class A common stock at a purchase price of $87.20 per share. The news sent shares of CoreWeave, which has seen prices plummet over concerns about its debt load, higher in early trading. The new investment is in addition to the $6.3 billion commitment Nvidia made last year to use CoreWeave’s infrastructure through 2032.
The expanded partnership points to an AI market increasingly focused on AI infrastructure and the build-out of AI factories and data centers.
The demand for land and real estate to build these data centers has intensified over the past year. Microsoft revealed plans for a $17.5 billion AI infrastructure investment in India in December. Meanwhile, OpenAI’s Stargate initiative is underway with construction on an AI data center in Texas already started and five more data centers planned in partnership with Oracle and SoftBank in Ohio, New Mexico, Texas and Wisconsin. For its part, Nvidia has numerous infrastructure partnerships. The AI hardware and software giant revealed plans to invest up to $100 billion in OpenAI and supply chips to the generative AI vendor to build out its own AI data centers.
Nvidia’s Support
With the expansion of its partnership with CoreWeave, Nvidia is acting more like a co-developer, helping the neocloud provider secure both compute power and real estate, said Nick Patience, an analyst at Futurum Group.
“Power is the new bottleneck here and real estate to a certain extent,” Patience said. “The two things are closely linked because you get a lot of data centers being put where the power is.”
Moreover, Nvidia will use CoreWeave’s software in its reference architectures, elevating the neocloud provider from a pure service provider to a technology partner. Nvidia’s reference architecture is a blueprint that helps organizations build AI factories.
“This is … basically an endorsement by Nvidia saying [CoreWeave’s] software is worth something,” Patience said.
For CoreWeave, the stamp of approval from Nvidia helps it differentiate from competitors such as Lambda Labs and Nscale.
CoreWeave’s access to the new Nvidia Vera Rubin platform also gives it a competitive advantage.
“Potentially, enterprises could look upon CoreWeave for more than just GPUs,” Patience added.
Advantages for Nvidia and Challenges
For Nvidia, the deal is also an avenue to promote its Nemotron family of open models.
“Nvidia certainly wants the world to know that it’s much more than a GPU company,” Patience said. “This gives them another channel for distribution of their software and models.”
However, the deal raises the now-familiar issue of circular financial arrangements, Patience said. Since Nvidia is investing in a company that uses the money to buy Nvidia chips, then Nvidia is essentially getting that money back.
Another challenge for CoreWeave is that it could become dependent on Nvidia as a supplier.
