Today’s news that AI cloud computing startup CoreWeave raised a whopping $1.1 billion round of funding at an eye-watering $19 billion valuation (four months ago it was worth $7 billion) likely had many wondering: Who the heck is CoreWeave?
In the spirit of Field of Dreams’ “If you build it, they will come,” imagine the following scenario: A company founded by three commodities traders works out of a New Jersey data center. Their cryptocurrency mining hobby morphs into an Ethereum mining company using Nvidia GPUs (graphics processing units) to verify blockchain transactions. The company pivots to build a specialized cloud infrastructure with massive amounts of GPUs on top — which turns out to be perfect as generative AI booms at the end of 2022. AI companies around the world suddenly have a need for speed to train and run massive models, and they line up to rent access to highly-sought-after GPUs on this new cloud. This creates a gold mine of demand more glittering than the trio could ever have imagined.
That, in a nutshell, is the unlikely story of Roseland, NJ-based CoreWeave, which was founded in 2017 and has now raised a total of $3.5 billion to develop and build hardware and software to support what CEO and co-founder Michael Intrator said in an interview with Fortune is “a very specialized use case in the cloud” — training and running large AI models — which requires “massive scale” infrastructure.
“We are building the largest supercomputers on the planet and we’re building them again and again and again,” he said. “That’s what CoreWeave is uniquely qualified to do…which is why we’ve been able to scale the way we have.” CoreWeave does not host websites or data lakes, for example, the cloud infrastructure is purposefully built to power “some of the world’s most sophisticated and computationally intensive workloads.”
In December 2023, the company announced that it closed a secondary investment of $642 million, following a $420 million primary round led by Magnetar in April 2023. In August, CoreWeave secured a $2.3 billion debt financing facility led by Magnetar and Blackstone. Also in the last year, CoreWeave increased its data center presence from three to 14 (and is currently building 14 more), and quadrupled its employee headcount to over 550.
The $1.1 billion in new funding was led by Coatue, with participation from Magnetar,as well as Altimeter Capital, Fidelity Management & Research Company, and Lykos Global Management. A press release said the new funding will be used to “support the rapid growth across all areas of the business, and CoreWeave’s expansion into new geographic regions to meet the explosive demand for GPU accelerated cloud infrastructure worldwide.”
The Maserati, not minivan, of AI cloud
Intrator told Fortune that he always explains the power of CoreWeave by putting it in car terms. The legacy cloud, he argued, was a minivan. “You can get your kids to the soccer game, you can drag your college roommate’s couch across campus, you can do anything with it and it was incredibly effective,” he said. “We didn’t build that.”
Instead, the CoreWeave team went back to the drawing board to figure out what companies building and running AI models would need. “What do you build? A Maserati,” he said. “We built a highly performant solution that doesn’t have to make all the compromises you make to build a vehicle that can do all these different things. We just want to go fast.”
What people also tend to not “get” about CoreWeave, Intrator argued, is how fundamental the company’s offering is underneath so much of the generative AI boom. “The way I think about it is as a series of base layers,” he explained. “There’s Nvidia and the GPU intellectual property. There’s TSMC as the physical build [of the GPUs]. There’s datacenters. And there’s us.”
In many ways, he continued, CoreWeave is “an incredibly pure, distilled moment of people’s investment in AI.” That is, CoreWeave is incredibly specialized, serving the world’s consumption of artificial intelligence—building it and running it. “We’re not the only ones” doing so, he added, “But we’re one of very few that can build the scale.”
No one knew how much the need for scalable AI infrastructure would boom
CoreWeave’s secret sauce, though, is also access to Nvidia GPUs and its technology: The company certainly benefited from Nvidia’s effort to stay dominant in the AI space — and to compete against other cloud providers like Amazon developing its own chips. Nvidia invested in CoreWeave in 2023 and allotted the company a generous number of GPUs (though Intrator insisted that “we don’t get preferential access to their GPUs—they’re not putting us in front of Amazon.”)
The eye-watering valuation that CoreWeave announced on Wednesday makes sense in the context of everything associated with Nvidia, said Daniel Newman, chief analyst at Futurum Research. “Because right now, [GPUs] are a scarce commodity,” he said. But, Newman wonders, is CoreWeave’s access to Nvidia silicon and its focus on hyper-specific AI workloads, a sustainable long term business, especially once AI chip scarcity eases up?
“I just feel like Microsoft, Google, AWS and others aren’t going to stand by idly and let their [cloud] customers go elsewhere,” he said (Microsoft reportedly inked a deal with CoreWeave a year ago to boost GPU capacity for OpenAI that could be worth billions, but that, Newman said, was just to address short-term capacity challenges). But, he added, “there’s no question that compute is a good market to be in.”
Intrator emphasized that the company has also done well because it has executed on its business plan. “The most knowledgeable investors, the most knowledgeable students of this space, have come back and invested in us time and time again,” he said.
He also admitted that he and his co-founders, Brian Venturo, who serves as chief strategy officer, and Brannin McBee, chief development officer, had no idea what the market was going to become when they were mining cryptocurrency. They understood that GPUs could have alternative uses if you built software that was “endlessly scalable,” and that Ethereum was just one of several “silos” in the cloud.
But they didn’t realized that one of those silos, AI, was going to become world-changing. “I would argue that no one understood it,” Intrator said. “Everybody was caught on their back foot.”
