CoreWeave Builds Ecosystem Momentum

Money continues to pour into AI datacenter service provider CoreWeave in a trend that heralds a potential IPO next year and could boost the fortunes of many of its suppliers and partners.

The Roseland, N.J.-based company, founded in 2017, last week closed a $650 million revolving credit facility to fuel the buildout of datacenters in the U.S. and Europe. The week before, Cisco was said to be considering investing in CoreWeave. Further, word is that employees may soon tender shares worth up to $500 million on a company valuation of $23 billion.

Behind all this activity is demand for hyperscaler and large enterprise processing of AI models in so-called AI factories—datacenters specifically designed for these workloads. These new facilities rely on compute, storage, and networking components as well as energy-saving datacenter technologies.

CoreWeave Tide Lifts Many Boats

CoreWeave’s success is a rising tide lifting a range of boats. CoreWeave boasts that it has been among the first providers of services based on NVIDIA’s Blackwell GPUs. Specifically, CoreWeave is offering two Blackwell configurations: the NVIDIA HGX B200 and the NVIDIA GB200 NVL72, both with NVIDIA Quantum-2 InfiniBand networking.

CoreWeave also has chosen Nokia for optical and IP transport networking across its global backbone of interconnected datacenters. The deal has helped boost Nokia’s prospects as a provider of AI networking.

CoreWeave's fortunes could also improve the prospects of competitors, including Lambda Labs, which counts Amazon and MIcrosoft among customers for its GPU-based AI processing services.

One of CoreWeave’s biggest contributions is to the datacenter construction economy. This past August, a $5 billion joint venture agreement between Blue Owl Capital Inc., Chirisa Technology Parks, and PowerHouse Data Centers was drawn up to support CoreWeave’s buildout of datacenters through 2026. Included in the plan is the construction of datacenters in Richmond, Va., along with future facilities in New Jersey, Pennsylvania, Texas, Kentucky and Nevada.

By the end of 2024, CoreWeave plans to expand its footprint of global datacenters from about 14 earlier this year to 28. Ten more datacenters are planned for 2025, including ones planned in a $3.5 billion project focused on Europe.

Energy Savings Takes Priority

CoreWeave’s datacenter rollout will incorporate liquid cooling technologies from unnamed vendors as part of an overall effort toward addressing the power challenges presented by AI datacenters. In a blog on the topic, Jacob Yundt, director of compute architecture at CoreWeave described the issues:

“In data centers today, you can’t fit the same number of NVIDIA H100 servers or GB200 Superchip compute trays in a cabinet as you could earlier GPU generations because of the higher power and cooling demands. Not only does liquid cooling enable more efficient heat dissipation of the latest chips, it also saves us the power that was earlier being spent on fans. The improved thermal efficiency and power savings allow us to be more generous with how many GPUs we can fit into a rack—which means more GPUs for our customers.”

Support from Many Sources

CoreWeave’s latest credit facility was led by led by JPMorgan Chase, Goldman Sachs, and Morgan Stanley, with participation from Barclays, Citi, Deutsche Bank, Jefferies, Mizuho, MUFG, and Wells Fargo. In addition, the company has raised $12.7 billion in equity and debt financing from Blackstone, Magnetar, Coatue, and others. The anticipated investment by Cisco is still a question.

All of the funding and projected activity by CoreWeave has fueled talk of a potential IPO next year. As reported by Futuriom, CoreWeave has hired personnel specifically for the purposes of investigating SEC filings and investor relations.

Futuriom Take: CoreWeave’s progress in funding and buildout of its liquid-cooled datacenters augurs more activity among its suppliers and possibly and IPO in 2025.