CoreWeave goes multi‑cloud
- CoreWeave announced a direct interconnect to Google Cloud and cross‑cloud Slurm‑on‑Kubernetes capabilities at Google Cloud Next. (datacenterknowledge.com) - The company is reframing itself from a GPU lessor to connective infrastructure that links specialised compute to hyperscaler environments. (datacenterknowledge.com) - That strategy reduces cloud‑switching costs and lets customers keep existing data gravity while accessing specialised accelerators. (datacenterknowledge.com)
CoreWeave used Google Cloud Next this week to pitch itself as a bridge between clouds, not just a place to rent Nvidia chips. (coreweave.com) The company said it is launching CoreWeave Interconnect, a private fiber link into Google Cloud, with private preview planned in select regions. It also introduced SUNK Anywhere, which extends its Slurm-on-Kubernetes software across CoreWeave, Google Cloud, Amazon Web Services, Microsoft Azure, and on-premises systems. (coreweave.com) Google Cloud’s side of the tie-up is Partner Cross-Cloud Interconnect, a service that lets customers connect workloads across providers without routing over the public internet. CoreWeave said the new setup is aimed at training and inference jobs that need to move between specialized accelerators and data already sitting in a hyperscale cloud. (ciodive.com) Slurm is the job scheduler many research labs use to queue giant computing runs, while Kubernetes is the software layer companies use to manage containers. CoreWeave has been combining the two so customers can run artificial intelligence jobs with Slurm-style scheduling on Kubernetes-based infrastructure. (wandb.ai) That matters because many companies now keep data, storage, and software tools in one cloud but still chase scarce graphics processing units wherever they can get them. Data Center Knowledge reported that third-party connectivity providers can add months of delay and extra cost when teams try to stitch those environments together. (datacenterknowledge.com) Google has been building more of its own artificial intelligence infrastructure stack around that same problem. At Next last year, Google Cloud rolled out Cluster Director and later added managed Slurm features for large-scale AI systems, giving CoreWeave a clearer on-ramp into Google’s existing AI infrastructure tooling. (cloud.google.com) CoreWeave is making the argument at a moment when it has more scale to sell. The company reported $5.13 billion in 2025 revenue and a $66.8 billion revenue backlog on February 26, 2026, after going public on Nasdaq on March 28, 2025 under the ticker CRWV. (investors.coreweave.com) (coreweave.com) The shift also answers a criticism that “neocloud” providers are just leveraged landlords for scarce chips. CoreWeave’s pitch now is that the harder product is the connective layer: the networking, storage movement, and scheduling software that lets customers treat several clouds like one pool of compute. (creativestrategies.com) Google Cloud Next ’26 opened in Las Vegas on April 22, and CoreWeave used the event to attach itself to one of the biggest cloud ecosystems in the market. The closer CoreWeave can sit to where customers already keep their data, the easier it is to sell specialized compute as an add-on instead of a migration. (googlecloudevents.com) (datacenterknowledge.com)
