CoreWeave and IBM to build Nvidia G200 AI supercomputer

CoreWeave has announced it will partner with IBM to deliver Big Blue with a new AI supercomputer to train its open-source Granite models.

As part of the build, the Nvidia-backed AI firm will deliver the supercomputer with one of the first Nvidia GB200 Grace Blackwell Superchips. It will also be equipped with Nvidia GB200 NVL72 systems and interconnected with Nvidia’s Quantum-2 InfiniBand network.

โ€œWe are thrilled to partner with IBM, a long-time pioneer of innovative technology solutions, to push the boundaries of artificial intelligence,โ€ CoreWeave CEO and Co-Founder Michael Intrator said.

โ€œThis collaboration is a testament to CoreWeaveโ€™s ability to deliver some of the worldโ€™s most advanced AI cloud solutions and will combine our strengths in engineering and product development. We look forward to deepening our relationship with IBM to drive transformative innovation together.โ€

According to IBM, the supercomputer will also leverage IBM Storage Scale System featuring NVMe SSDs, which is touted to deliver high-performance storage for AI research and development, data analytics and other demanding workloads.

What the IBM partnership means for CoreWeave customers

Under this agreement between the pair, CoreWeave customers can access the IBM Storage platform within CoreWeaveโ€™s dedicated environments and AI cloud platform, IBM said.

โ€œWe are excited to work with CoreWeave on state-of-the-art AI hardware and software to unlock new capabilities for future generations of IBM Granite models,โ€ IBM Research AI VP Sriram Raghavan said.

โ€œCoreWeaveโ€™s cutting-edge platform can augment IBMโ€™s organic capabilities to help build advanced, performant, and cost-efficient models for powering enterprise AI applications and agents.

โ€œIn turn, IBM Storage is enabling a new world of possibilities for AI by offering IBM Storage Scale System to enhance CoreWeaveโ€™s comprehensive suite of developer-focused AI capabilities.โ€

Raghavan added that IBM will use the supercomputer to advance open technologies such as Kubernetes to power AI computing in a hybrid cloud environment.

Earlier this week, Reuters cited an initial report by The Information, that said the first shipments of racks with Blackwell chips have been overheating and exhibiting glitches. As a result, top customers such as Microsoft and Amazon are delaying orders, according to The Information. However, the report did not name IBM as one of those customers.

Aimee Chanthadavong
Aimee Chanthadavong

Aimee Chanthadavong has been a journalist, editor and content producer for more than a decade. During that time she's covered enterprise technology for premium websites such as ZDNet and InnovationAus as well as food and travel for Broadsheet and SBS.