IBM and CoreWeave, a leading AI specialty AI cloud provider, together announced a collaboration to deliver one of the first NVIDIA GB200 Grace Blackwell Superchip-enabled AI supercomputers to market.
The system, equipped with NVIDIA GB200 NVL72 rack-scale systems and interconnected via NVIDIA Quantum-2 InfiniBand networking, will support the training and deployment of IBM’s Granite models — IBM’s series of open-source, enterprise-ready AI models.
Let’s take a quick look at the deal.
The Collaboration
The partnership leverages CoreWeave’s highly performant, resilient, and secure AI cloud platform to address the computational demands of training and refining IBM’s Granite models. The NVIDIA GB200 Grace Blackwell Superchip provides the advanced processing power needed for these tasks, while NVIDIA’s Quantum-2 InfiniBand ensures seamless connectivity for high-speed data transfer.
This collaboration allows IBM to accelerate the development of Granite models to enhance their utility across hybrid and multi-cloud environments. Using CoreWeave’s infrastructure, IBM can efficiently iterate and optimize Granite to meet evolving enterprise requirements.
Advanced Storage and Open Ecosystems
In addition to leveraging CoreWeave’s hyperscale infrastructure, the new supercomputer will integrate IBM’s Storage Scale System, using NVMe-attached flash storage to handle data-intensive AI workloads.
Granite’s deployment is also aligned with IBM’s commitment to open technologies and hybrid cloud solutions. The collaboration incorporates Kubernetes-based management, enabling Granite models to be developed and deployed flexibly across cloud and on-premises environments. This approach underscores the importance of interoperability and scalability in modern AI solutions.
Background: IBM Granite Models
Granite models are part of IBM’s broader AI and hybrid cloud strategy, emphasizing scalability, transparency, and compliance with enterprise standards.
IBM Granite is a series of open-source, enterprise-ready AI models designed by IBM to meet enterprise-specific needs. These models balance high performance, safety, and cost-efficiency, making them well-suited for real-world enterprise use cases.
Unlike larger, generalized AI models, Granite deliverss state-of-the-art performance without requiring excessive computational resources. This design enables faster training, deployment, and inference while minimizing costs.
Granite models are built with enterprise applications in mind, such as:
- Automating workflows
- Advanced data analytics
- Powering intelligent agents and chatbots
- Enhancing decision-making processes
Analyst’s Take
For CoreWeave, the partnership reinforces its position as a leader in hyperscale AI cloud solutions, capable of supporting enterprise-scale AI projects. The deal also underscores IBM’s dedication to advancing enterprise AI by leveraging state-of-the-art technology to deliver tailored solutions for specific business challenges.
The collaboration is a strong example of how specialized AI infrastructure and enterprise-focused model development can work together to address real-world needs. Granite is both a flagship workload for CoreWeave’s advanced infrastructure and a demonstration of IBM’s ability to combine performance, safety, and cost-efficiency in AI applications.
The CoreWeave-IBM partnership also highlights the growing importance of collaborations in driving AI innovation. By combining CoreWeave’s cutting-edge cloud platform with IBM’s enterprise-ready AI models, the two companies aim to set new benchmarks in AI infrastructure and application development.
Granite models trained on this advanced system will empower businesses to adopt AI confidently, with solutions that are not only powerful but also practical, safe, and cost-effective. The new partnership between CoreWeave and IBM exemplifies the future of enterprise AI — one that balances technological sophistication with real-world utility.
