CoreWeave adds NVIDIA AI Enterprise and Cloud Functions to its AI cloud

What CoreWeave is adding

CoreWeave says the goal is to help customers build, deploy and manage AI applications more easily. The company argues that as AI models become more capable and adoption accelerates, optimizing and scaling inference grows more complex. In its words, customers often spend significant time managing deployments and struggle to optimize model-serving performance.

NVIDIA AI Enterprise is described in the announcement as a cloud-native software platform that streamlines the development and deployment of production-grade, end-to-end generative AI pipelines. It includes NVIDIA NIM microservices, NVIDIA NeMo, AI frameworks, reference applications and workflows, SDKs, libraries, and infrastructure management.

NVIDIA Cloud Functions, a core technology of DGX Cloud Serverless Inference, is an auto-scaling AI inference solution. According to CoreWeave, NVCF simplifies hosting inference workloads such as NIM microservices by reducing the time needed to deploy optimized infrastructure. NIM microservices are containerized AI models and inference engines that can be deployed and scaled on NVIDIA GPUs.

How customers get started

The support runs on CoreWeave Kubernetes Service (CKS). CoreWeave says CKS includes ready-to-use components such as optimized networking and storage interfaces, GPU drivers, Slurm-on-Kubernetes integration, and built-in observability tools.

The setup process described in the announcement has a few steps:

• Customers install the NVIDIA Cluster Agent Operator, available from NVIDIA's NGC Platform, so an existing CKS cluster can act as a deployment target for NVCF.
• With a registered CKS cluster and the Cluster Agent installed, an NVCF function can be created and deployed through the NVIDIA Cloud Functions console or the NGC CLI.
• When creating a function in the UI, customers can pick "Elastic NIM" under Function Type to use NVIDIA performance-optimized NIM microservices, which are part of NVIDIA AI Enterprise.

NVCF on GB200 NVL72

CoreWeave highlights its work on NVIDIA GB200 NVL72, noting that it was the first cloud provider to make instances based on that system generally available. As part of the enablement work, the company says its engineers deployed and ran NVCF on GB200 NVL72-accelerated CoreWeave instances running CKS.

The engineers created a new cloud function tied to a custom model and deployed it on those instances, ready for invocation. An image in the announcement shows a DeepSeek R1 model deployed on a CoreWeave instance accelerated by NVIDIA GB200 NVL72.

What it means for customers

For teams already running on CoreWeave, the update adds another route to deploy inference workloads using NVIDIA's software stack, without leaving the CKS environment. CoreWeave says the combination should help customers bring solutions to market at higher performance and lower costs, though the announcement gives no benchmark figures or pricing to back that up.

The company invites partners and customers to get in touch to explore the offering. The announcement does not give details on pricing or on any regions where the support is available.