CoreWeave has introduced its Zero Egress Migration (0EM) program, a true no-egress-fee data migration offering designed for organizations moving large AI datasets from third-party clouds. The program allows customers to transfer data directly into CoreWeave’s cloud. At the same time, CoreWeave covers the egress charges for the initial migration of AI workloads from providers such as AWS, Azure, Google Cloud, IBM, and Alibaba.
0EM is delivered as a fully managed service that focuses on secure, high-speed, and verifiable data movement. CoreWeave coordinates the end-to-end process, including secure data paths, performance tuning, and dataset integrity validation at scale. This structure reduces operational risk and internal engineering overhead typically associated with large data migrations while eliminating a significant cost barrier: egress fees.
A Single Global Dataset for AI
Once migrated, data lands in CoreWeave AI Object Storage, which is designed to present a single global dataset rather than fragmented copies spread across regions or providers. This architecture helps reduce capital and operational overhead caused by data duplication, manual tiering, and sprawl across multiple environments.
A key component is CoreWeave’s Local Object Transport Accelerator (LOTA) technology. LOTA is engineered to deliver up to 7 GB per second of throughput per GPU, regardless of the data’s physical location. In practice, this helps keep GPUs continuously fed during training and inference, thereby improving utilization rates and reducing idle time. The combination of AI Object Storage and LOTA targets both cross-cloud flexibility and predictable performance for multi-cloud AI pipelines.
Addressing Structural Limits of Legacy Cloud Storage
CoreWeave’s leadership has been explicit that conventional cloud storage architectures were not built for the AI era, where rapid movement of massive datasets within and across clouds is mandatory. High egress and transfer fees, along with rigid architectures, can inhibit experimentation and slow production deployment of AI applications.
Through AI Object Storage and the 0EM program, CoreWeave is attempting to reshape this model by removing financial, physical, and operational bottlenecks around data mobility. The focus is on enabling AI teams to move data where it delivers the most value without being constrained by punitive network fees or complex, manual migration projects.
Operational Control, Visibility, and No Lock-In
The 0EM program is designed to work alongside existing cloud strategies rather than requiring a complete cutover. Customers can keep active accounts with their current cloud providers during and after migration, maintaining flexibility for multi-cloud or hybrid architectures. CoreWeave does not impose exit penalties, which is a notable feature for technical and commercial decision-makers concerned about vendor lock-in.
To support operational transparency, CoreWeave provides a migration dashboard that exposes real-time visibility into data transfer progress, throughput, performance metrics, and integrity checks. This level of monitoring is essential for teams that must validate large-scale moves to meet compliance, internal audit, or strict SLA requirements for production AI workloads.
Integrated AI Platform with Proven Performance
0EM extends CoreWeave’s broader effort to assemble a full AI-focused cloud stack. The platform already spans high-performance compute, multi-cloud-compatible data storage, and a software layer that allows builders to develop, test, and deploy AI workloads at scale.
Recently, CoreWeave introduced ServerlessRL, a fully managed, publicly available reinforcement learning capability that targets complex training workflows without requiring users to stand up and manage bespoke RL infrastructure. The company’s technology credentials are reinforced by industry benchmarks, including an industry-leading MLPerf result for AI workloads and Platinum rankings in SemiAnalysis ClusterMAX 1.0 and 2.0. These rankings are widely regarded as a reference for AI cloud performance, efficiency, and reliability, which matters for teams standardizing on a long-term AI infrastructure partner.
Availability
CoreWeave’s 0EM program is generally available to all customers starting today. Organizations interested in the program can learn more by visiting the program webpage or by engaging directly with CoreWeave’s sales and solutions engineering teams.
Amazon