CoreWeave Offers to Cover Egress Fees for Petabyte-Scale Data Migrations
CoreWeave has launched the Zero Egress Migration ([0]EM) program, a service that moves large datasets from major cloud providers to CoreWeave and, according to the company, covers the egress fees for leaving the existing cloud. CoreWeave says this can save customers up to $1 million on a typical transfer.
- CoreWeave launched the Zero Egress Migration ([0]EM) program for moving large datasets from major clouds to CoreWeave.
- The company says it will cover egress fees when customers leave their existing cloud, saving up to $1 million on a typical transfer.
- Customers provide read-only credentials; migrations typically exceed 1 PB per day with checksum validation, according to CoreWeave.
- Data in CoreWeave AI Object Storage carries no CoreWeave egress fees, wherever it is used.
- CoreWeave claims up to 7 GB/s per GPU throughput, 11 nines durability and 99.9% uptime for the storage service.
What the [0]EM program covers
CoreWeave describes [0]EM as a secure, white-glove service for moving large-scale datasets from major cloud providers directly to CoreWeave. The company points to the cost of data mobility as a central problem for teams scaling AI. According to CoreWeave, transferring training data or model outputs between clouds can cost up to $0.12 per GB in egress fees, which can turn a single petabyte move into a six-figure expense.
CoreWeave also criticizes the terms of existing offers. In its words, most "free data transfer" offers waive egress fees only if customers agree to close their entire accounts as part of a one-time switch. The company argues that this slows innovation and traps data behind commercial fine print.
Under [0]EM, CoreWeave says it will pay the egress fees when customers leave their existing cloud. Once data is stored in CoreWeave AI Object Storage, customers will not pay CoreWeave egress fees, no matter where the data is used. The company says teams can transfer data at petabyte scale and keep full access to their existing environments.
How the migration works
CoreWeave says the process is designed to remove manual effort and uncertainty. Customers provide read-only credentials, and CoreWeave handles the rest. The company says it runs end-to-end checksum validation to confirm integrity and accuracy across every object.
- Migrations typically exceed 1 PB per day, according to CoreWeave.
- The company claims zero data loss, with full fidelity verified at every stage.
- A real-time dashboard shows progress, performance and data integrity.
- CoreWeave's expert team is available through a dedicated Slack channel.
The stated aim is to let platform leaders migrate massive datasets quickly and start running AI workloads with minimal interruption.
AI Object Storage performance claims
The destination for migrated data is CoreWeave AI Object Storage. According to CoreWeave, it delivers up to 7 GB/s per GPU of throughput with 11 nines durability and 99.9% uptime. The company says compute and storage can scale together, keeping throughput and efficiency consistent as workloads grow to hundreds of thousands of GPUs. It also claims linear performance gains with each additional GPU, in contrast to legacy hyperscalers, where it says throughput per GPU declines at scale.
The announcement includes a comment from Xander Dunn, Member of Technical Staff at Periodic Labs: "CoreWeave AI Object Storage delivers the throughput and reliability our research pipelines depend on, balancing speed and efficiency across active and inactive data. It's allowed us to experiment faster while keeping costs under control."
What it means for AI teams
For teams that build across clouds, the program targets two obstacles named in the announcement: steep egress fees and rigid lock-in. CoreWeave positions [0]EM as part of its mission to build what it calls the Essential Cloud for AI and to embrace a multi-cloud world. Teams interested in joining are directed to contact CoreWeave. The announcement does not give eligibility criteria or pricing beyond the figures above, so details of who qualifies would need to be confirmed with the company.
