The preferred storage option that doesn’t lock you in: zero egress fees.

In a rapidly-evolving AI landscape where new tools are offered, you cannot afford to be locked into platforms. Store data in R2 to reduce dependency on any specific GPU vendor now, and future-proof your data strategy.
Unlocking value from your data with AI means your storage needs will continue to grow. Choose a trusted solution that scales with you. Store petabytes of high-value, frequently accessed datasets and checkpoints in R2, and focus on gaining deeper insights.
Latency and speed are critical for AI workloads. Cloudflare’s global network delivers datasets from R2 with fast downloads and low-latency access. Use built-in security and orchestration tools to serve models, run inference, and deliver great experiences.

AI & ML teams rely on a multi-cloud architecture for training and processing requests, due to the price and availability of GPUs in the cloud. These distributed architectures require moving data between clouds, which lead to high bills in egress fees. As companies’ data grows, storage and egress get even more expensive.
The teams at Character AI migrated to R2 to store training and production data, and make it accessible to any cloud without incurring egress fees.
Now that R2 is the glue between their clouds, their teams are able to move their data freely and preserve their technical optionality. When cloud and compute needs change in the future, their data and workloads remain flexible.

Migrate to R2 as a strategy to keep your data workloads flexible and compute agnostic
Enjoy the benefits of preserving your technical optionality with immediate ROI on every request served from R2 post-migration. You can quickly offset any migration expenses with savings from egress and storage costs.
Cloudflare’s data migration UI and API services help you quickly and easily copy objects into R2 from any S3-compatible object storage provider including MinIO, Wasabi, Backblaze B2, and DigitalOcean Spaces.
Built for one-time bulk transfers. Ideal for fast and full migration.
Learn moreBuilt for on-demand transfers. Helps realize immediate value from migrating "hot" data first.
Learn more“We are now able to store our training & production data in R2 for access by any cloud, without egress fees, and get the best prices and performance across multiple cloud providers”
– Founding Researcher, Character AI
Store and access massive AI training datasets in Cloudflare R2 without worrying about egress fees. R2’s global, S3-compatible storage lets you move data freely between compute environments, making multi-cloud and hybrid AI training more affordable and flexible
See how Cloudflare R2’s zero egress fee model can dramatically lower your AI training data storage and transfer costs compared to AWS S3. Free up budget for compute and scale your experiments without hidden fees
Gain visibility and control over how your AI apps and training pipelines access data stored in R2. AI Gateway provides observability, security, and traffic management - helping you optimize data usage and safeguard your training endpoint.
Run inference on your AI models close to where your training data lives in R2, leveraging Cloudflare’s global network. Reduce latency and simplify deployment by serving models and accessing data wherever your users or compute clusters are located.
Automate data preprocessing, transformation, and orchestration for your AI training workflows with Cloudflare Workers. Build serverless applications that operate directly on data in R2, streamlining your pipeline from data ingestion to training
Speed up LLM training and reduce storage costs by pairing MosaicML’s cloud-agnostic training platform with Cloudflare R2. Stream data and checkpoints efficiently between compute clusters and storage, enabling flexible, cost-optimized AI development
Jumpstart your projects with Cloudflare for startups. Access credits and support to leverage R2 for scalable, affordable training data storage as you grow your business.