Cloudflare, a leader in global developer platforms, announced the acquisition of Replicate, the platform for running AI models, to integrate Replicate’s tools directly into Cloudflare’s ecosystem. The integration aims to make deploying AI models as easy as running a single line of code while leveraging Cloudflare’s global network and serverless infrastructure.
Existing Replicate users will see uninterrupted APIs and workflows with added performance and reliability. Workers AI users gain access to an expanded model catalog and the ability to run fine-tuned and custom models on the Workers AI platform.
“The AI Revolution was not televised, but it started with open source,” the announcement noted. Open-source collaboration in machine learning enabled rapid innovation across image, video, audio, and language models. Replicate solved a critical challenge by abstracting complexity, allowing developers to run thousands of models via a simple API call, including proprietary models like GPT-5 and Claude Sonnet.
Cloudflare’s AI Cloud now combines Replicate’s catalog and community with serverless GPU inference, AI Gateway control, Data Stack (Vectorize and R2), and orchestration tools like AI Search, Agents, and Workflows. This integration enables developers to run models globally, fine-tune them, and seamlessly connect AI to applications through Cloudflare’s developer stack.
“This acquisition brings the entire Replicate catalog to Workers AI, providing ultimate flexibility: run models in Replicate’s environment or Cloudflare’s serverless platform from one interface,” the release said. Fine-tuning and custom model deployment will become fully supported, using Replicate’s Cog system for reproducible and seamless integration.
The combined platform allows developers to store outputs in R2 or Vectorize, trigger inference from Workers or Queues, manage AI agent state via Durable Objects, and create real-time generative applications with WebRTC and WebSockets. A unified AI Gateway provides observability, prompt management, A/B testing, and cost analytics across all models, whether running on Cloudflare, Replicate, or other providers.