Cloudflare Unifies AI Gateway and Workers AI

Cloudflare is consolidating its AI offerings, merging AI Gateway and Workers AI into a unified control plane. This move aims to simplify how developers connect to and manage AI models, whether they are hosted by Cloudflare or external providers. The unification means a single point for observability, billing, and security, reducing complexity for users.

The company, known for its edge computing and cybersecurity services, has seen its users increasingly converge on similar goals with its distinct AI products. AI Gateway offered proxying to any model provider with built-in features like logging and security. Workers AI, on the other hand, provided inference-as-a-service using Cloudflare's managed GPU infrastructure. Now, these will operate under a singular management layer.

Simplified Access and Observability

A key benefit of this integration is the introduction of a unified entry point through both a Workers binding and a REST API. Developers can now use a single AI binding to call either AI Gateway or Workers AI. Cloudflare also shipped a concept of a "default" gateway, which automatically provides observability and logging for Workers AI usage even without explicit setup. This means every request is logged, token counts are tracked, and cost attribution is available from the first authenticated request.

Previously, using Workers AI directly meant operating without the detailed insights offered by AI Gateway. Now, by specifying `gateway: { id: 'default' }` in the binding or using the `cf-aig-gateway-id: default` header in API calls, users automatically inherit AI Gateway's visibility features. This upgrade is significant for teams debugging model behavior or auditing AI output, moving from "flying blind" to having full visibility into latency, token usage, and error rates.

For those needing more granular control, the option to create named gateways remains. This allows for custom caching rules or traffic splitting by application, simply by changing a parameter in the request.

Unified Billing and Credits

The convergence also extends to billing. Cloudflare is enabling the use of AI Gateway credits for Workers AI usage. Previously, these credits were only applicable to external model providers like OpenAI or Anthropic. Now, users can load credits into a wallet and spend them across any supported provider, including Cloudflare's own hosted models. This unified billing approach simplifies financial management and is being incentivized with elevated rate limits on Workers AI models for users employing this unified billing path.

This strategy reflects a broader trend in the AI infrastructure space. As the number of model providers and specialized AI services grows, platforms that can abstract away the complexity of integration and management become increasingly valuable. Companies like Microsoft (NASDAQ:MSFT) with Azure AI, and Alphabet Inc. (NASDAQ:GOOGL) with Google Cloud AI, offer broad suites of tools, but Cloudflare's approach focuses on the edge and developer-centric tooling.

The Future of Model Routing

Looking ahead, Cloudflare plans to introduce model-first routing. This will allow users to define routing rules based on the specific AI model being requested, further enhancing flexibility. The move to a single control plane positions Cloudflare to offer a more comprehensive solution for developers building AI-powered applications, whether they're running models at the edge or calling out to external services. For startups, this consolidation could lower the barrier to entry for adopting sophisticated AI capabilities, abstracting away infrastructure concerns and simplifying cost management.