Cloudflare Launches Global AI Inference Network with 1ms Edge Latency for Serverless AI
Cloudflare has launched a global AI inference network spanning 330 cities, enabling sub-1ms AI inference at the edge for serverless AI applications.
In Brief
Cloudflare Workers AI 2.0 launches a global inference network across 330 cities with sub-1ms edge latency. Supporting 100+ models from Meta, Mistral, Google, and Stability AI, with pricing from $0.000001 per inference. Processing 1 billion inferences daily.
Cloudflare has launched Workers AI 2.0, a global AI inference network spanning 330 cities across 120+ countries, enabling sub-1ms AI inference at the network edge. The platform allows developers to run AI models without managing any infrastructure.
The network supports 100+ pre-trained models from Meta, Mistral, Google, and Stability AI, including Llama 4, Mistral Large 3, Gemini Nano, and Stable Diffusion 4. Models run on NVIDIA H100 GPUs deployed in Cloudflare's edge locations, with automatic scaling and failover.
Information
Workers AI 2.0 pricing starts at $0.000001 per inference for small models — making AI inference 100x cheaper than traditional cloud GPU providers.
Cloudflare claims that the network reduces AI inference latency by 90% compared to traditional cloud providers by running models physically closer to end users. The platform also includes built-in DDoS protection, rate limiting, and data residency controls.
“AI inference should be as fast and accessible as serving a web page. Workers AI 2.0 makes it possible to run AI models at the edge with millisecond latency and zero infrastructure management.”
— Matthew Prince, Cloudflare CEO
Major customers include Discord, which uses the platform for real-time content moderation; Shopify, for AI-powered product recommendations; and Stripe, for fraud detection. Cloudflare reports processing over 1 billion AI inferences per day on the network.
Key Takeaways
- 01Global AI inference network across 330 cities in 120+ countries
- 02Sub-1ms inference latency at the edge — 90% reduction vs traditional cloud
- 03100+ pre-trained models from Meta, Mistral, Google, and Stability AI
- 04Pricing from $0.000001 per inference — 100x cheaper than cloud GPU providers
- 05Processing over 1 billion AI inferences per day for customers like Discord and Shopify
Frequently Asked Questions
What is Workers AI 2.0?
Workers AI 2.0 is Cloudflare's serverless AI inference platform that runs models at the network edge across 330 global cities, eliminating infrastructure management.
What models does Workers AI 2.0 support?
The platform supports 100+ models including Meta's Llama 4, Mistral Large 3, Google's Gemini Nano, and Stability AI's Stable Diffusion 4, with new models added weekly.
Sources
Topics
Stay Ahead with TECHRADAR360
Get the latest tech news delivered to your inbox every morning.
Related
NVIDIA Unveils Blackwell Ultra GPU Built on 3nm Process — 2x AI Performance Leap
AMD Ryzen 9000 Series Breaks Gaming Benchmarks with Zen 6 Architecture
