Coding agents now trigger half of Vercel deployments
Vercel’s deployment data suggests coding agents are already changing downstream developer infrastructure, not just how code gets written.
AI coding tools are beginning to reshape the infrastructure around software delivery, not just the editor where code is written.
What happened
Vercel says its platform now handles around 6M deployments per day, with half triggered by coding agents. The company also says more than 1T tokens per day flow through its AI gateway.
The data points to a production stack where models, gateways, sandboxes, data systems and agent frameworks can increasingly be combined and swapped independently.
Why it matters
If coding agents are already responsible for half of daily deployments on a major developer platform, agentic software creation is moving beyond experimentation.
That changes where infrastructure demand appears. More autonomous coding can increase the need for deployment systems, testing environments, sandboxes, observability and model-routing layers.
It also raises a strategic question over where value sits: inside the model itself, or in the orchestration and deployment infrastructure that lets developers switch between models and agents.
The bigger picture
The developer stack is becoming more modular around AI.
As companies use multiple models and agent frameworks, infrastructure providers may benefit by sitting between those systems and production. The long-term winner may not be one model provider controlling the entire stack, but a broader ecosystem of interchangeable tools around agentic software development.