Vercel CEO Guillermo Rauch says the AI sector is moving beyond experimentation and into large-scale deployment, with his company handling 6 million deployments a day and more than 1 trillion tokens daily through its AI gateway. Rauch said about half of those deployments are triggered by coding agents, which he described as one of the clearest production uses for AI. He also identified internal corporate assistants as a second major use case, framing both as evidence that businesses are now focusing less on pilot projects and more on operational systems.
Rauch said that shift has pushed security and governance to the forefront. He pointed to Vercel products including Eve, a framework for defining agent instructions in natural language, and Vercel Sandbox, which is designed to restrict what data an AI agent can access or send out. He argued that such controls are increasingly important for companies with sensitive proprietary information, citing the risk that a misconfigured coding tool could expose large internal codebases to cloud-based model training.
On competition, Rauch said customers are moving away from relying on a single AI lab and instead assembling systems from interchangeable parts such as models, data platforms, gateways and sandboxes. He said developers can now choose among providers including OpenAI, Anthropic and Gemini, while also adopting open models such as Deepseek and GLM-5.2. Rauch contended that production users are increasingly making decisions based on price and performance, and he said this has contributed to growth for Gemini and other alternatives.
At the same time, Rauch acknowledged that the major model companies are expanding into areas that overlap with infrastructure providers. He cited OpenAI's release of tools that can publish directly to the web as an example of how AI labs are broadening their reach. Rauch argued that the industry is now deciding whether models and agents will remain tightly bundled within a single provider's ecosystem or become more modular components that companies can combine more freely. He said Vercel is positioning itself around the latter approach, while presenting open protocols and data portability as better suited to long-term enterprise use.