Guillermo Rauch says AI agents now trigger more than half of all Vercel deployments

Vercel CEO Guillermo Rauch announced that AI agents now trigger more than half of the platform's 6 million daily deployments, up from under 3% six months ago, marking a tipping point in the agent era. At Vercel's Ship conferences in June, Rauch revealed the company's annual recurring revenue surged 240% to $340 million by February 2026, and its AI gateway processes a trillion tokens daily. Vercel launched an open-source agent framework called eve and Vercel Services to position itself as the operating layer for production AI agents.

Vercel's platform crossed a tipping point this year: coding agents now initiate more than 50% of its 6 million daily deployments, up from under 3% six months ago, and CEO Guillermo Rauch is repositioning the company as the operating layer of the agent era. The numbers Rauch has been sharing publicly are hard to argue with. At Vercel's Ship conference in London on June 17 and then in New York on June 30, he laid out what the company is actually seeing at infrastructure scale: 6 million deployments a day, more than half triggered by coding agents rather than human developers typing commands. A trillion tokens flowing through Vercel's AI gateway daily. Thirty percent of the apps running on the platform originated from an AI agent, not a person. Annual recurring revenue at $100 million when 2024 opened, running at $340 million by the end of February 2026, a 240% increase in roughly two years. Those aren't projections. That's the platform's current shape. The strategic argument Rauch is making from those facts is equally direct. As he told TechCrunch in early July, the industry is splitting into two camps: one that wants to bundle the AI model and the agent into a closed product stack, and one that insists on keeping them separate. He's firmly in the second camp. "Do you get all your intelligence from one place? Or do you get a module or a library or a building block from one provider, and then you build on top of it?" That question, in Rauch's telling, is the real infrastructure fight of 2026. At Ship, Vercel didn't just talk about agents. It shipped for them. The announcements included eve, an open-source agent framework Rauch described as "Next.js for agents," where an agent is essentially a directory with durability, sandboxed compute, and human-in-the-loop approvals built in from the start. Alongside that came Vercel Services, which lets teams deploy backends and frontends as a single coordinated project where components communicate privately without ever touching the public internet. Vercel Connect replaced long-lived credentials with scoped, short-lived tokens and full audit trails, which is the kind of thing enterprises need before they'll run agents in production at scale. Taken together, the stack is a deliberate answer to what Rauch described after ShipNYC as the industry's shift from prototyping AI agents to actually running them in production. Building an agent is now the easy part. The hard part is keeping it secure and reliable once real users depend on it - and observable enough that you can tell when it isn't. That's where Vercel is placing its weight. Production AI teams, he pointed out, now need model routing, cost monitoring, and gateway controls - evals included - not just a favored model contract with one lab. His AI gateway handling a trillion daily tokens reflects exactly that, teams routing across OpenAI, Anthropic, Google's Gemini, DeepSeek, and others simultaneously rather than betting on a single provider. What this means for the incumbent developer tools stack The deployment-composition flip from 3% to 50% in six months is the kind of number that reshapes how you think about every company positioned around the developer workflow. Netlify, AWS Amplify, Cloudflare Pages - all built for a world where humans write code and push it. Vercel built for that world too, of course, but it moved aggressively into the agent layer while its competitors were still refining their edge networks and pricing tiers. Cloudflare has the infrastructure muscle and 300-plus edge locations. AWS has the enterprise relationships. Neither has announced anything resembling eve or the Agent Stack Vercel just shipped. That window won't stay open indefinitely. Rauch hinted at IPO readiness at the HumanX conference in San Francisco in April, though he declined to give a quarter. The company was last valued at $9.3 billion after a $300 million Series F led by Accel in September. That valuation was set before the agent deployment numbers hit their current levels. The gap between the last round and the company's current trajectory is the cleanest argument for why the IPO conversation has started. Frankly, the more interesting question isn't whether Vercel goes public. It's whether the bet holds. The company's pitch rests on agents consistently choosing Vercel as their deployment target, which means staying ahead of the labs themselves, since OpenAI, Anthropic, and Google all have obvious incentives to build hosting infrastructure that keeps developers inside their own ecosystems. Rauch's model-agnostic, open-framework positioning is the direct counter to that pull, but it requires developers to keep choosing the neutral layer over the convenient one. So far, the numbers say they are. Also read: Bankr bot's return to X after suspension exposes how fragile crypto's social layer really is https://startupfortune.com/bankr-bots-return-to-x-after-suspension-exposes-how-fragile-cryptos-social-layer-really-is/ • A Chinese AI model just turned a routine training trick into a Washington national-security crisis https://startupfortune.com/a-chinese-ai-model-just-turned-a-routine-training-trick-into-a-washington-national-security-crisis/ • Two-thirds of all venture capital is now flowing to AI startups and non-AI founders are feeling it https://startupfortune.com/two-thirds-of-all-venture-capital-is-now-flowing-to-ai-startups-and-non-ai-founders-are-feeling-it/