Vercel CEO Guillermo Rauch on Building the Internet for Agents

Financial & Valuation Milestones

• Vercel recently closed a $300 million Series F funding round, elevating its valuation to $9.3 billion.
• The company reported approximately $600 million in annualized recurring revenue (ARR).
• Vercel has surpassed 2.5 billion total deployments: taking 10 years to reach the first 1 billion, followed by an additional 1.5 billion in just 10 months.
• The current customer base includes over 485,000 users, ranging from indie developers to enterprise giants like OpenAI, Meta, and Mercedes-Benz.

The Agentic Shift (3% to 60%)

• At the start of the year, only 3% of Vercel deployments originated from AI coding agents; that figure has now risen to 60%.
• This shift represents a transition from human-driven shipping to a "supersonic tsunami" where machines write code for machines to run.
• Vercel is positioning itself as a "universal field" for agentic code, effectively becoming the infrastructure where AI agents deploy their output.

Strategic Pivot to Agentic Infrastructure

• Agent-Centric UX: Vercel is redesigning interfaces from human-centric (HTML/Pixels) to agent-centric (MCP, precise data signals), ensuring agents can consume and interact with services without human friction.
• Headless Capabilities: Companies are moving toward "headless" operations where agents interact directly with back-end systems (e.g., Salesforce, Stripe) via APIs, bypassing traditional UIs to purchase goods or retrieve data autonomously.
• Protocol Evolution: The company highlights the rise of Machine Communication Protocol (MCP) and the Machine Payment Protocol, which allow agents to execute transactions and tool usage directly.
• Revenue Model Shift: The vision is a "lethal combination" where customers buy the Vercel platform and then generate custom software via agents, eliminating the need to purchase third-party SaaS applications.

Security, Sandboxes, and Governance

• Autonomy Risks: The industry is grappling with "agent autonomy," where agents can execute long-horizon tasks without human intervention, increasing the risk of unintended actions or prompt injections.
• Sandbox Challenge: Vercel launched a $1 million "Sandbox Challenge" to identify vulnerabilities in their security environment, resulting in a critical Linux vulnerability discovery by a sophisticated attacker.
• Verification Systems: Vercel is developing "proof systems" to mathematically verify code correctness, aiming for machine-generated code that offers verifiable certainty similar to mathematical proofs (referencing OpenAI's Navier-Stokes proof).
• No Data Retention: The company guarantees zero data retention for enterprise clients, ensuring customer tokens and code are not used for model training by third parties.

Open vs. Closed Model Dynamics

• Market Reversal: Early in the year, traffic to Vercel's AI Gateway was dominated by proprietary models (OpenAI, Google, Anthropic); this has flipped, with 80% of traffic now utilizing open models.
• Reasons for Open Model Adoption: Enterprises favor open models for cost efficiency, data sovereignty, and the ability to swap models via a "Resilient Gateway" with a single line of code.
• Vercel's Stance: The company supports a "best-of-all-worlds" approach, refusing to lock customers into a single model provider to encourage competition and price negotiation.

Hiring and Organizational Philosophy

• New Skill Requirements: Vercel is prioritizing "agentic engineering" skills—the ability to command swarms of agents to build software factories—over traditional individual coding output.
• Communication Value: There is a heightened emphasis on high-quality human communication and "truth-telling," countering the trend of AI-generated content.
• Forward Deployment: Sales and engineering roles are merging into "forward-deployed" positions where engineers migrate enterprise stacks (e.g., Mercedes) in minutes rather than months, using agents to rewrite code instantly.
• Workforce Outlook: Despite the efficiency of AI, the company notes a paradox where CEO and engineering workload remains intense (7/7 schedules) due to the excitement and rapid pace of innovation.

Future Technology: Fluid Compute

• Vercel is defining the next evolution of serverless as "Fluid Compute," where infrastructure expands and contracts like a liquid to handle the massive, variable compute demands of agentic AI.
• This infrastructure is designed to handle heterogeneous workloads, including sandboxes, functions, and static sites, optimizing for the physics of AI token consumption.

Leadership & Personal Philosophy

• Transparency Culture: Founder Guillermo Rauch attributes Vercel's success to "working public," transparency, and open-source contributions, which create a positive feedback loop with AI training data.
• Performance Focus: Rauch emphasizes focusing on foundational health (diet, exercise, basic principles) over "last mile" optimizations, a philosophy he applies to both personal life and business strategy.
• Code as Liability: A core belief is that "code is a liability"; while agents can write code faster, the focus must shift to verifying and securing that code to prevent data leaks or security breaches.