Vercel CEO Guillermo Rauch on How One Person Plus AI Agents Could Build a $1 Trillion Company

Watch the original interview: YouTube

In a conversation published by Delphi, Guillermo Rauch, founder and CEO of Vercel and creator of Next.js, talks about how to compete with a giant, why trust is the moat for a business in a world of AGI, what it means when your next customer is an agent, and how he is raising six children to think when AI can do almost everything. The interview runs about 55 minutes.

This article is our own summary, with short quotes and timestamps from the video's transcript. Watch the full interview above for everything that was said.

The short version

• Position for the incoming wave. Don't fight the wave that is losing steam; be the platform for the next one (3:04 to 5:38).
• One person plus agents. Rauch hopes to see "the single person trillion dollar company" built on Vercel (6:04).
• Master something first. Don't be a jack of all trades; pick a niche you can master, then expand in concentric circles (10:36 to 15:07).
• Trust is an output, not an input. You earn it through details, security and documentation, not "crunches" like meetups or tweets (18:05 to 21:01).
• Agents are customers now. Build for them, keep your own judgment, and expect them to bring you trade-offs (27:46 to 39:02).
• Verify, don't just generate. The focus shifts from frameworks to verification and evaluation (33:45 to 37:05).
• Delete tools that refuse to speak agent. MCP is "almost like the new HTTP" (43:19 to 45:19).

What he said, by topic

Competing with a giant (0:39 to 5:38)

The host describes migrating a codebase from AWS to Vercel. Rauch says he never saw himself as competing with the solutions he was leaving behind, but as building the next generation. He explains that older technology never fully disappears, "it just loses interest and attention," and his advice to startups is to be positioned for the incoming wave rather than compete with one that is losing steam. He describes Vercel's idea of "self-driving infrastructure," now called agentic infrastructure.

One person doing the work of thousands (5:38 to 6:54)

He points to Instagram's tiny team using Amazon S3 instead of racks of storage. His hope is that on Vercel "we'll see maybe one day the single person trillion dollar company." He adds this does not dismiss teams, but he sees "one person plus agents" doing work that used to take hundreds or thousands of people.

Why enterprises copy startups (6:54 to 10:10)

Enterprises look to innovators for inspiration, so focusing on the world's innovators brings everybody eventually: a "trickle up effect." He spent years pushing headless architectures, where a business exposes APIs and builds many experiences on top. With agents, he says, everyone now needs APIs, CLIs and MCPs, and the world is becoming more generative.

Why "AWS but better" was naive (10:10 to 17:15)

Early on Vercel pitched itself as all of AWS, only better. What helped was mastering a few things and picking relatable examples. With enterprise prospects he starts with the customer experience or business ROI, not the technology. E-commerce is the clearest case: "time is money," and a faster site makes more money. On expanding, he likes concentric expansion from a niche and cautions that unexpected adjacencies are risky for a company's identity, which he calls "brand permission."

Trust is the moat after AGI (17:15 to 24:24)

Rauch says trust is likely one of the most valuable moats for a post-AGI business, but "trust is an output not an input." Brand, trust and developer experience are measured after the work is done. His analogy: you don't get abs by doing crunches. You work on the foundations. For trust that means security, fixing bugs, simplicity and great documentation, not activities like hosting meetups. People build trust from many scattered interactions that add up to a gut feeling. He says design-first craft for developer products was a "pretty wild" idea at the time and is canon now.

Confronting more pain (24:24 to 27:46)

He argues you should be dissatisfied with your own product and ask whether you are "confronting as much pain" as you could: reading customer complaints and cancellation feedback. Vercel has an internal Slack channel called "scream" for bad experiences working there, and he reads every post.

Your next customer is an agent (27:46 to 33:45)

Vercel's aim has been to lower the bar so more people can build while keeping the ceiling high. He prefers the word "builder," and thinks technical and non-technical users are merging. On agents that keep choosing Vercel, he says agents should simply choose the best products and hopes nobody hacks it, including the model labs. He wants agents to become an open, liquid product-discovery system. He also describes a "power and autonomy paradox": the best experience is delegating many choices, but the agent must stay aligned with your intent and push back when needed.

AI writes 2 million tests, then ships garbage (33:45 to 37:05)

He is shifting focus from frameworks to verification and evaluation, creating checks and balances for agents. He favors end-to-end tests close to the user experience, and criticizes agent-written unit tests that check trivial things. He wants people to keep understanding the system, not surrender that.

The future is capital allocation (37:05 to 43:19)

Humans will increasingly decide where to allocate tokens, compute and time, for example into security loops. Agents will come back with trade-offs and ask for more capital. He has invested in a few companies building fully autonomous companies on Vercel, where agents are given seed capital and compete, with success so far only at small scale. On build versus buy, he tells people to think about the full system, including maintenance, and says throw away SaaS when several bespoke tools do the same thing and exist only because "the person doesn't know how to write code."

Delete agent-hostile tools (43:19 to 47:30)

Rauch says "everything is code," and calls the coding agent the browser of AGI. He deletes tools that don't expose data and capability to agents, saying they are like companies that opted out of the early internet. He likens MCP to a new HTTP and wants sites to serve agents digestible text. Agents call APIs at far greater rates than humans, so rate limits, consistency and compute matter. He points to Vercel's fluid compute, which can scale to zero or across many regions.

Raising six kids (49:50 to the end)

He says each of his children has a different personality and that he wants to help them find the overlap of what they love, what is good for the world and what creates value. He wants kids to build a solid foundation and then use agents as "powerups" of intellect and creativity, like the personal computer and Google before.

Chapter guide

• 00:00 Intro
• 0:39 How to Compete With a Giant (AWS)
• 5:38 One Person Doing the Work of Thousands
• 6:54 Why Enterprises Copy Startups
• 10:10 Why "AWS but Better" Was Naive
• 17:15 Trust Is the Moat After AGI
• 24:24 You're Not Confronting Enough Pain
• 25:54 The Slack Channel Called "Scream"
• 27:46 Your Next Customer Is an Agent
• 30:05 Why AI Agents Keep Picking Vercel
• 33:45 AI Writes 2M Tests, Then Ships Garbage
• 37:05 The Future Is Just Capital Allocation
• 43:19 Delete Agent-Hostile Tools Like the Plague
• 44:17 MCP Is the New HTTP
• 45:19 The Internet Runs on APIs That Lie
• 47:30 The Night His House Caught Fire
• 49:50 Raising Six Sub-Agents

What it means for your business

1. Aim at the next wave. Build for where customers are going, not for a migration from the last era.
2. Earn trust through the work. Treat trust and brand as outputs of security, details and documentation.
3. Make your product agent-ready. Expose data and actions through APIs and MCP, and keep them fast and consistent.
4. Invest in verification. Agents can generate a lot; checks and evaluations decide whether it is good.
5. Decide where to spend compute. Capital allocation across tokens, security and speed becomes a human job.
6. Cut tools that don't fit. If a tool can't work with agents, question why you still pay for it.

Video details

• Title: Vercel CEO: How 1 Person + AI Agents Could Build a $1T Company | Guillermo Rauch
• Guest: Guillermo Rauch, founder and CEO of Vercel
• Published by: Delphi on YouTube, October 29, 2026
• Length: about 56 minutes
• Watch: youtube.com/watch?v=W9YfKZaK7KI
• Note: quotes come from the video's automatic transcript and are lightly cleaned. Check the video for exact wording.

Want to put this to work?

We help owners turn AI trends into a plan, a product and a launch. Book a free call or see our services.