Interview

Vercel raises $300M primary and $300M secondary at $9.3B valuation, bets on AI cloud and agentic SDK

Oct 27, 2025 with Guillermo Rauch

Key Points

  • Vercel raises $600M at $9.3B valuation, betting that AI agents will replace web pages as the internet's fundamental primitive and that every software category will become AI-native.
  • Vercel's AI SDK is growing faster than anything in the company's open-source history, positioning itself as the default package for developers building agents across startups and enterprises.
  • Vercel launches an agent marketplace and plans selective acquisitions to automate cloud operations, viewing infrastructure incidents as workloads that should route to agents before humans.
Vercel raises $300M primary and $300M secondary at $9.3B valuation, bets on AI cloud and agentic SDK

Summary

Vercel has closed a $600M capital event, split evenly between $300M in primary funding and $300M in secondary, at a $9.3B valuation. The round signals investor conviction that the shift from web pages to AI-native software represents a platform-level transition comparable to the birth of cloud computing.

CEO Guillermo Rauch frames Vercel's strategic pivot around a single thesis: every category of software will become AI-native, and the agent will replace the web page as the fundamental primitive of the internet. Vercel is positioning itself as the infrastructure layer for that shift, which it calls the "AI cloud."

AI SDK as the Growth Engine

The most concrete product signal is Vercel's AI SDK, which Rauch describes as growing "basically vertically" at a rate the company has never seen across its open-source history. He positions it as the React of AI, the default package developers reach for when building agents. Growth is coming from both net-new startups building agent-first products and enterprises leapfrogging conventional digital transformation to deploy agents directly.

The underlying infrastructure demand is changing too. Traditional Vercel workloads were short-burst, CPU-bound rendering jobs. Agentic workloads run for minutes, hours, or days, a fundamentally different compute profile that Rauch calls "workflows" and that requires new cloud services built around them.

Infrastructure Strategy: Hyperscalers Over Neo-Clouds

Despite the emergence of neo-cloud data center operators, Rauch is explicit that Vercel is staying on top of hyperscalers for now. His reasoning centers on two factors: enterprise data proximity, which favors established cloud regions for security and regulatory compliance, and latency, which is directly costly when agents are continuously fetching context and running retrieval-augmented generation. He describes this as a "rich get richer" dynamic that currently favors AWS, Azure, and Google Cloud.

Vercel's abstraction layer means it can route workloads across providers without locking developers into a specific model or infrastructure vendor, a flexibility Rauch views as a durable competitive advantage.

MCP and the ChatGPT Distribution Play

Vercel is betting on MCP (Model Context Protocol) as the successor protocol to HTTP for agentic communication. At Next.js Conf, Rauch demonstrated deploying an app inside ChatGPT via Vercel's MCP integration, including running Doom inside ChatGPT, enabled by OpenAI's recently launched in-app embedding capability. He sees this as the early infrastructure for a distribution model where consumer AI interfaces like ChatGPT surface Vercel-hosted applications, including v0, Vercel's vibe-coding and prototyping tool, directly inside chat sessions.

OpenAI's app platform currently operates in developer mode with relatively open guidelines, though an App Store-style approval process is expected. Rauch views MCP as the protocol layer enabling agent-to-agent communication at scale, citing a Figma-to-Next.js conversion workflow as a concrete near-term example.

Marketplace and M&A

Vercel launched an agent marketplace at its ship.ai conference, offering one-click installation of AI agents and infrastructure services. The marketplace fills gaps in Vercel's native offering, with early integrations including on-demand browser services and AI-assisted security auditing tools like Corridor. Because Vercel holds deployment URLs, source code access, and production data, Rauch argues it occupies a uniquely leveraged position to host an ecosystem of agents that automate cloud operations.

The longer-term vision is that infrastructure incidents, currently handled by on-call engineers, get routed first to agents. Rauch points to the recent US-East-1 outage as the kind of event that should trigger an agent response before a human one.

On M&A, Rauch acknowledged the AI tooling landscape is oversaturated across every subcategory, from QA to testing to security, and signaled Vercel will pursue selective strategic acquisitions. The marketplace, however, is the primary near-term vehicle for ecosystem expansion rather than direct acquisition.