Jared Palmer (Microsoft/GitHub, ex-Vercel) on open source business models, switching costs, and why enterprise shouldn't come too early

Summary

Jared Palmer — now at Microsoft/GitHub after co-creating Next.js and leading product at Vercel — offers a clear-eyed view of open source business models and why enterprise ambition, pursued too early, can kill a project.

The conventional assumption that free open source users will naturally convert into paying customers is, in Palmer's view, a founding mistake. The expectations are incompatible: developers adopt open source because it's free, and pivoting them to a paid tier without laying that groundwork from day one tends to collapse trust and conversion rates simultaneously. The better structure is to define the boundary early — what's open, what's paid — and build toward a symbiosis between the two, rather than hoping one becomes the other.

Vercel itself is the illustrative case. The company is not an open source business. Next.js is open source, but Vercel is a commercial deployment platform. Next.js functions as a demand-generation satellite, not the product: it drives developer attention and creates a feedback loop through internal dogfooding, while Vercel captures value through its hosting infrastructure. Palmer describes the model as "framework-defined infrastructure" — build a framework, deploy with zero configuration, and let the platform handle scale. The analogy he uses: you focus on cooking for four, Vercel worries about feeding the room.

Going enterprise too early

On the question of a barbell strategy — open source broadly, enterprise contracts immediately — Palmer is cautious. Large contracts early can be devastating to momentum because they pull focus away from building the community inertia that makes the product defensible. On TurboRepo, Next.js, and v0, Palmer says enterprise wasn't launched for almost a year, and even then he believes it came too early. The right move was to delay further.

ChatGPT is the reference point: enterprise was held back long enough that employees were already embedded in the product before their companies were asked to pay. By the time the CISO conversation happened, the usage data was already there.

AI adoption beyond software

On how AI adoption spreads outside software engineering, Palmer points to Claude for Excel as a credible wedge into finance. He spent time at Goldman Sachs before moving into software, and argues that banking analysts will be genuinely excited by deep AI integration into Excel workflows. The underlying observation is that coding agents — the same runtime powering coding copilots — are the foundation being adapted vertically. Over the next year or so, he expects model labs to build vertical harnesses for banking, healthcare, and consulting by extending that same core, rather than building category-specific models from scratch.