While everyone’s fixated on Anthropic’s potential $900B valuation, the real story is happening in the trenches of AI-powered coding. Replit CEO Amjad Masad’s recent comments about the Cursor deal aren’t just sour grapes—they reveal a fundamental shift in how AI coding tools will be distributed and monetized.
Table of Contents
The Distribution War Nobody Saw Coming
Cursor’s rapid rise isn’t about superior AI models—it’s about understanding developer workflows better than anyone else. While Replit built a browser-based IDE betting on cloud-first development, Cursor embedded itself directly into VS Code, the editor 73% of developers already use daily.
This is Microsoft’s old playbook: don’t force users to switch platforms, become indispensable within their existing workflow. Cursor gets to $100M+ ARR not by building a better development environment, but by making the existing one smarter.
Why Anthropic’s $900B Valuation Changes Everything
Here’s what Masad and most of Silicon Valley missed: Anthropic’s astronomical valuation isn’t about competing with OpenAI—it’s about owning the entire AI application stack. When foundation models become $900B businesses, the real money isn’t in the picks and shovels (like Replit’s IDE), it’s in controlling how those models reach end users.
Cursor likely commands 20-30% revenue share with whatever LLM provider they use. Multiply that across millions of developers, and you’re looking at a $10B+ intermediary business. Replit, despite their technical sophistication, is fighting for scraps in the infrastructure layer while Cursor owns the relationship with the actual humans writing code.
The Business Model Inversion
Traditional SaaS taught us to build complete solutions and charge monthly subscriptions. AI coding tools are inverting this model entirely:
Old model (Replit): Build everything, own the environment, charge $20-50/month
New model (Cursor): Embed in existing tools, charge per AI interaction, take percentage of compute costs
Cursor’s genius is recognizing that developers don’t want another platform—they want their existing platform to be smarter. This creates a paradox: the more successful foundation models become, the more valuable the distribution layer becomes, even if it’s technically “just a plugin.”
What This Means for Every AI Startup
The Replit vs Cursor dynamic is playing out across every industry where AI is being integrated. Companies building “AI-first” platforms are losing to companies making existing platforms AI-enabled.
Look at Adobe vs Midjourney. Adobe didn’t build the best image generation model, but they embedded decent AI directly into Photoshop—the tool designers already know. Result? Adobe’s creative revenue grows 15% year-over-year while standalone AI art tools fight for attention.
The pattern is clear: distribution beats differentiation in AI applications, especially when that distribution happens inside tools people already can’t live without.
The $10 Trillion Prediction
If Anthropic hits $900B, and OpenAI follows suit, we’re looking at $2T+ in foundation model value by 2028. But here’s the contrarian bet: the companies that capture the most value won’t be the model builders—they’ll be the interface owners like Cursor.
Every $100B in foundation model value creates roughly $50B in interface/distribution value. Masad’s mistake wasn’t technical—it was strategic. Replit optimized for building a better mousetrap when they should have optimized for controlling the path to the cheese.
FourWeekMBA AI Business Intelligence — strategic analysis of the moves that matter.