SpaceX Bought Cursor for AI Coding Leadership

Diving deeper into

Why SpaceX bought Cursor

Document
establishing & leading the AI coding as the top AI application layer category versus Cognition’s Devin
Analyzed 4 sources

Cursor won the first big wave of AI coding by owning the place developers already live, the editor, not by building the most autonomous agent first. That made it a broad application layer category, like Copilot but more native and more agentic, while Devin started narrower as a higher autonomy coding worker. Cursor then pulled ahead by turning the IDE into a control room for multiple agents, terminal actions, and web search, which let it absorb more agent behavior without asking users to leave their normal workflow.

  • The revenue pattern shows why Cursor became the category leader. It reached $100M ARR by the end of 2024 with about 360,000 mostly individual developers paying $20 to $40 per month, then doubled to $200M ARR by March 2025 with about 720,000 paying users. That is consumer like adoption inside a B2B tool, which creates a much wider wedge than a pure agent product sold task by task.
  • Product shape mattered. Early Copilot mainly completed code inline. Cursor added Composer, then made Agent Mode the default, with parallel agents, terminal access, and web search inside the IDE. That blurred the line between assistant and agent, while keeping the familiar editor, diff, and approval loop that developers already trust.
  • Devin helped define the agentic end of the market, but the center of gravity moved toward tools that combine autonomy with tight human oversight. In production usage, teams often use Claude Code or Codex for heavy agent work, while Cursor remains a lighter interface choice for users who want an IDE centric workflow. That suggests the winning application layer is the place where people review, steer, and approve agent work, not just the agent itself.

From here, AI coding is likely to consolidate around the system that owns both execution and review. Cursor established the category by making agentic coding feel like a better version of normal development, and the next phase is a fight to keep that control point as terminal native agents, model providers, and full workflow automation push farther beyond the editor.