Cursor vs Copilot: the revenue view

We index 44 businesses with revenue evidence that credit Cursor, and 5 that mention GitHub Copilot. We read all five by hand: exactly one was actually built on Copilot. So the honest headline is not that Cursor earns more, but that Cursor founders narrate their tool and Copilot users do not. Five cases is thin evidence, not a verdict.

Contents

• ·Which tool's builders publish more revenue evidence?
• ·What do the cases on each side actually earn?
• ·Where do the Cursor cases cluster?
• ·What do the evidence grades say?
• ·Which one should you actually pick?
• ·What these numbers cannot tell you
• ·Frequently asked questions

Which tool's builders publish more revenue evidence?

Cursor, by roughly nine to one, and the real gap is wider than 44 against 5. Our tool tags come from what a founder said on camera, so a mention is not a build. Five cases is small enough to audit in public, so here is every one of them.

• ·BlogToPin is the genuine article. Nick had the idea in December 2023 while still holding a 9-to-5 and built it on GitHub Copilot at nights and weekends, because Cursor and Claude Code did not exist yet. It reports $16K MRR, 400+ active subscribers, a $39/mo starter plan and roughly 10% monthly churn.
• ·Atlas matched because the product calls itself an AI e-commerce copilot. Same spelling, different thing.
• ·Windsurf names GitHub Copilot in its risk section, as the competitor with Microsoft's balance sheet behind it.
• ·Claude corporate team training names Copilot as the tool enterprises are already locked into by procurement.
• ·Enderance mentions Copilot to say the opposite. The game was built in Unity before any of this existed: "no Copilot, no Cursor," just forums and Stack Overflow.

Our read: this is a narration gap, not an earnings gap. Copilot is autocomplete inside an editor a company already pays for, and nobody films a build-in-public video about an IDE plugin. Cursor is something people switch to, and switching is a story.

What do the cases on each side actually earn?

Cursor's cases top out at $440K/mo verified and cluster around $15K/mo. The one genuine Copilot build reports $16K MRR, which is a single point from a single founder, not a distribution. Read the evidence column before the revenue column, every time.

Grades, on first use: ✅ verified means the figure was corroborated beyond the founder's word, usually a dashboard read on camera. 🗣 founder-reported means the operator stated it and we could not check it. 📎 creator-relayed means someone passed on another person's number. 🔮 unproven means no figure was disclosed, or the figure is a price rather than a receipt.

Across the Cursor cases that publish a single unambiguous monthly dollar figure, the median is $15K/mo. BlogToPin lands $1K above it. That is the strongest thing this comparison can say, and what it says is: nothing. One case cannot beat or lose to a cohort of 44.

Where do the Cursor cases cluster?

Around shipped screens, run by one person. Of the 44 Cursor cases, 15 are SaaS and 9 are consumer apps, 27 are solo operations, and 16 sit at Tier 1, our copy-this-now grade. The rest spread across 5 scale references, 4 AI websites, 4 platform plugins and 3 directory sites.

The build stories are unusually short, which is the point of the tool. Prayer Lock took three days to write in Cursor with Sonnet, no login and no database, and reports $21K/mo against 58K downloads in six months. Stoppr rebuilt a competitor screen for screen in two and a half weeks and reports $12,000/mo. Profit AI describes its entire build as downloading a client spreadsheet as a CSV, uploading it to Cursor and saying build it, then fixing everything the model got wrong.

The Copilot side cannot be clustered, only described: one solo platform plugin, shipped at nights and weekends by a founder with a day job. Anyone charting five cases as a category is charting noise.

What do the evidence grades say?

That the Cursor side is better supported, and that the Copilot side barely has a side. Cursor breaks down as 10 verified, 26 founder-reported, 4 creator-relayed and 4 unproven. Copilot holds exactly one verified figure, and it belongs to Windsurf, which only names Copilot as a competitor.

Strip the mentions away and the Copilot column holds one founder-reported number. We will not dress that up as a trend. We will say this much: it is an honest number, with a $39/mo plan, 400+ subscribers and a churn figure the founder volunteered, which is more disclosure than most $250K claims carry.

For the ranking across every tool we track, see AI coding tools revenue, ranked in our AI coding tools hub.

Which one should you actually pick?

Pick on the work in front of you, not on our case counts. The evidence supports one narrow claim: people shipping paid products alone have published far more results with Cursor. It supports nothing about which tool writes better code inside a company codebase, because those people do not publish.

You are shipping a paid product alone

Start with Cursor, because the analogues exist and you can read them. Gravl at $440K/mo keeps the model scoped to named files rather than letting it roam the project. Cursor Directory at $34,000–35,000/mo was a three-hour weekend build. Copy the working method, not the revenue.

You write code inside a company codebase

Stay on whatever your employer already licenses, most likely Copilot, and ignore this page's counts entirely. Nothing in five cases argues for a procurement fight. BlogToPin proves a salaried developer can ship a $16K MRR product on Copilot at nights and weekends, which is the only claim our data actually supports.

You are choosing what to build, not what to build with

Then the tool is the least interesting variable here. Every case above earns from a buyer, a channel and a price, and none of them earn from an editor. Look at Sync2Sheets at $9K/mo with ~90% margins: the leverage is the niche, not the IDE.

What these numbers cannot tell you

They cannot tell you a failure rate, and they cannot tell you which tool a working developer prefers. We index businesses that published a number, almost all of them on YouTube. Our read is that Copilot's audience skews toward salaried engineers, whose employer's codebase is not a content channel, and that alone could explain the entire 44-to-5 gap.

Attribution is loose in both directions. Three of the five Copilot cases also sit on our Cursor list, because founders use several tools and say so. A case tagged Cursor usually means mostly Cursor, plus whatever else was open. For what each tool does, GitHub's own Copilot documentation and our explainer on what Cursor AI is are better sources than any revenue table, including this one.

If you want the deeper per-tool reads, see apps built with Cursor and what is Windsurf AI. More in AI coding tools.

Frequently asked questions

Does Cursor make its users more money than GitHub Copilot?

The evidence cannot answer that. We hold 44 Cursor cases with revenue evidence and one genuine Copilot build, so there is no Copilot distribution to compare against. The single Copilot case reports $16K MRR, which sits just above the $15K/mo median of the Cursor cohort. One case beats nothing and loses to nothing.

Why do you only have five Copilot cases?

Because our index is built from founders who published their numbers, mostly in video interviews. We think Copilot's audience skews toward employed developers in private codebases, and that work does not become a revenue breakdown on YouTube. The scarcity reflects who talks, not who earns. We would rather show five and say so than pad the list.

Can you build a paid product on Copilot alone?

Yes, and one founder documented it. BlogToPin was built on GitHub Copilot at nights and weekends from December 2023, before Cursor and Claude Code existed. A scrappy version came together in seven days, most of it was rewritten, and the shippable product took about two months. It now reports $16K MRR on a $39/mo starter plan.

What does third-party verified mean here?

It means the revenue figure was corroborated beyond the founder's own claim, most often a dashboard read on camera or a store payout screen. Founder-reported means the operator stated it and we could not independently check. Creator-relayed means a third party passed on someone else's number. Unproven means nothing was disclosed, or the figure quoted is a price rather than revenue.