A note on beat: I usually cover clinical AI and the patients on the receiving end of it. Cursor is not a health story, so I am writing this one straight, with the same skepticism I bring to a Phase II readout. The patient population here is the working software engineer, and the standard of care, until very recently, was a free editor (most often Visual Studio Code) plus an autocomplete plugin like GitHub Copilot bolted on top.
That is the world Cursor is trying to redraw. The product, built by San Francisco-based Anysphere and founded in 2022, is itself a fork of Visual Studio Code with a deeper layer of AI woven through it [Wikipedia]. It runs on macOS, Windows, and Linux, and the pitch on the homepage is blunt: an editor that "learns how your codebase works, no matter the scale or complexity" [Cursor]. The company's stated mission, repeated across its careers pages and blog, is "to automate coding" [Cursor blog]. That is an unusually direct thing for a developer tools company to say out loud, and it is the frame for everything else.
The bet
Cursor sells a familiar artifact (a desktop code editor) with an unfamiliar promise (that a model sitting inside it can reason about an entire repository and act on it through natural language). The wedge is individual developers who download the app, like it, and bring it into their teams. From there, Cursor sells a Teams plan at $40 per user per month and an Enterprise tier with custom pricing and account management [Cursor docs]. The Stripe case study describes Cursor moving from "hundreds to thousands of enthusiastic Stripe employees" using the editor inside one customer [Stripe]. That bottom-up motion, individual to team to enterprise contract, is the same shape that carried Slack, Figma, and Notion into the Fortune 500, and Cursor is openly hiring a Software Engineer, Enterprise to harden that path [Cursor careers].
The revenue line has moved fast. Stripe, which processes Cursor's payments and therefore has a clean view of the number, reports that Cursor crossed "over $100 million in annual recurring revenue in 3 years" [Stripe]. A third-party tracker pegs the user base at roughly 850,000 in mid-2025, which should be treated as directional rather than audited [Shipper.now].
Why it could be big
The capital behind this bet is now substantial enough that the company has the runway to be wrong once or twice and still recover. Funding has stacked in three visible tranches.
| Round | Date | Amount | Lead |
|---|---|---|---|
| Series B | 2023 | $105M | Undisclosed [Cursor blog] |
| Series C | May 2025 | $900M | Thrive Capital |
| Series D | November 2025 | $2.3B | Accel, Coatue |
Three rounds in roughly eighteen months, led by Thrive, then Accel and Coatue, is the funding pattern of a company that investors believe is compounding into a category leader rather than fighting for a niche. The thesis those checks underwrite is straightforward: if a meaningful share of professional software gets written with an AI agent in the loop, the editor that hosts that loop is one of the most valuable surfaces in enterprise software. Cursor is also training its own coding models and operating "infrastructure that supports billions of requests per day," according to its LinkedIn page, which suggests the company intends to own more of the stack than a thin UI layer over someone else's API [LinkedIn].
The team and traction
Anysphere is based in San Francisco and was founded in 2022 [Wikipedia]. The current job board shows the shape of the company today: open roles for a Software Engineer in ML Research, a Software Engineer for Growth, a Software Engineer for Enterprise, a Design Engineer, and a Product Engineer [Cursor careers]. That mix, research plus growth plus enterprise plus design, is what a company looks like when it is trying to do three hard things at once: keep a model edge, keep the consumer funnel humming, and stand up a real go-to-market for six- and seven-figure contracts. The 262,000-plus follower count on the company's LinkedIn page is a soft signal, but it is consistent with the developer mindshare implied by the revenue ramp [LinkedIn].
The honest counterfactual
The most credible bear case is pricing friction at the individual tier, which is the soil the enterprise motion grows out of. A widely upvoted thread on Cursor's own community forum captures the complaint: a Pro user warned of being "projected to reach your usage limits by 7/30/2025" and questioned the policy directly, with the title of the thread arguing the company risks losing customers if it does not respond quickly [Cursor forum]. Usage-based limits on AI products are genuinely hard, because inference costs are real and variable, and competitors including GitHub Copilot face the same math. The bull answer, supported by the Stripe revenue figure and the Series D, is that the bottom-up funnel is still converting fast enough at the team and enterprise tiers ($40 per user per month and up) to absorb churn at the edges [Cursor docs] [Stripe]. Whether that remains true if a free or cheaper alternative reaches parity on codebase understanding is the open question, and it is the one a Series D investor is paying $2.3B to bet against.
What to watch
Three things over the next twelve months. First, whether Cursor's in-house coding models, the work the ML Research team is hiring for, start to show measurable advantages over the frontier models the product also routes to [Cursor careers]. Second, whether the Enterprise tier produces a named anchor customer beyond the Stripe case study, ideally in a regulated industry where the procurement bar is higher [Stripe]. Third, whether the pricing structure stabilizes in a way that quiets the forum without compressing margins [Cursor forum]. The Series D buys time for all three. It does not settle any of them.
Pulse Raman