Cursor’s Rapid Growth

Cursor has become one of the most lucrative AI unicorns, with annual revenue projected to exceed $2 billion by 2025, according to a report from Bloomberg. Approximately 60% of its revenue comes from enterprise clients, including new users and existing clients increasing their subscriptions.

The company’s rapid growth has attracted attention from major venture capital firms like Accel, Andreessen Horowitz, and Thrive Capital. In a funding round led by Accel and Coatue last November, Cursor was valued at $29.3 billion, making it one of the most valuable AI startups in the U.S.

Cursor was founded in 2022 by four MIT alumni initially focused on building models to assist mechanical engineers in designing physical parts. However, they quickly pivoted to develop a popular product: a code editor that has gained significant traction. CEO Michael Truell described Cursor as the “Google Docs for programmers,” a collaborative editor where humans and AI improve code together.

Truell and his co-founders have closely followed OpenAI’s advancements in AI since 2020, even before the release of ChatGPT. They recognized the impending explosive growth in this field and were inspired to create an AI coding startup after witnessing the success of Microsoft’s GitHub Copilot.

Cursor’s founders, along with many of its 400 employees, are in their mid-twenties, creating a startup atmosphere reminiscent of an elite university campus. Employees often work late into the night, shower at the office, and live nearby.

This drive and unique product vision led to an annualized revenue of $100 million by early 2025, which soared to over $1 billion by November. The latest funding round pushed the company’s valuation close to $30 billion, making its founders billionaires and placing Cursor among the world’s 20 most valuable private companies.

Challenges from Competitors

However, recent months have seen rising concerns about Cursor’s future. On January 5, 2026, employees returned from the holiday weekend to an all-hands meeting where a slide titled “Wartime” was presented. During the break, employees tested Anthropic’s latest Opus 4.5 and discovered its coding capabilities had advanced to the point where developers no longer needed to check code line by line. Instead, they could issue high-level commands to autonomous agents and receive completed functions, which poses a significant threat to Cursor’s core product concept.

Cursor aims to automate 95% of engineers’ tedious tasks, allowing them to focus more on the creative aspects of coding. Truell has stated, “I believe that soon, a single engineer will be able to build systems far more complex than what a powerful team can currently achieve.” However, if AI can operate independently, the necessity for a code editor becomes questionable, leading to doubts about Cursor’s core product vision.

At the all-hands meeting, Cursor’s leadership warned of upcoming turbulence, indicating that projects might be canceled and priorities could shift. The underlying issue is that AI programming is transitioning from “assisting in writing code” to “agents completing tasks.”

In early 2025, Anthropic previewed a new product called Claude Code to its largest client, Cursor. Claude Code is a command-line tool that allows developers to quickly deploy multiple coding agents. Although it initially seemed that Claude Code would not compete directly with Cursor’s code editor, the situation has changed dramatically. Claude Code’s annual revenue surpassed $1 billion within six months and reached $2.5 billion last month, outpacing Cursor.

Is Cursor Facing Extinction?

With strong competition from Claude and Codex, is Cursor truly on the verge of disappearing? In February 2025, following the release of a more advanced version of Opus, many startup founders began to claim that their teams had abandoned Cursor, believing that model creators like Anthropic and OpenAI would absorb the coding layer themselves.

Jerry Murdock, co-founder of Insight Partners, stated on the 20VC podcast that many companies view Cursor as outdated. In February, over 90 employees at mortgage startup Valon canceled their Cursor subscriptions, opting instead for Claude Code’s powerful agents to achieve complete automation of their workflows, including data migration and bug fixes. Valon CEO Andrew Wang noted that these tasks were completed ten times faster.

Despite the dire situation, some data suggests that the narrative of “Cursor is dying” may not hold up. Recent discussions on the 20VC podcast revealed a significant disconnect between social media sentiment and actual business performance. While developers on platforms like X seem to believe that Claude Code has completely overshadowed Cursor, financial data indicates that Cursor’s annual revenue has reportedly doubled from $1 billion to $2 billion in the past 90 days, with rumors of a new funding round valuing the company at $50 billion.

The disconnect arises from different operational logics in the startup ecosystem versus the enterprise market. Individual developers may switch tools frequently, but large financial institutions like Barclays require lengthy procurement processes and compliance checks before adopting new coding tools. Once they integrate Cursor, they are unlikely to switch to a new tool overnight.

Cursor’s Strategy for Survival

Cursor’s leadership recognizes that the future of software development is not merely about writing code. To adapt to this trend, they are enhancing their R&D capabilities, aiming to surpass competitors like Anthropic and OpenAI in releasing the best coding models through research and proprietary data. They are also prioritizing contracts with large enterprises, as these contracts are more stable than consumer subscription services.

Currently, Cursor’s growth is accompanied by significant anxiety. Insiders report that the company’s focus on revenue tracking has become too fragmented, leading to a halt in daily data updates on their Slack channel. An internal directive reads, “Delete Cursor,” with new tasks named “P0: Build the best coding model,” indicating a shift towards developing intelligent agents similar to Claude Code and Codex.

Last week, Cursor announced a major update to its “cloud agents” product, allowing multiple agents to handle different tasks in dedicated workspaces simultaneously. The leadership believes that enterprises will value products that are not limited to a single model provider, and as model capabilities improve, the market landscape may favor any player, making this an increasingly important issue for developers.

Cursor is also working to reduce its reliance on Anthropic and OpenAI. Their philosophy is that even if competitors invest in larger, cutting-edge models, smaller, specialized coding models trained on proprietary data can still compete effectively.

Currently, about 20 AI researchers are working on Cursor’s Composer model, which is built on powerful Chinese open-source models like DeepSeek, Kimi, and Qwen, modified through additional training and reinforcement learning using Cursor’s proprietary data.

These efforts have shown promise: Composer 1.5 is fast and the second most popular model on the platform, with operating costs significantly lower than Anthropic’s large models. However, for developers, using Composer 1.5 remains costly, with a price of $3.5 per million input tokens, compared to OpenAI’s GPT-5.3 Codex at $1.75 on the Cursor platform.

Community Perspectives

In the developer community, discussions about Cursor’s potential demise are intensifying. Some developers argue that Cursor’s “moat” is quite fragile. One user humorously remarked that Cursor “once shone brightly but only for two minutes,” as its advantages quickly diminished with the release of more powerful models.

Others have provided detailed evaluations from a user experience perspective. A long-time user noted that while they primarily use Claude Code and OpenCode now, they still believe Cursor’s user interface is the best among similar products. However, they feel that Cursor’s commercial strategy missteps have accelerated user attrition, particularly after the company imposed stricter limits on subscription features.

On a technical level, this developer acknowledges Cursor’s core capabilities, particularly in code indexing and embedded search, which outperform traditional tools like grep in accurately locating code context. They suggest that if Cursor could further integrate browser automation capabilities, such as built-in Playwright or Chrome’s MCP interface, its development workflow management could become even stronger. However, they emphasize that the key factors attracting developers to other tools are not just the terminal tools themselves but the performance leaps of models, especially Claude Opus.

Some users have expressed dissatisfaction with pricing and product models. One developer noted that rather than subscribing to Cursor, they would prefer to use the coding tools offered by model creators. They commented, “Through Codex or Claude Code, developers can get more token usage for the same cost and typically receive traffic quotas calculated by the hour or week, with additional charges for more computing power. In this model, traditional monthly subscriptions seem outdated.”

Meanwhile, some developers maintain a cautious outlook on Cursor’s future. One programmer mentioned that as cutting-edge model labs continue to release stronger coding agents, tools like Cursor could indeed face marginalization. They noted that even though over 90% of their project code is generated by models, reliable software development still requires strict oversight, systematic code review, and engineering management, areas where IDEs excel.

However, there are also more pessimistic voices. Some users believe that Cursor has always been a “quick cash-out” project without long-term growth potential.

Additionally, a senior developer recently posted a video on YouTube explaining their reasons for “quitting” Cursor, revealing the engineering challenges faced by this star product amid rapid iterations.

In the latter half of 2024, Cursor won many loyal users due to its seamless compatibility with the VS Code plugin ecosystem and its “black magic” Tab code completion model. However, as versions evolved, Cursor’s reputation began to decline. The developer pointed out that as Microsoft tightened control over core VS Code plugins, Cursor’s underlying compatibility issues arose, even leading to system crashes. Changes in commercial strategy also triggered a crisis of trust:

• Pricing betrayal: The cancellation of unlimited fast requests effectively shifted to expensive API billing without transparent communication.
• UI bloating: The aggressively pushed Agent window in version 2.0 was criticized for its frequent layout changes that disrupted developers’ muscle memory.
• Performance drain: Multiple Chromium processes significantly increased hardware resource consumption, turning what was once a productivity tool into a burden that slowed system performance.

Returning to the terminal and embracing CLI agents prompted this developer to completely abandon Cursor. They found that CLI agents perform more naturally in handling long-term tasks and parallel development, with subscription costs significantly lower than API billing models. They ultimately opted for a minimalist “dark side” setup: Neovim + Tmux.

This shift represents a new trend: since stuffing agents into editors leads to bloating, it is more effective to bring lightweight editors directly into the terminal where agents reside.

The developer concluded that while Cursor remains friendly for beginners seeking a visual interface and single-task focus, it has drifted far from its initial lightweight appeal for advanced developers pursuing extreme performance and automated workflows. In this race of AI coding tools, more flexible and transparent open-source CLI tools are becoming the new favorites for efficiency enthusiasts.