Anysphere, the company behind the popular AI coding assistant Cursor, is not considering an IPO anytime soon, co-founder and CEO Michael Tuell said on stage at Fortune’s AI Brainstorm conference on Monday.
After reaching $1 billion in annual revenue in November and raising $2.3 billion last month at a valuation of $29.3 billion, Tuell said his company is instead focused on building more capabilities.
For example, he noted that Cursor’s homegrown LLM is tailored to support specific products. Cursor acknowledged the existence of these models in a blog post in November, saying, “Our in-house models currently generate more code than nearly every LLM in the world.”
His comments about these models came up during a Fortune event when the founder was asked how he plans to compete with the LLM manufacturers he relies on, as larger manufacturers OpenAI and Anthropic offer their own AI coding products.
Truell likened his coding product to a “concept car,” but his product is a production car.
“It’s like taking a concept car out with the engine and everything around it, rather than building the entire end-to-end car,” Tuell said. “What we do is take advantage of the best intelligence that the market has to offer from different providers, and we also create our own product-specific models here and there. We take that and build and integrate it together to build the best tools to work with the AI and the final UX as well.”
Cursor’s reliance on competitors and the need to build its own LLM has been the subject of speculation among Silicon Valley VCs since it was reported earlier this year that OpenAI was considering Anysphere as an acquisition target. Anysphere rejected the idea. (This was around the same time Windsurf’s OpenAI deal also didn’t materialize, and the founders ended up joining Google.)
tech crunch event
san francisco
|
October 13-15, 2026
The problem, investors told TechCrunch, was that AI coding editors were losing money due to high fees paid to model makers. In Cursor’s case, instead of selling it, in July it adjusted its price to the model used and directly passed on the API fees that model makers charge users. This change from all-inclusive subscription pricing (and the shockingly high bills some customers faced) caused an uproar among some users.
Asked about the pricing confusion on Monday, Tuell said, “When we started Cursor, we were using Cursor to ask simple questions about JavaScript, and now we’re using Cursor to do hours of work. So the pricing model has had to shift for us and others in the space, and that means moving more towards a consumption model,” he said.
Truell added that one of the tools the company is working on is cost management tools like cloud computing that allow companies to monitor total usage and keep track of the bills their engineers are running.
“We have a whole team in-house dedicated to enterprise engineering and building things like spend management and billing groups and visibility,” he said.
Additionally, Cursor said it is focusing on two key areas in the coming year. One is to handle more complex agent functionality.
“We want to take end-to-end tasks, tasks that are simple to specify but very difficult to actually execute, and have them performed entirely by Cursor, such as bug fixing,” Truell explained.
He specifically wants Cursor to be able to fix the kinds of bugs that are easy to explain but would require “someone to spend weeks and run the code thousands of times” to handle. “We want Cursor to do that end-to-end,” he said.
Another area he mentioned was the idea of ”thinking about teams as atomic units that we serve,” but it wasn’t explained in much detail. This is in contrast to providing services to individual programmers, and should be a hint at how well a company’s business is doing.
In addition to cost monitoring capabilities, Truell said he wants Cursor to be able to handle more parts of the software development lifecycle beyond code creation. He cited Cursor’s code review product as an example, which some customers use to analyze every pull request, whether it’s written by AI or a human. (A pull request is when a programmer submits code for review before it is merged into the main project.)
“So as we add more features like that, you’ll see us start to help the whole team even more,” he promised.
Meanwhile, major competitors are also preparing for a world of agents with complex tasks. Amazon just released a coding tool that already promises to run you for days on end.
Just this week, AI power players including Anthropic, OpenAI, Microsoft, and AWS launched a new consortium under the Linux Foundation to develop open source agent interoperability standards. They also contributed to several major projects, including Anthropic’s hugely popular Model Context Protocol (MCP).
His plans for this year won’t ensure Anysphere leads Cursor’s major model maker competitors. But they need to keep the company in the race.
