Media

Cursor seeks 50 billion valuation

Investors fund AI coding growth as margins depend on enterprise, Google folds web pages into Chrome AI Mode instead of sending users away

Images

Marina Temkin Marina Temkin techcrunch.com
Image description Image description the-decoder.com

Cursor seeks new funding at 50 billion valuation, AI coding boom meets browser-level capture of the web, publishers face traffic loss as answers stay inside platforms

Cursor, the AI-assisted coding company formerly known as Anysphere, is in talks to raise at least $2 billion at a $50 billion valuation, according to TechCrunch, a jump from the roughly $29 billion post-money valuation it received six months ago. The four-year-old startup is forecast to end 2026 with more than $6 billion in annualized revenue run rate, sources told the outlet, after reaching about $2 billion in annualized revenue in February. Thrive Capital and Andreessen Horowitz are expected to lead the round, with Battery Ventures joining and Nvidia also expected to participate.

The numbers matter because they describe a product that is still expensive to run. TechCrunch reports Cursor only recently moved from negative gross margins—where serving users costs more than what they pay—toward slight profitability, helped by a proprietary “Composer” model introduced last November and by routing some requests to cheaper third-party models, including China’s Kimi. Even now, the company is said to be profitable on large enterprise accounts while losing money on individual developer subscriptions. That split is a clue to where the market is heading: the consumer tier becomes a funnel and a training ground, while the business tier becomes the cash register.

Cursor’s attempt to rely less on outside model providers is also a defensive move. If your core feature depends on Anthropic or OpenAI, your main supplier can become your competitor overnight—by bundling similar capabilities into its own tools or by changing pricing and access. The incentive is to build enough proprietary capability to keep the margin and control the roadmap, without paying the full cost of training frontier models from scratch.

At the same time, Google is tightening the loop between search and browsing so that “visiting a website” increasingly happens inside a Google-controlled interface. The Decoder reports that Google is integrating its AI Mode more directly into Chrome: clicking a link from the AI chat can open the target page in a side panel while the conversation continues, and a new “plus” menu can feed open tabs, images, and PDFs into AI Mode as context. Google frames this as convenience—keeping context while reading—but the practical effect is that the AI response becomes the primary product and the website becomes reference material.

Google disputes studies suggesting AI answers reduce outbound traffic, but the direction of product design is clear. When the answer, the follow-up questions, and the next recommendation all happen in the same chat box, publishers lose the one thing that made the open web economically viable: a direct relationship with the reader. Even if a page view is technically counted, the user’s attention stays with the platform’s interface, not the site’s navigation, ads, or subscription prompts.

Put together, Cursor’s enterprise economics and Google’s interface changes show the same shift: value accrues to whoever controls the layer where users work and decide, not to whoever provides the underlying content or model access. For developers, that means the coding environment becomes the gatekeeper. For publishers, it means the browser and search box become the newsroom’s front page.

Cursor’s new round is not final, TechCrunch notes, and Google’s AI Mode changes are rolling out first in the US. But the pattern is already visible: the internet’s “open” layer is being repackaged as a set of inputs for a handful of paid interfaces.