JetBrains is selling independence as the rest of AI coding picks sides
JetBrains is making a new argument for why developers should care who owns their coding tools. Cursor is tying future model training to xAI’s Colossus infrastructure. Windsurf was split between Google and Cognition last summer. Copilot has always been Microsoft and OpenAI. JetBrains argues it is now the only major independent AI coding-tool vendor left.
“JetBrains is the only independent vendor of the tooling, AI tooling for software developers,” said Mikhail Vink, JetBrains’ VP of business development, in a conversation at Google Cloud Next. “There is no one else.”
Vink walked through the consolidation map as he sees it. Microsoft Copilot is tied to OpenAI. Cursor’s parent, Anysphere, is committed to training future models on xAI infrastructure. Google took key Windsurf talent and a technology license last summer, while Cognition acquired Windsurf’s product, IP, brand, and business.
“There is some kind of lab or some kind of hyperscaler behind every tool,” Vink said. “And JetBrains ends up [being] the only independent vendor where we have this ability and option to work with whatever models and agents we like.”
That neutrality is built into the product. JetBrains’ first-party agent, Junie, defaults to Gemini Flash via a Google Cloud partnership but can also run against models from Anthropic and OpenAI. Internal JetBrains teams use Claude Code, Codex, and Junie interchangeably depending on the task, Vink said. The pitch is that none of those choices have to be permanent.
The reason JetBrains can afford to stay model-neutral, Vink argued, is that it never raised venture capital. The company has been profitable since year one. It has 16 million users and more than 300,000 commercial customers from a 26-year-old IDE business that funded the AI work. “So that funded the current AI journey for us,” Vink said.
JetBrains is leaning into what it announced in March as JetBrains Central, a governance and execution layer for AI coding agents.
That funding model has limits. JetBrains is not training its own foundation model, and Vink said the company has no plans to do so. Instead, JetBrains is leaning into what it announced in March as JetBrains Central, a governance and execution layer for AI coding agents. Central is meant to give enterprise customers a single place to manage who can use which agent, what it costs, and what gets billed. Anthropic, OpenAI, and Google Cloud are all named launch partners.
The independence pitch is tied to a pricing argument. Vink said per-seat pricing does not map cleanly to agentic coding because one task might cost a few cents while another can burn through hundreds or thousands of dollars in model usage, depending on the codebase, context window, and task. Central is JetBrains’ answer to that problem: a control plane for AI governance, agent execution, analytics, and consumption-based billing across whichever models a company chooses.
Independence is also a marketing line. It’s the line available to a vendor that doesn’t own a model. The question is whether enterprise buyers care. Vink argues developers already do. “There is not too much loyalty to a specific model or a specific tool at this point,” he said.
“Developers can use their OpenAI model today, and they can switch to the Anthropic model tomorrow because it’s better.” If teams are swapping models month to month, a vendor that is wedded to one of them is a tax on that switching.
