JetBrains Net Loss: How AI Coding Agents Upended the IDE Giant

A Record Year That Ended in the Red

JetBrains, the Czech-founded maker of IntelliJ IDEA, PyCharm and a long list of developer tools, has reported the first net loss in its history. The company finished 2025 roughly $14 million in the red, even though revenue climbed to a record $710 million 1. For a business long admired for staying profitable and independent without venture capital, the result marks a symbolic break.

The cause is not shrinking demand for its products but the cost of keeping up. Both reports tie the loss to heavy spending on artificial intelligence, framed as the price of competing with AI-first editors such as Cursor 12. Revenue grew, but margins were consumed by a rush to build agentic tooling.

Where the Money Went

The spending went into three main efforts. Junie, JetBrains' coding agent, now ships across its full IDE lineup and reportedly has about 343,000 users 2. JetBrains Air is a standalone agentic development environment 2. Mellum is a coding model the company has released as open source 2.

This is a large R&D bet for a firm of JetBrains' size: an in-house model, an agent embedded in every product, and an entirely new environment built at the same time. Taken together, these projects suggest the company concluded that waiting for the market to settle was riskier than taking a hit to profit.

The Competitive Squeeze

The two accounts agree on the basic pressure but frame it differently. One centers on Cursor, the AI-native editor, and treats the loss as the real cost of fighting it 1. The other contrasts JetBrains' loss with Cursor reportedly reaching $4 billion in annualized revenue. It then argues the more serious threat comes from somewhere other than rival IDEs: the command line 2.

That second argument deserves attention. Anthropic's Claude Code runs in the terminal, where it can read an entire repository, plan multi-step changes, run shell commands, edit files, execute tests and retry after failures, all without an IDE open 2. OpenAI has rebuilt its Codex CLI into a full-screen terminal interface. It adds parallel agent management, voice input and the ability to control sessions from a phone 2. Google's Gemini CLI runs locally and offers what is described as the most generous free tier available 2.

The argument is that if agents do most of the work, the editor becomes a place to review output rather than write code. That would erode the core value of a premium IDE subscription. One account goes as far as declaring the IDE era over 2. That conclusion reaches further than the financial figures alone can support.

JetBrains' Answer: Become the Control Room

JetBrains' strategic response looks like an attempt to absorb that shift instead of resisting it. In March 2026 it launched Air as an "Agentic Development Environment" in which developers guide agents and review their work rather than writing code themselves 2. Air is agent-agnostic. Through the Agent Client Protocol it supports Codex, Claude, Gemini CLI and JetBrains' own Junie 2.

The product has three parts. Air in JetBrains IDEs handles agent orchestration inside the editor. Air Teams coordinates work between humans and agents. Air Governance covers policy, auditability and cost management 2.

The governance layer may be the most telling choice. Individual developers can switch freely between free terminal tools. Large organizations, however, need oversight of what autonomous agents change, what they cost and who approved their work. By positioning itself as a neutral orchestration and compliance layer above competing agents, JetBrains is targeting a role that the model vendors themselves are less naturally suited to fill.

Reading the Loss

A $14 million loss on $710 million in revenue is a thin deficit, not a collapse. Record revenue shows that customers have not abandoned the company. The loss looks more like a deliberate investment year than a sign of distress, though the scale of competitors' growth explains the urgency.

The harder question is structural. JetBrains built its business on being the best place for humans to write code. If agents in terminals increasingly do that writing, the company has to win on coordination, review and control instead. Air's multi-agent, governance-heavy design suggests JetBrains understands this. Whether enterprises will pay for that layer, or whether the AI labs will bundle it into their own tools, will likely decide if 2025 becomes a one-off dip or the start of a longer decline.