JetBrains Warns AI Agents Are Heading for a Cloud-Style ROI Crisis

The Numbers Behind the Warning

The January 2026 JetBrains AI Pulse survey, conducted across 11,000 developers worldwide, paints a striking picture. 90% of developers now use AI tools at work. That number has climbed steadily from 46% in early 2024 to 75% in mid-2025 to where we are today. AI adoption in software development is no longer a trend; it is the baseline.

But the data on coding agents specifically reveals a more complicated story. 22% of developers already use AI coding agents like Cursor, Claude Code, or Codex in their daily work. 66% of companies are either evaluating or planning to adopt agent-based development tools within the next 12 months. That is a massive wave of adoption happening simultaneously across the industry.

The problem is that almost no one is tracking what this costs or what it returns.

The Cloud Parallel

JetBrains draws an explicit parallel to the early days of cloud computing. When AWS launched in 2006, companies rushed to migrate workloads without building the financial controls and governance structures needed to manage spending at scale. The result was years of cloud bill shock, with companies discovering they were spending 2-5x more than expected on infrastructure they barely understood.

AI agents are on the same trajectory, but faster. The costs are harder to measure because they are not just financial. Every agent session consumes API credits, compute resources, and developer time for oversight and review. But no one is tracking these costs in a unified way.

A single complex agent task might consume $2-5 in API credits across multiple model calls. At enterprise scale, with hundreds of developers running dozens of agent sessions per day, that adds up to thousands of dollars per week. Without centralized tracking, these costs hide in individual team budgets and expense reports.

The productivity side is equally murky. Everyone has anecdotes about agents saving hours of work, but systematic measurement is rare. How do you quantify the value of code that an agent wrote correctly on the first try versus code that required three rounds of human correction? What about the hidden cost of developers spending time reviewing and fixing agent-generated code that looked right but contained subtle bugs?

What JetBrains Central Does

JetBrains Central is designed to address this gap. It is a governance and execution platform that provides centralized control over AI coding agents across an organization.

The platform offers three core capabilities. First, cost tracking and budgeting. Central monitors API spending across all connected agents, breaks down costs by team, project, and task type, and enforces spending limits that prevent runaway bills. Finance teams get the same kind of visibility into AI agent spending that they eventually got with cloud cost management tools.

Second, policy enforcement. Central allows organizations to define rules about which agents can be used for which tasks, what data agents can access, and what security constraints apply to agent-generated code. This is critical for companies in regulated industries where AI-generated code needs to meet compliance requirements.

Third, agent coordination. As organizations adopt multiple coding agents, developers increasingly waste time context-switching between tools with different permission models, cost structures, and context management approaches. Central provides a unified interface that abstracts away these differences, letting developers focus on shipping code rather than managing agent configurations.

The Early Access Program launches in Q2 2026 with a limited group of design partners. JetBrains has not yet announced pricing, but the company has indicated that Central will be available as both a standalone product and as part of its existing IDE subscriptions.

Why This Matters Now

First, agent capabilities are expanding rapidly. The launch of Cursor 3 on April 2, 2026, with its parallel agent execution model, means that a single developer can now spin up 5-10 agents simultaneously. Each agent makes dozens of API calls per minute. Multiply that across a 500-person engineering organization and you have thousands of concurrent agent sessions consuming compute resources around the clock.

Second, the shift from copilots to agents changes the cost dynamics fundamentally. Copilots like GitHub Copilot suggest code completions that developers accept or reject in real time. Agents operate more autonomously, running for minutes or hours on complex tasks, consuming API credits continuously. A single failed agent task that goes off the rails can burn through more compute than a developer's entire daily copilot usage.

Third, enterprise adoption is accelerating faster than governance can keep up. The JetBrains survey found that 66% of companies are planning to adopt agents, but fewer than 15% have any formal policies in place for managing agent usage, costs, or security. This gap between adoption and governance is exactly the pattern that produced the cloud ROI crisis.

The Developer Experience Perspective

From a developer's perspective, the governance problem manifests as friction. A developer using Cursor for frontend work, Claude Code for backend refactoring, and Codex for CI/CD integration has to manage three separate tool configurations, three sets of API credentials, and three different billing accounts.

Each tool has its own way of handling context, its own permission model, and its own approach to code review. Developers spend time managing these differences instead of writing code. This coordination overhead can erase the productivity gains that motivated agent adoption in the first place.

JetBrains Central's promise is to reduce this friction by providing a single pane of glass for all agent interactions. Whether you are using Cursor, Claude Code, Codex, or JetBrains' own AI Assistant, Central provides unified cost tracking, consistent policy enforcement, and standardized workflows.

The Competition

JetBrains is not alone in recognizing this problem. GitHub is expanding its Copilot governance features for enterprise customers. Atlassian is building agent management into its platform. Several startups have emerged specifically to address AI agent cost management and compliance.

But JetBrains has two structural advantages. First, its IDEs are where developers actually write code. IntelliJ IDEA, PyCharm, WebStorm, and the rest of the JetBrains IDE family are used by millions of professional developers daily. Integrating agent governance directly into the development environment eliminates the friction of adopting a separate governance tool.

Second, JetBrains has years of experience building tools for professional development teams. The company understands that governance cannot come at the cost of developer experience. If the governance layer adds friction, developers will find ways to work around it, which defeats the purpose.

What Engineering Leaders Should Do Now

Start tracking agent costs now. Even if you are early in adoption, establish baseline measurements for API spending per developer, per project, and per task type. You cannot manage what you do not measure.

Define clear policies for agent usage. What tasks are appropriate for autonomous agent execution? What requires human oversight? What data should agents never access? Writing these policies down before adoption scales is far easier than retroactively imposing rules on entrenched habits.

Pilot multiple agents with measurement. Do not standardize on a single agent tool without comparative data. Run structured pilots where different teams use different tools for similar tasks, measuring both productivity gains and total cost of ownership.

Invest in code review capacity. Agent-generated code needs review, and the volume of code being produced is increasing. Teams that invest in review tooling and review culture now will be better positioned to absorb the increased throughput that agents enable.

Plan for the governance layer. Whether it is JetBrains Central, a competitor, or something built in-house, every organization adopting agents at scale will need a governance platform. Start evaluating options now rather than scrambling when the CFO asks why the AI bill tripled last quarter.

The Bigger Picture

The AI coding agent market is growing faster than most people realize. Cursor raised a $900 million round in 2025 at a $9 billion valuation. Anthropic's Claude Code has become a significant revenue driver. OpenAI is positioning Codex as a core enterprise product. Google, Microsoft, and Amazon are all investing heavily in agent-based development tools.

This level of investment means agents will get dramatically more capable over the next 12-18 months. The agents of late 2027 will make today's tools look primitive. But capability without governance is how you get the cloud crisis all over again.

JetBrains is right to sound the alarm now. The industry has a brief window to build the governance infrastructure needed before agent adoption scales to the point where it becomes a systemic problem. The companies that invest in governance early will reap the benefits of agent productivity without the cost surprises. The ones that don't will learn the same expensive lesson that cloud pioneers did a decade ago.

Sources

• •JetBrains Central announcement: blog.jetbrains.com
• •The New Stack: JetBrains AI agents ROI crisis: thenewstack.io
• •JetBrains AI Pulse survey data (January 2026, 11,000 developers)
• •Business Insider: AI isn't killing coding jobs: businessinsider.com
• •WebProNews: AI boosts developer jobs 20%: webpronews.com
• •MCP adoption statistics: 97M monthly SDK downloads, 5,800+ community servers (March 2026)