The $60B Fight for Your Code Editor

GitHub Copilot switched to usage-based billing on June 1, 2026 . The headline price stayed at $10 per month for Pro users. The fine print changed everything.

One developer estimated that agentic coding sessions—where Copilot plans, researches, and executes multi-step tasks—routinely consume $30 to $40 per session . A Pro user with $10 per month in credits hits their ceiling in a single working session . The flat-rate era is over. The metered era just started, and the AI coding assistant market hit $12.8 billion in 2026 with 85% of developers using AI tools .

Three names now dominate the space, each with a different bet on how developers will build software. GitHub Copilot leads on raw users with 4.7 million paid subscribers and 75% year-over-year growth. Cursor leads on revenue, hitting $2 billion in annual recurring revenue with over 1 million paying users. Claude Code leads on satisfaction, capturing 46% in the JetBrains April 2026 survey versus Cursor at 19% and Copilot at 9% .

The numbers tell a story about control. Who owns the coding layer, who captures the value when software creation shifts from human-only work to agent-led execution, and who pays when the bill arrives.

Can Flat Pricing Survive Agentic Workflows?

GitHub's explanation is direct: "Copilot simply is not the same product it was a year ago—it now powers far more complex, agentic workflows that consume far more compute" . The old model counted premium requests. The new model counts tokens—input, output, and cached—at rates that vary by model.

Copilot Pro now includes $10 in monthly AI Credits. Pro+ includes $39 . One AI credit equals one cent. Claude Sonnet 5 costs $2.00 per million input tokens and $10.00 per million output tokens through August 31, 2026 . The math gets messy fast.

One developer reported using around 360 AI credits in a single normal development day, intentionally reducing usage of certain workflows just to stay within budget, effectively changing how Copilot is used in practice—from seamless development assistant to something constantly monitored in terms of cost . That's the friction GitHub introduced to fix its unit economics.

The backlash was immediate. Within hours of the change going live, the phrase "What a joke" was trending across GitHub's own community forum, where one of the most-upvoted comments called the new model "a precise double strike" against individual users .

Who's Winning the Developer Satisfaction War?

Market share and satisfaction diverged sharply in 2026. GitHub Copilot's share among professional developers fell from 67% to 51% between late 2025 and 2026, according to the Stack Overflow Developer Survey. In the same period, Cursor debuted at 18% adoption among dedicated AI-native IDEs, while Claude Code reached 10% on its first appearance .

When JetBrains asked developers with more than ten years of professional experience which AI tool they would choose for daily work, 46% picked Claude Code and 9% picked Copilot . That's not a preference gap. That's a trust gap.

Startups pick Claude Code for agentic, multi-step coding tasks—refactor X across the codebase, ship a feature end-to-end—better than competitors . For one-line autocomplete, Copilot and Cursor still win on speed . The tools are splitting by use case, not replacing each other.

Most teams stack them: 70% of engineers use 2-4 AI coding tools simultaneously, with the dominant pattern being Cursor for editing plus Claude Code for complex tasks . The single-tool era never arrived. Developers are building their own workflows from multiple vendors, and Q1 2026 developer surveys found that experienced developers run an average of 2.3 AI coding tools at once .

What Happened to Cursor's $60 Billion Valuation?

Cursor reached a valuation of $29.3 billion, surpassed $3 billion in annual recurring revenue by early 2026, and then saw a June 16, 2026 announcement that SpaceX will acquire Cursor at a $60 billion valuation . The reporting remains ambiguous around xAI and SpaceX involvement, but the headline is clear: Cursor moved into the orbit of the most capital-intensive tech players on earth.

Anysphere, the company behind Cursor, went through an unprecedented revenue trajectory in 2025 for business-to-business software, with revenue doubling every two months between June 2025 and February 2026: $100 million in January 2025, $500 million in June, $1 billion in November, $2 billion by February 2026. No other B2B SaaS company had ever reached $2 billion in ARR in less than 24 months from commercial launch .

Cursor is a fork of VS Code that replaces the editor entirely, keeping Microsoft's familiar layout and extensions intact while making AI features first-class rather than bolted on. The flagship feature is Composer, which proposes multi-file edits in a single pass . Cursor announced a Teams update in June 2026: Standard seats now split usage into two pools—Composer/Auto (first-party Cursor models) and Third-Party API (Claude, GPT, Gemini). A new Premium seat tier at $120/seat/month covers developers running agents all day .

The pricing shift mirrors GitHub's. Flat rates couldn't survive agentic workloads. The difference is Cursor built for agents from day one. GitHub bolted them onto an autocomplete tool.

Are Developers Getting Priced Out of AI Assistance?

Roughly 48% of AI-generated code has security flaws and 75% of senior developers still review every snippet before merging . Reviewing now beats writing for most senior engineers, by an 11.4 to 9.8 hours per week margin in early 2026 surveys . AI didn't eliminate the work. It shifted where developers spend time.

Among developers using AI coding agents, 84% use or plan to use AI tools according to the 2025 Stack Overflow Developer Survey, with 51% using them daily. Yet only 29% trust AI output to be accurate . Adoption is universal. Confidence is not.

GitHub Copilot crossed 20 million users with 1.3 million paid subscribers, making it the most widely adopted AI coding tool by market share at 42%. The agent mode, shipped in late 2025, turns Copilot from an autocomplete tool into a full coding partner . But Microsoft has released Copilot's agent mode and integrated third-party models through Bring Your Own Key, yet the feeling is one of catching up: the legacy product is being extended, not rethought .

The competitive pressure is real. At-work usage of Claude Code grew 6x in under a year, from 3% in mid-2025 to 18% in April 2026, with 75% of startups reporting Claude Code as their primary AI coding tool versus Copilot's 56% in 10,000+ employee enterprises . Startups move fast. Enterprises move cautiously. The tools are splitting by company size.

What Changed This Week

GitHub reopened Copilot sign-ups after a six-week pause, usage-based billing is now live across all tiers, and developers are learning what their actual monthly costs look like under the new model. Cursor finalized its Teams pricing split between first-party and third-party model usage. Claude Code maintained its satisfaction lead while GitHub and Cursor both moved to credit-based systems that punish heavy users. The flat-rate era ended quietly, replaced by a metered model that aligns vendor costs with developer behavior—but introduces friction that wasn't there before.

What to Watch

Microsoft Build 2026 runs June 2-3 in San Francisco at Fort Mason, with expectations for a major Copilot agent refresh and probably another pricing change . Anthropic is likely to ship a Claude Mythos public release this year, which will widen the coding lead further . Cursor will keep building agentic tooling and is the most likely to acquire its way into JetBrains and mobile editors . Watch how usage-based pricing shakes out over the next quarter, whether agentic security incidents become real news, and which tool wins the JetBrains and mobile market. The AI coding wars are no longer about features. They're about who can make the unit economics work without pricing developers out of the tools they've come to depend on.