Claude Code hit $2.5B ARR in 9 months. Cursor needed 24.
These three tools are not really competing for the same buyer. Cursor owns the IDE-native developer workflow and enterprise seat deals. Claude Code is a terminal-first agent that's eating GitHub commits at a rate no IDE can match. Windsurf has the deepest compliance stack but spent 2025 losing its CEO to Google and its independence to Cognition. The winner depends entirely on what you're actually trying to do.
Key takeaways
Winners by dimension
Side-by-side
| Cursor | Claude Code | Windsurf | |
|---|---|---|---|
| Free tier | Limited Agent requests, limited Tab completions, no frontier models, no MCPs | Free Claude.ai access with limited usage; Claude Code itself requires Pro ($20/mo) or API credits | Unlimited Tab autocomplete, SWE-1-mini access, browser previews, app deploys, MCP support |
| Starting paid price | $20/mo (Pro) | $20/mo (Pro, billed monthly) or $17/mo billed annually; API from $1/MTok input | $20/mo (Pro) |
| Pricing model | Per-seat with monthly usage credit pool; overage billed at API rates. 1x/3x/20x multiplier tiers. | Per-seat subscription (Pro/Max/Team) plus pay-per-token API. Two separate billing tracks. | Per-seat with daily and weekly quota refresh. Unlimited Tab autocomplete on all plans. |
| Power user max individual plan | $200/mo Ultra (20x usage multiplier) | $200/mo Max 20x plan | $200/mo Max plan (heavy quota) |
| Team plan price | $40/user/mo (Teams) | $20/seat standard or $100/seat premium billed annually (Team) | $40/user/mo (Teams, up to 200 users) |
| Flagship AI agent | Composer 2 (proprietary MoE, parallel subagents, Mission Control, Cloud Agents) | Claude Code CLI (terminal-native, autonomous multi-step, Managed Agents API) | Cascade plus Devin cloud agent (Agent Command Center, Spaces, parallel sessions) |
| IDE and editor surface | VS Code fork (standalone IDE only); JetBrains via ACP partnership | CLI/terminal-native; VS Code and JetBrains extensions; no standalone IDE | VS Code fork plus plugins for 40+ IDEs (JetBrains, Vim, Neovim, Xcode) |
| Enterprise compliance certifications | SOC 2 Type II, GDPR, CCPA; zero data retention with LLM providers; no FedRAMP or HIPAA | SOC 2, HIPAA-ready at Enterprise tier, GDPR, CCPA; available on AWS Bedrock, GCP Vertex, Azure | SOC 2 Type 2, FedRAMP High, HIPAA, ITAR; hybrid deployment option at Enterprise tier |
| On-premises or hybrid deployment | No. Cloud-only on AWS. No VPC or on-prem option. | Via AWS Bedrock, Google Vertex AI, Azure Foundry. No on-prem IDE. | Yes. Hybrid deployment option available at Enterprise tier. |
| Proprietary AI model | Composer 2 (MoE agentic model), proprietary RL-trained Tab autocomplete model | Claude Opus 4.7, Sonnet 4.6, Haiku 4.5 (Anthropic's own frontier models) | SWE-1 (flagship), SWE-1.5 (13x faster than Sonnet 4.5), SWE-1-mini (free tier) |
| Open roles as hiring signal | 68 open roles across sales, engineering, ML research, and field marketing globally | 445 open roles at Anthropic across research, engineering, policy, and enterprise sales | 15 open roles; CEO departed to Google July 2025; company acquired by Cognition |
| Notable integrations count | 47 integrations: GitHub, Slack, Linear, Figma, Datadog, Vercel, Snowflake, Stripe, Shopify, AWS, Azure | 31 integrations: AWS Bedrock, GCP Vertex, Azure, Snowflake, Salesforce, Slack, M365, Google Workspace | 20 integrations: GitHub, GitLab, Figma, Slack, Stripe, SonarQube, Snyk, Playwright, Neon |
Who should pick whom
What we found
Three tools, three fundamentally different bets
Cursor is betting that the IDE is the permanent home of software development and that winning the editor wins the developer. Claude Code is betting the opposite: that the terminal and the agent will displace the GUI editor entirely. Windsurf is caught between those two positions, trying to be an agentic IDE while also being the compliance-first option for regulated industries. Each tool is genuinely best for a different buyer, which is why head-to-head benchmarks mostly miss the point.
The pricing crisis Cursor still hasn't fully resolved
Cursor's June 2025 billing overhaul was a turning point. Users who expected 500 fast model requests per month suddenly found themselves burning through $20 of API credits in a handful of Claude Opus prompts, with automatic overage charges they hadn't consented to (per FinTech Weekly). The result: a 1.7/5 Trustpilot score across 203 reviews. Cursor refunded affected users and apologized, but the incident revealed the structural tension in any tool that resells expensive frontier model compute at a flat rate. Windsurf's quota-refresh model and Claude Code's transparent per-token API pricing both sidestep this problem more cleanly.
Claude Code's real moat isn't the model, it's the commit share
The number that should reframe this entire comparison: Claude Code accounts for an estimated 4% of all GitHub public commits worldwide as of February 2026, with projections of 20%+ by year-end (per SemiAnalysis). That's not a chat interface metric. That's code actually landing in production repositories. Claude Code went from general availability in May 2025 to $2.5B ARR by February 2026, per Anthropic's Series G announcement. Cursor's $2B ARR is impressive. Claude Code's $2.5B in 9 months is a different category of growth entirely.
Windsurf's compliance stack is real, but its stability is not
Windsurf holds FedRAMP High, HIPAA, ITAR, and SOC 2 Type 2 certifications. That combination is genuinely rare for an AI code editor and makes it the only viable option in this group for U.S. government, defense, and strict healthcare deployments. The problem is organizational: in July 2025, OpenAI's $3B acquisition deal fell through, Google poached the CEO and several senior employees, and Cognition acquired the company days later (per CNBC). With only 15 open roles as of April 2026, Windsurf's hiring signal is the weakest of the three by a wide margin.
Who should pick what
Individual developers who want the best daily coding flow in an IDE should pick Cursor Pro at $20/mo. The G2 score, the enterprise case studies, and the JetBrains Jan 2026 survey (18% workplace usage, tied with Claude Code) all point to a product that works. Teams in healthcare, government, or defense should evaluate Windsurf Enterprise first, given its FedRAMP High and ITAR certifications Cursor doesn't offer. Developers who need the sharpest reasoning on hard architectural problems, or who want to run autonomous agents from the CLI, should use Claude Code. The 42% code generation market share Anthropic holds (per Menlo Ventures 2025) exists for a reason.
Sources & references
Every claim in this report was triangulated against 14 third-party sources (analyst reports, developer surveys, news coverage, and pricing pages). Sources are listed below in citation order.
- Cursor AI Statistics 2026: Users, Revenue and Adoption(headline | winners | side_by_side | key_stat)
- 2025 Mid-Year LLM Market Update: Foundation Model Landscape(winners | narrative | side_by_side | key_stat)
- 2025 Stack Overflow Developer Survey(winners | narrative | side_by_side)
- Anthropic raises $30 billion in Series G funding at $380 billion post-money valuation(headline | winners | narrative | key_stat)
- Anthropic Just Hit $14 Billion in ARR. Up From $1 Billion Just 14 Months Ago.(headline | winners | narrative | key_stat)
- Claude Code is the Inflection Point(headline | narrative | key_stat)
- Cursor Faces Backlash Over Pro Plan Pricing Shift(winners | narrative | side_by_side)
- Cognition to buy AI startup Windsurf days after Google poached CEO(winners | narrative | side_by_side)
- Best AI Coding Agents for 2026: Real-World Developer Reviews(narrative | personas | takeaways)
- Cursor vs Windsurf: 7 Tests, 1 Clear Winner 2026(side_by_side | narrative | winners)
- Best AI Coding Tools 2026: Cursor, Copilot and Claude Ranked(narrative | personas | side_by_side)
- Cursor's Dead and Claude Code Killed It(narrative | takeaways)
- Cursor Pricing Page (live)(side_by_side | winners)
- Is Cursor's Pricing Worth It? Total Cost and Competitors 2026(winners | narrative)
Want to share this?
Two tools, same price, wildly different growth curves. Claude Code didn't beat Cursor at the IDE game. It skipped it. Here's what that means for your team's tooling decision:
Want this kind of report on your competitors?
ClientCues runs deep AI scans + side-by-side comparisons every week.
Try ClientCues Free