JetBrains AI Pulse Wave 2: Claude Code and Cursor tied; Copilot stalls

JetBrains shipped Wave 2 of their AI Pulse survey in April 2026 — the most rigorous developer-tool usage study currently published. n > 10,000 professional developers, fielded January 2026. The headline numbers reset what “the AI coding leader” means.

The numbers

Adoption levels:

• 90% of professional developers use at least one AI tool regularly at work
• 74% use specialized dev-focused AI tools (not just general chatbots)
• 85% report regular use of AI in some part of their workflow

Tool awareness (which tools devs have heard of):

• GitHub Copilot — 76% (still highest)
• Cursor — 69%
• Claude Code — somewhat lower, but rising

Work adoption (which tools devs actually use at their day job):

• GitHub Copilot — 29% (still #1, but stalled year-over-year)
• Claude Code — 18% (tied for #2)
• Cursor — 18% (tied for #2)

Productivity impact:

• Nearly 9 in 10 respondents save at least 1 hour per week
• 1 in 5 save 8+ hours per week

What changed from Wave 1 to Wave 2

The previous Wave 1 survey (mid-2025) had Copilot at 35%, Cursor at 14%, Claude Code at 9%. Wave 2 (April 2026) shows:

• Copilot: 35% → 29% (down 6 points, meaningful stall)
• Cursor: 14% → 18% (up 4 points)
• Claude Code: 9% → 18% (up 9 points — fastest growth in the survey)

Translation: Copilot is losing professional-developer share while Claude Code is doubling. The trajectory is clear even if Copilot still leads on absolute numbers.

Why this matters

Three things this data tells us that other surveys don’t.

1. Copilot’s plateau is real. Anecdotally everyone has been saying “Claude Code is taking over” since mid-2025. Wave 2 is the first methodologically-rigorous data confirming it. Copilot’s 6-point drop in 9 months is the first negative-growth quarter for an AI coding tool that previously had the entire market.

2. The “agentic” workflow is winning. Both Cursor and Claude Code are agent-first products. Copilot (still primarily an autocomplete + chat product) is the autocomplete-first incumbent. The data suggests professional developers are voting with their adoption — agents over autocomplete.

3. The market has fragmented to a real three-way race. Six months ago, “Cursor or Copilot” was the question. Now it’s “Cursor, Copilot, or Claude Code” — and increasingly the answer is “all three, used together.” Multiple-tool stacks are becoming the norm among heavy users.

The productivity numbers are real

The “1 in 5 save 8+ hours per week” stat is the most important non-headline number. That’s a full work day per week per developer. Even at conservative assumptions (a $100K/year developer’s hour costs ~$50), 8 hours/week saved = ~$20K/year of value per developer.

For an organization paying Copilot Business at $19/seat/month ($228/year), the math is overwhelmingly positive. Most “should we pay for AI coding tools” debates ended in Q1 2026 because of data like this. The remaining debates are now about which tool, not whether.

What this means for buyers

If you’re choosing one AI coding tool today:

• Default consumer pick: Claude Code at $20/month — fastest-growing, most-loved by power users, Opus 4.7 launched April 16
• Default enterprise pick: GitHub Copilot — IT-friendly procurement, IP indemnity, multi-IDE coverage, multi-model
• For frontend / rapid iteration: Cursor — Cursor 3 (April 2 launch) + Composer 2

If your org is paying for Copilot: Don’t switch yet. Copilot’s still #1 by adoption and the multi-model integration (Claude, GPT-5) makes it competitive. But evaluate Claude Code as a complementary tool — the JetBrains data suggests dual-tool adoption is where serious developers are going.

If you’re not paying for any AI coding tool: Start with Gemini CLI (free, 1,000 requests/day with Google account). When you outgrow it — usually within 1-2 months — upgrade to Claude Pro ($20) for Claude Code or Cursor Pro ($20) based on workflow preference.

If you run an engineering team:

• The data justifies AI coding tool spend. The numbers are not borderline.
• Don’t standardize on one tool. The “Cursor + Claude Code” or “Copilot + Claude Code” dual-stack is increasingly common.
• Watch for the next Wave 3 in late 2026 to see if Copilot’s stall continues.

The contrarian read

JetBrains is not a neutral observer — they make AI coding tools (JetBrains AI Assistant). Their methodology is solid (n > 10,000, rigorous sampling) but their incentives matter when interpreting the framing. The “Copilot is stalling” narrative serves JetBrains’ competitive interests.

Two caveats to weight:

1. Awareness ≠ adoption ≠ retention. The survey measures awareness and current work adoption. Long-term retention (do users stick with these tools 12+ months later?) is the harder question. We won’t have that data until Wave 3.

2. “Most loved” surveys are emotional. Claude Code’s 46% “most-loved” rating in the previous JetBrains survey reflected user sentiment. Wave 2’s adoption numbers reflect actual usage. They’re correlated but not identical.

What to watch in Wave 3 (late 2026)

• Does Copilot’s adoption continue dropping or stabilize?
• Does Claude Code’s growth continue or hit a ceiling around 25-30%?
• Where does OpenAI Codex land? (It wasn’t separately broken out in Wave 2 in the same way; the 4M weekly active developers figure suggests significant growth.)
• Do dual-tool stacks (Cursor + Claude Code, Copilot + Claude Code) become the norm or stay an edge case?
• What’s the average save-per-week trend? If it climbs from “1 in 5 save 8+ hours” to “1 in 3 save 8+ hours,” the productivity case for AI coding tools becomes essentially unanswerable.

For now, Wave 2 is the cleanest data point we have on the actual professional adoption of AI coding tools. It will be the chart you see cited everywhere for the next six months.

Related:

• Best AI coding tools in 2026 — the ranked picks
• Claude Opus 4.7 launch
• The $20 AI coding tier is collapsing
• GPT-5.4 vs Gemini 3.1 Pro vs Claude Opus 4.7 benchmarks

Questions or corrections? Email Pick Right. Want the full list? See all news.