How much did Cognition raise in May 2026?

Cognition announced on May 27, 2026 it raised more than $1 billion at a $25 billion pre-money / $26 billion post-money valuation. The round more than doubled its valuation from roughly nine months earlier. Cognition is the maker of Devin, the autonomous AI software engineer, and now also owns Windsurf (acquired in 2025 after the Google acqui-hire fallout). The company reported $492 million in annualized revenue run-rate and 50% month-over-month enterprise usage growth for the prior six months.

Is Devin better than Cursor or Claude Code?

Different shape. Devin is an autonomous remote agent — you give it a Linear/Jira-style task and it produces a pull request, often with multiple cloud agents running in parallel. Cursor is an interactive AI IDE you drive in real time. Claude Code is a terminal-based agent with filesystem and Git access that runs locally. For long-running, ticketed, enterprise-style work where async parallelism matters, Devin wins. For interactive pair-programming and tight iteration, Cursor or Claude Code wins. Many large teams now use Devin for backlog burndown and Cursor/Claude Code for active development.

How much does Devin cost in May 2026?

Devin starts at $20/month for the entry plan (introduced with Devin 2.0 in April 2025). Enterprise pricing scales by parallel agent count and usage volume. Cognition does not publicly disclose enterprise list prices but customers like Mercedes-Benz, NASA, Goldman Sachs and Santander run multi-seat, multi-agent deployments. For comparison: Cursor Pro is $20/month plus usage, Cursor Ultra $200/month, Claude Code is metered via Claude Pro $20 / Max 5x $100 / Max 20x $200 (changing to a separate credit pool June 15, 2026).

Who owns Windsurf in May 2026?

Cognition AI owns Windsurf. After Google's acqui-hire deal for Windsurf's leadership team in 2025 left the product in limbo, Cognition acquired the remaining Windsurf assets in July 2025. As of May 2026, Cognition is rolling out Devin capabilities to Windsurf in stages — partial integration is live, full convergence of the IDE surface with Devin's autonomous agent backend is still in progress. Cursor remains the dominant AI IDE in seat count, but Cognition's combined Devin + Windsurf revenue is a credible challenger.

Quick Answer

Devin $26B vs Cursor vs Claude Code: AI Coding May 2026

Published: May 28, 2026

Devin ($26B) vs Cursor vs Claude Code: AI Coding May 2026

Cognition just raised $1 billion at a $26 billion valuation on May 27, 2026. The maker of Devin — and now also Windsurf — reported $492M ARR and 50% month-over-month enterprise growth for six straight months. Here’s how Devin actually compares to Cursor and Claude Code right now.

Last verified: May 28, 2026.

TL;DR table

	Devin (Cognition)	Cursor	Claude Code (Anthropic)
Shape	Autonomous remote agent	Interactive AI IDE	Terminal-based agent
Best for	Ticketed/backlog work, async parallelism	Active pair-programming, iteration	Local repos, Git workflows, CLI users
Entry price	$20/mo	$20/mo (Pro)	$20/mo (Claude Pro)
Top tier	Enterprise (custom)	$200/mo (Ultra)	$200/mo (Max 20x)
Model	Multi-model (Claude Opus 4.7 dominant)	Composer 2.5 + Claude/GPT/Gemini	Claude Opus 4.7 / Sonnet 4.6
Parallel agents	Yes (cloud, many)	Yes (Cursor 3 Agents Window, local + cloud)	Yes (subagents)
Notable customers	Mercedes-Benz, NASA, Goldman Sachs, Santander	Most YC + frontier AI startups	Anthropic itself, many indie devs
ARR (estimate)	$492M (disclosed May 2026)	$500M+ (reported)	Bundled in Anthropic revenue
Valuation	$26B (May 27, 2026)	~$10B+	Anthropic ~$900B funding round May 2026

What Devin actually is

Devin is an autonomous software engineering agent. You assign it a task — a bug ticket, a feature spec, a refactor — and it goes off and produces a pull request. The interaction model is closer to managing a junior engineer than driving an IDE.

Devin 2.0 (April 2025) added an integrated IDE for AI-agent collaboration starting at $20/mo. Since then, Cognition has pushed hard on:

Parallel cloud agents — running many Devin instances at once on different tasks
Enterprise governance — audit logs, RBAC, deployment in customer VPCs
Multi-model backbone — Devin uses whatever frontier model is best for the task (Claude Opus 4.7 is heavily weighted in May 2026 for SWE-bench performance)

The funding announcement disclosed that 90% of Cognition’s own production code is written by Devin. That’s a meaningful signal: the team uses what they sell.

What Cursor actually is

Cursor is the dominant interactive AI IDE. Cursor 3 (April 2026) shipped the Agents Window, letting developers run multiple AI agents in parallel inside the editor — both local and cloud. Composer 2.5 (Cursor’s in-house model, with Kimi K2 lineage) launched in May 2026 as a low-cost default for routine work, with Claude Opus 4.7 / GPT-5.5 / Gemini 3.5 Flash on tap for harder problems.

Cursor’s strength is interactive density — autocompletion, inline edits, multi-file refactors that you drive live. It’s the default AI IDE for most YC startups and frontier AI labs.

What Claude Code actually is

Claude Code is Anthropic’s terminal-based agent. It runs locally, reads and writes files, executes shell commands, drives Git, and orchestrates subagents. It’s the “agent that lives in your repo” model.

Claude Code is most popular with:

CLI-first developers (founders, sysadmins, infra engineers)
Anyone who values a thin agent over a fat IDE
Teams already on Anthropic’s stack (Claude.ai + Claude Code + API)

Important pricing change: starting June 15, 2026, programmatic Claude usage (including Claude Code) moves off the shared subscription pool and onto a separate monthly credit pool — $20 / $100 / $200 by tier, metered at full API rates with no rollover. Plan accordingly.

Where each one wins

Devin wins

Backlog burndown. “Here are 40 tickets, work through them” is Devin’s home turf. Cloud parallelism beats local serial.
Enterprise procurement. Mercedes-Benz, NASA, Goldman Sachs, Santander aren’t running individual devs — they’re running fleets of agents with audit trails.
Async management. Assign work, walk away, come back to PRs. Manager-of-AI workflow.
Long-horizon tasks. Devin’s harness is built for hour+ runs without supervision.

Cursor wins

Active development. When you’re actually writing code and need millisecond-level AI feedback, Cursor’s IDE integration is unbeaten.
Model choice. Cursor’s model router gives you the cheapest model that does the job — Composer 2.5 for routine, Claude/GPT/Gemini for hard.
Solo and small-team dev velocity. Most indie founders and small startups ship faster with Cursor than with autonomous agents.
Cost predictability (with Pro $20/mo + per-token).

Claude Code wins

Terminal-native workflows. SSH into a server, run Claude Code, ship a fix. No IDE in the loop.
Anthropic stack lock-in. If you’re paying for Claude Max anyway, Claude Code is included (until June 15 billing split).
Subagent orchestration. Spawning specialized child agents for parts of a task is genuinely good in Claude Code.
Git-heavy work. The Git tooling is the cleanest of the three.

The bigger picture

In May 2026, the AI coding market has three winners coexisting:

Cursor owns the interactive seat (highest user count)
Devin owns the autonomous enterprise seat (highest ARR-per-customer)
Claude Code owns the terminal / power-user seat (highest revenue retention per developer)

These are not zero-sum. Large enterprises (the Cognition customer roster makes this explicit) typically buy all three:

Cursor for active developers
Devin for backlog and async work
Claude Code for SREs, infra, and CLI-heavy engineers

The Cognition valuation of $26B says the autonomous-agent surface is now a real category — not just a feature of an IDE. Expect Cursor and Anthropic to push harder on parallel cloud agents in response.

Verdict

You want autonomous backlog burndown: Devin.
You want fast interactive dev: Cursor.
You want terminal-native agentic Git: Claude Code.
You’re a large enterprise: all three — they’re complementary, not competitive, at scale.

The Cognition funding round confirms the autonomous-agent category is venture-validated. For solo and small teams, Cursor or Claude Code still ships more value per dollar. For 500+ engineering orgs, Devin’s parallel cloud model is now the new default for ticketed work.

Sources: TechCrunch (May 27, 2026), Bloomberg (May 27, 2026), Cognition AI funding announcement, Toolradar Windsurf vs Cursor 2026, Anthropic billing change announcement (May 13, 2026), Cursor 3 release notes.