Cognition
Cognition makes Devin, an AI agent that writes and ships code for engineering teams.
On the record
Checkpoint: Google's Jules autonomous-agent tier reaching feature parity with Devin
The call
Cognition has real revenue and real government logos, but it is racing model labs that can build its own product as a side feature.
Cognition built Devin, an AI tool that writes code for software developers and companies. It has paying customers including government agencies, and it is growing fast. However, Google, Microsoft, OpenAI, and Anthropic have all launched similar tools or acquired the teams that built them, and these companies have more money and users to draw from. The main thing protecting Cognition is that its customers have built their work habits around Devin and some need government-approved security features, but Google is specifically trying to copy this by hiring the people who originally created similar technology. Over the next year or so, if Google launches an AI coding agent that works as well as Devin, Cognition will have to prove its customers stay loyal and keep spending, or it loses its reason to exist as an independent company.
Why
01
AI pressure: Anthropic's Claude Code, OpenAI's Codex, and Google's Jules. Rival coding agents from the frontier labs now ship the same autonomous-agent workflow Devin sells, direct from the model makers Cognition depends on. Could AI do the core job? Yes, likely.
02
Bigger players: Google (Jules, built from the $2.4B Windsurf acquihire) and Microsoft/GitHub Copilot. Google paid $2.4B for Windsurf's founding team to build an in-house agent, and Copilot already sits inside every GitHub repo Cognition wants to win. Could a giant copy it? Yes, easily.
03
What protects it. Devin Desktop workflow lock-in, ITAR/government contracts (Army, Navy, NASA-JPL, Anduril), and proprietary SWE models that cut frontier-model cost dependence.
What would change this call
Google's Jules ships an autonomous-agent tier matching Devin's capabilities within 12-18 months. Would show the $2.4B Windsurf-founder acquihire is translating into direct competitive parity.
Cognition's enterprise net revenue retention rate reported after the Windsurf-to-Devin-Desktop rebrand. Would show whether rebrand disruption caused customers to churn to Cursor or Copilot.
Cognition's gross margin trend after wider deployment of proprietary SWE-1.x models. Would show whether it can escape dependency and pricing pressure from frontier-model providers.
Cognition's enterprise net revenue retention rate reported after the Windsurf-to-Devin-Desktop rebrand. Would show whether rebrand disruption caused customers to churn to Cursor or Copilot.
Cognition's gross margin trend after wider deployment of proprietary SWE-1.x models. Would show whether it can escape dependency and pricing pressure from frontier-model providers.
Key numbers
- Devin ARR ~$492M annualized
- $26B valuation at ~53x revenue
- Google paid $2.4B to acquihire Windsurf's team
Named risks
- Frontier labs (Google, Anthropic, OpenAI) shipping competing autonomous coding agents.
- Valuation doubled from $10.2B to $26B in about eight months at a 53x revenue multiple.
- Reported 80-hour workweek culture and staff terminations following Windsurf acquisition.
- Windsurf-to-Devin-Desktop rebrand risking enterprise churn to Cursor or Copilot.
- AI gross margins structurally lower than SaaS due to frontier-model cost dependency.
Rivals, as read in July 2026
| Rival | Direction | Evidence |
|---|---|---|
| Cursor (Anysphere) | contesting | Reached ~$1B ARR at $29.3B valuation, Nov 2025, more revenue than Cognition's $492M in May 2026 |
| Google (Jules) | gaining on | Paid $2.4B to acquihire Windsurf's founding team, Jul 2025, building an in-house rival agent |
| Anthropic (Claude Code) | contesting | Actively shipping and iterating on a competing autonomous coding agent as of 2026 |
| OpenAI (Codex) | contesting | Shipped competing autonomous coding agent, actively targeting same enterprise accounts |
| Microsoft/GitHub (Copilot) | contesting | Moved Copilot to usage-based AI Credits billing on June 1, 2026 amid competitive pressure |
Questions people ask
- Is Cognition profitable in 2026?
- No profitability figures have been disclosed. Cognition reported roughly $492M annualized revenue as of May 28, 2026, but AI coding tools typically run 45-65% gross margins versus 80-90% for traditional SaaS, and frontier-model costs remain a stated pressure on Cognition's economics.
- Who is Cognition's biggest competitor?
- Cursor (Anysphere) is the closest rival, reporting about $1B ARR at a $29.3B valuation in November 2025, more revenue than Cognition's $492M as of May 2026. Google, Anthropic's Claude Code, OpenAI's Codex, and Microsoft's GitHub Copilot are also directly competing for the same enterprise coding-agent accounts.
- Could AI replace Cognition itself?
- Cognition's own product, Devin, is built to automate the coding work its rivals also target, and CEO Scott Wu said on May 29, 2026 that Devin 'naturally owns tasks end to end' while disclaiming intent to replace humans. Frontier labs like Google, Anthropic and OpenAI already ship competing autonomous coding agents.
- Who owns Cognition and is it public?
- Cognition AI, Inc. is privately held and venture-backed, not publicly traded. Its May 27, 2026 Series D raised over $1B at a $25B pre-money, $26B post-money valuation, co-led by Lux Capital, General Catalyst, and 8VC.
- What is Cognition's valuation and revenue multiple?
- Cognition was valued at $26B post-money as of May 27, 2026, implying roughly a 53x multiple on its reported $492M annualized revenue. That compares to rival Cursor's roughly 30x multiple on about $1B ARR at its November 2025 Series D.
- Did Google try to buy Cognition's Windsurf deal out from under it?
- Yes. In July 2025, Google paid $2.4B to acquihire Windsurf's CEO/co-founder and core research team, while Cognition separately acquired the remaining Windsurf IP, brand, roughly 240+ staff, $82M ARR and 350+ enterprise customers, reportedly for about $250M.
Timeline
- 2026-07-27Launched SWE-1.7 coding model priced at $1.97 per task, undercutting rivals
- 2026-06-04Launched AI Productivity Guarantee offering up to $10M credit for shortfalls
- 2026-06-02Rebranded Windsurf as Devin Desktop with Rust-based local agent
- 2026-06-01GitHub Copilot moved to usage-based AI Credits billing
- 2026-05-27Raised $1B Series D at $25B pre-money, $26B post-money valuation
- 2026-02-26Launched Cognition for Government targeting federal and defense customers
- 2025-11-19Reports of 80-hour workweek mandate, about 30 terminated, 200 given ultimatum
- 2025-09-10Raised $400M Series C at $10.2B valuation, led by Founders Fund
- 2025-07-15Acquired Windsurf IP and staff; Google paid $2.4B to acquihire its founders
- 2025-03-01Raised $300M Series B at $4B valuation, led by 8VC and Lux
Revision history
29 July 2026
First assessed. Losing its protection. The text above is fixed as of this date and is not edited afterwards.
In the same band
Sources
- Cognition Announces $1B Funding Round at $25B Valuation... (2026-05-28)
- Cognition AI scores $400M at $10.2B valuation..., TFN (2025-09-10)
- Cognition's $26B Raise Bets Agent-First Architecture Beats IDE Tools (2026-05-29)
- Cognition, maker of the AI coding agent Devin, acquires Windsurf (2025-07-15)
- Windsurf Joins Cognition After Google Hires CEO in $2.4B Deal (2025-07-16)
- Cognition's Scott Wu says AI coding agents shouldn't replace humans (2026-05-29)
- The 8 best Devin Fusion alternatives in 2026 (2026-07-01)
- This AI startup will give you up to $10 million if its AI tool fails to be productive... (2026-06-05)
- AI coding startup Cognition raises $1B at $25B pre-money valuation (2026-05-27)
- Cognition - Headcount, Funding & Open Jobs (2026-07-19)
- Cursor vs Copilot vs Claude Code (2026-06-23)
- Windsurf Alternatives 2026 (2026-07-14)