In July, Cognition bought a company called Windsurf.
Google had been trying to hire the team, but Cognition got there first.
After the acquisition, Cognition’s enterprise revenue doubled.
Doubled.
Most acquisitions slow companies down. This one accelerated everything.
It tells us Cognition isn’t only building features. They’re assembling something bigger. Something that can work inside large, messy, real-world organizations.
Now, Cognition AI just raised $400 million at a valuation that may make your head spin.
$10.2 billion dollars.
But here’s what actually matters for your business: they’re selling something that writes code for you.
Real code.
Production code.
The kind that companies like Goldman Sachs and Cisco are actually using right now.
This feels different from all the AI hype we’ve been swimming through.
Let me show you why we’re going to look at this today.
🧠 Cognition’s $10B Jump
Cognition AI went from stealth startup to $10.2 billion powerhouse by giving every developer a tireless teammate named Devin.
Key Facts
- 💰 70× revenue growth, reaching $73 million ARR nine months after launch
- 🔁 Less than $20 million spent to date—remarkable capital efficiency
- ⭐ Customers already include Goldman Sachs, Cisco, and Palantir
Tell it to “build an inventory tracker for our Shopify shop” in plain English.
Then watch it work.
It figures out what you need. Creates a code repository. Writes the code. Tests everything. Fixes bugs. Deploys to the cloud. All while you focus on bigger problems.
And under the hood it’s name is Devin.
I wrote about Devin almost a year and a half ago. And at the time my prediction was simple “… one day we’ll all have our own personal Devin bots to help with coding our apps and websites.”
Funny enough, that one day didn’t even take a full calendar year.
This isn’t just cool technology. It’s a new way for teams to work.
Goldman Sachs, Cisco, and Palantir already put Devin on their payroll.
These companies don’t throw money at unproven tools. They’re voting with real budgets that this model delivers results.
But what really caught my attention? 👀
Cognition kept its net burn under $20 million while scaling this fast. In today’s venture world, that’s like finding a unicorn riding another unicorn.
How They Pulled It Off
- Product-as-Teammate: Devin behaves like a dependable junior developer – writing, testing, and deploying code while senior staff focus on architecture and user experience.
- Smart M&A: The quiet acquisition of Windsurf doubled revenue and boosted Devin’s model quality without a drawn-out integration.
- Frugal Velocity: While many AI startups burn cash on compute processes, Cognition keeps burn low, which investors see as proof the business model works, not just the demo.
✅ Reality Check
I need to be straight with you about the challenges here.
First, this technology has limits. An AI agent can’t match a senior developer’s intuition on complex architecture decisions. If you’re building software for medical devices or financial trading systems, you still need human expertise at every step.
Don’t be cavalier and put your clients at risk at the sake of novelty.
Second, there are reports of intense working conditions at Cognition.
80hr weeks & layoffs.
That kind of pressure can affect vendor reliability. Before you bet big on any vendor, ask hard questions about their support structure and ability to scale with you.
Third, data governance matters.
Who owns the code the AI generates?
Where does the training data come from?
What happens to your proprietary information?
Get these answers in writing before you start. This is why I fall back to using Replit for my projects.
🖼️ The Bigger Picture
Cognition sits in an interesting spot in the AI landscape.
OpenAI and Anthropic are building the foundational models. They’re the engine manufacturers. Cognition is building the car.
OpenAI just raised funds at a 300B valuation. Anthropic is valued at 183B. These are the companies making AI possible.
But Cognition and companies like it are making AI practical.
They’re taking those powerful models and turning them into tools that actually do work.
Real, billable work.
👀 Have you seen this?
Greg Isenberg on Twitter shared this finding and there’s a LOT of money in this one graph. A breakdown of all the super-groups in how people use ChatGPT. At 800 million Monthly Active Users this is a big insight into user dynamics and realtime trends that can help your business.
🙌🏻 Interested in Joining a Mastermind I Mentor In?
Last week I mentioned the Mentor Collective Mastermind and had some incredible conversations in the replies.
When my mentors Chris & Lori Harder asked me to be a coach in their newest program, the Mentor Collective Mastermind, I honestly just couldn’t say no.
It’s been an incredible group to be part of, and I’ve loved coaching on tech & AI inside. Think:
→ Bi-weekly coaching with Chris & Lori
→ Weekly access to their expert team (me included)
→ A community of entrepreneurs actually scaling
→ Real-time support so you’re never stuck
Doors open Sept 21 & the waitlist is here. Would love to support you inside.
✌️ Let’s Talk
Could an AI agent handle 30% of the work that’s on your plate? 50%? Even 20% would free up your team for more important things.
The companies using these tools aren’t waiting for perfect. They’re starting small, learning fast, and scaling what works. The question isn’t whether AI will change how we build software. It’s whether you’ll be ahead of that change or behind it.
If you’ve already tried AI coding tools, I want to hear about it. What worked? What didn’t? What surprised you?
And if you haven’t tried them yet, what’s holding you back?
This shift is happening whether we’re ready or not. But together, we can figure out how to make it work for our businesses, our teams, and our customers.
P.S. Our community is growing and we’re welcoming new members in weekly. Join us 🔗
