00:00Cognition has raised over a billion dollars at a $26 billion valuation. Lux Capital, General Catalyst, and 8VC led the
00:08new round.
00:09Cognition CEO Scott Wu is here to discuss the news. We also give an update on the company's AI coding
00:14agent, Devin.
00:16Not just one agent. I think Caroline made a good point. It's an army of agents, and we'll get to
00:20that.
00:20Start with some basics, Scott. A billion dollars, big valuation. Why'd you do that?
00:25Yeah, absolutely. Thank you so much for having me back. A few different reasons.
00:30First of all, the growth that we've seen in the business has been incredible.
00:33I think really across the board, what we're seeing is that AI is doing real work at real companies everywhere.
00:39Every company in 2026 is a software company.
00:42We work with, for example, all the top five health insurers in the United States.
00:46We're seeing folks building way more tooling for all of their care providers.
00:49They're able to cut down on price and cover more people.
00:52We're working with banks on delivering software and making sure to give their customers access to what they need.
00:58We're seeing this with the Treasury and NASA as well.
01:02A couple of reasons for the raise.
01:03I think, first of all, we want to continue to grow aggressively, and this really allows us to do that.
01:08It allows us to scale our compute. It allows us to grow the team and so on.
01:12Second of all, it allows us to stay independent and to really continue on as an independent business, which is
01:17really, really important for us.
01:18It's going well as an independent business.
01:21What's interesting talking to you over, let's say, an aggregate over a period of a year is to track growth.
01:27Yeah.
01:27Right.
01:27So when you first start coming on the show, like beginning of 25, 24, the revenue run rate was a
01:32few million with respect.
01:34Then exactly a year ago, you're kind of at a run rate of about $37 million.
01:40Where's your revenue run rate now?
01:41So we're getting close to $500 million today.
01:43Wow.
01:44As you said, we've only been in business for about two years, and I think a lot of what it
01:47speaks to is just how much demand there is out there for all of this.
01:51And, you know, there's about 30, 35 million software engineers in the world today.
01:54We want to make all of them 10 times more efficient, and then we think there is a lot more
01:57than 10 times more software to build.
01:59There is so much demand, Scott, but there's also a pretty crowded market when it thinks of startups.
02:05And admittedly, you've been talking about how labs are sort of buying these startups, and we think about what DealCurse
02:09has just been doing over with SpaceX.
02:11But I'm interested as to how you see the threat from the big labs in and of themselves.
02:17Yeah.
02:17Yeah, absolutely.
02:18For us, it's actually the other way around.
02:19I think for us, it's, you know, being fully independent and fully neutral is actually the best way for us
02:24to be aligned with our customers.
02:26And so we work very closely with all these labs, OpenAI, Anthropic, but also Google, XAI, and so on.
02:31We have deep relationships with them.
02:33We work with them on research.
02:35And it allows us to work with our customers and provide them the best model for every different use case.
02:39And so Devon is a compound system that works with all of these different models.
02:42And because of that, we're able to be the Switzerland in the equation here.
02:48A Switzerland that sometimes makes the most of the disruption when it comes to talent and the like.
02:52I think about what happened last year, the WindSurf assets, the IP, the brand, the employees after Google took, well,
03:00the key CEO and leadership of that company.
03:02Will more M&A happen?
03:04Is that where some of the new funding will come into perspective?
03:09I'm sure there will be more.
03:10And, you know, there will be different things that come up.
03:12It's for us, you know, what we're personally most focused on is just continuing to grow the business and serve
03:19our customers as best as we can.
03:20You said, well, a heavy emphasis on independence.
03:23Yeah.
03:23And then you called yourselves the Switzerland of this space.
03:26Yes.
03:26You know, what do you think will happen if SpaceX doesn't acquire Cursor is the base question.
03:32But when I speak to, say, the engineering teams at NVIDIA, the reason they like Cursor was the freedom to
03:39swap in and out the underlying model, depending on what your coding objective was.
03:44I suppose that's one reason why they like Devon, right?
03:47But if Cursor becomes a part of SpaceX AI, you know, how do you see that changing the field?
03:53Data is so critically important, right?
03:56And who you are beholden to.
03:58Yeah, absolutely.
03:59No, I mean, it's a great question.
04:00I think in practice, I think there are a lot of great teams working on code, including, of course, the
04:04labs themselves.
04:05I think what we see is that the ecosystem is broad enough and vibrant enough that it makes sense for
04:11there to be different players in different positions.
04:13And so there are going to continue to be first-party products from the labs themselves.
04:18But as we're kind of saying, you know, to your point, I think having the ability to serve each of
04:22the different models and not just the ability, but also the neutrality in terms of being incentivized to just serve
04:27whatever is the best model for each case rather than a single, you know, a single series of models, I
04:31think, is an important position in the space as well.
04:34You would say that Devon is the first AI software engineer.
04:39Yeah.
04:39And we talked about the revenue run rate.
04:41Clearly, that's evidence of momentum and success.
04:44But are you able to sort of give any data on how pervasive Devon is, how it has changed the
04:50structure of different engineering orgs at software companies, at other technology companies?
04:56For sure.
04:56Yeah, I mean, we're seeing this across the board where as teams are adopting Devon and coding agents en masse,
05:03that in practice, they're able to do much more and execute much more aggressively on roadmap.
05:08Is this new code or is it going back over old code bases?
05:11It's both.
05:12And so, as you can imagine, the large majority of work is working on these existing code bases and continuing
05:17to build and add new features and so on.
05:20But even internally at Cognition, for example, more than 90% of the code that we write is written by
05:24Devon.
05:25And so, of course, we're using Devon all day when we're going and building Devon itself.
05:28And even with that productivity, you've been scaling the amount of people you hire, Scott.
05:34Just reflect on what this means in terms of how many software engineers we're going to need.
05:39Like this undergoing anxiety is the disruption that AI causes in all its ways across all these industries.
05:46Yeah, no, I think what we'll see is, of course, the job will evolve over time and we'll see some
05:51of the skill sets change.
05:52But I think we will have far more people doing this and building software and building products, not less.
05:56And, you know, my favorite side on this personally is today there's about 30 or 35 million software engineers in
06:01the world.
06:01Just 20, 25 years ago at the start of the century, it was under 1 million.
06:05And that's clearly come with a massive rise in the amount of software that we've produced.
06:10And I think as we continue to make it more and more efficient, we're actually just going to produce even
06:14more software, not less.
06:15Scott, very, very quick.
06:16What's the goal you've set the team for the balance of the year?
06:19One metric.
06:20Yeah, yeah.
06:21Look, I think for this year, we firmly intend to cross a billion in revenue run rate.
06:26We want to keep going further even beyond that.
06:28And from there, we just want to grow to as many of the companies in the world that we can.
Comments