I was just flummoxed by that tweet because it basically handed every other firm a weapon when they're competing on deals, saying, 'Is this the partner you want when things go bad?' And we have a special guest joining us this week, Dev, the returned CEO of MongoDB. And today we discuss OpenAI. They are closing the gap on Anthropic, nearing a $70 billion run rate. Then we have the Factory and the Cognition debacle. Salesforce then buys Listen Labs for $2 billion. Reflection: could they be the US's best hope for an open-source model that can compete with China? This and so much more in today's show.
I've never seen a CRO flip from one competitor to another.
I just think in the AI age, the definition of loyalty has changed.
The reputational damage that you can do in the long term is going to come back and bite you.
When agents pick you, it's a force of nature. Right now, there's a lot that's got to happen between now and the Anthropic IPO in the middle of November.
Ready to go, boys. It is so good to be back and we have a phenomenal guest joining us today. Dev, thank you so much for joining us, dude.
Thank you, Harry. Thanks for inviting me and being part of this amazing group.
Do you know what? Actually, I don't know if you guys know, but Dev gave me feedback that we needed to move on faster from just OpenAI and Anthropic. And we really took it to heart. And so, taking it to heart, I thought we'd start on OpenAI.
OpenAI nears a $70 billion run rate and goes for $30 billion at $1.4 trillion. But the IPO is going to wait until 2027 despite Anthropic going out in the next few weeks, it would seem. What do we think, boys?
The crazy thing to me when I think about the story is, I mean, OpenAI is back, right? $70 billion in revenue. But more importantly, if you look at your portfolio companies and others, I mean, Astro, Soul, and Luna are everywhere, right? Everyone's using them, everyone is using them. And when you think where we started this year, like things, a lot of things in B2B only worked like at the end of last year, and like only Anthropic could get this stuff to work and nobody could. And even coding at the start of this year only worked on Anthropic. And then we went into the summer and it kind of seemed like, you know, OpenAI in some ways was on the ropes, right? Because consumer was the wrong path and they couldn't get back. It's only October or something.
I mean, the rate of change here and the rate at which the OpenAI team got back in the game and made something that is truly epically competitive to the point where people are switching out their models for real because it's better and cheaper. I mean, it's crazy that pace at that scale. So, I mean, it's and it's $70 billion. It says it all, right? So, I mean, you counted them out three or four months ago, not literally, but figuratively. And now, if they launch a truly competitive set of models, LMS, it's like instantly massively competitive. It's crazy. So, maybe it's Captain Obvious, but massive kudos, I think.
Can I ask what happens if Anthropic's numbers for Q3 aren't blowout? Like every team that I speak to has at least deviated some meaningful portion from Anthropic to Codex and everyone says how amazing it is. What happens if they're not blowout numbers? Like the world really just assumes Anthropic always has.
Now, the interesting thing that we should also learn from this and having sold to developers the last 12 plus years is that there's not a lot of loyalty and developers are very quick to switch from one tool to another or frankly use multiple tools. And I think what we're seeing is that developers are quick to find the shiny new toy. I think the question is, you know, what will Claude or Anthropic come out with next and what are the other alternatives. The other thing that's I think interesting also is that what OpenAI has done is shown that when you have essentially a billion people using ChatGPT a week, that's a great distribution channel, but the potentially where the moat is, where the enterprise business is, you know, you have to sign contracts, you have to train thousands of employees, you have to build workflows, and it may be harder to switch the deeper you get into the enterprise. That's kind of a little bit of reaction I had.
Yeah. I guess in that sense, OpenAI's run rate's even more impressive, right? I mean, maybe they were already embedded. Maybe they're already through procurement, but that's pretty fast change in the enterprise too, right?
Can I ask, can I ask you a follow-up question? Sorry to interrupt, Rory. What from your seat, do you think that is a blocker for more progress in open weights models in the enterprise? Just literally the ability to get through procurement, get approved, or do you not? What do you, I mean obviously the explosion this year is crazy, but it is a hurdle to get over. What are you seeing from the MongoDB and Sequoia side?
Well actually from my side, I was just talking to our CIO and we were talking about token costs and token budgets, right? So cost is a big factor and that I think every enterprise has to consider. No one has unlimited budgets. The second thing is IP rights. You know, how we train these models, who gets the data, how proprietary is the data. And I think those two things are also pretty important points when it comes to what models you choose and who you work with.
As you often do, how you put a lot out there. And I think you're right going back to the, forget the valuation stuff which we haven't even discussed, just the raw number, the 70% Q on Q for OpenAI in Q3. I tweeted at the end of Q2 when the OpenAI numbers came out for Q1 and Q2 and the GAAP revenue growth from Q1 to Q2 was 18% quarter on quarter and you did that math and it was that implied, you know, just under 100% year-on-year growth rate which was massive deceleration at a time when Anthropic was still going 10x. So, like at the end of Q2, that was the state of the race. And I don't like to make this a horse race because you need to zoom out, but there's a part of it is a horse race, right? And if you noticed really early on in Q3, all the OpenAI team started really tweeting, 'We're having a killer July.' Do you remember Sarah Frier's comment on that? 'We're having a killer August.' It was clearly that they internalized that, you know, Q2 was the first time Anthropic's GAAP revenue was larger than them. And it just felt like they were pulling away, right? Just, you know, $11 billion versus $6.8 billion, growing 10x versus growing 2x. Suddenly a 70% number, no matter what, we'll talk about Anthropic in a second, but no matter what Anthropic does, at least changes that narrative back if it really is correct, and we don't know, but if GAAP revenue grows 70% Q on Q or even 60% up from 18% Q1 to Q2, that's a massive reacceleration and, you know, just starting with a huge win and the big freaking sigh of relief because, yeah, given the OpenAI forward purchases of just about everything from compute to data centers to land in Virginia, if they didn't grow at hyper speed, a lot of things come unstuck. So that's the first thing, which is if true, amazing for OpenAI, right? And then you write the second thing is, did that come, and this is where Dev's comment on developer loyalty is important, is that did it all come at the expense of Anthropic? In other words, you're asking the question, is this a rotating share game? They accelerate at 70% and then Anthropic slow down markedly, or did they both grow pretty well, in which case, you know, you can lean into the TAM. And you're right, it's now, I mean, if the OpenAI growth rate was the key missing card, you know, the missing data point, now that you've allegedly seen that card, and again, we're assuming the 70% is correct, you're exactly right, the next shoe to drop is, okay, OpenAI, you played a 70% card, now we're going to play our still 10x growth card, or is it going to be a 5x growth, or is it going to be, hey, we've slowed down a lot? It will be super interesting to see where that Anthropic Q3 number comes out. And I won't say it's no accident, but it's worth noting that the timing of the IPO now is such that that number will be in the prospectus, but it will be known at the time they go public.
If we disentangle the revenue from the price in the round, then Rory, as you quite rightly illustrated, I should have done, poor Harry, $1.44 trillion, the largest private round. $30 billion more. I mean, I thought we had dried up all private capital. Clearly, there's more. How do we think about this, Rory?
This would be the first ever venture round above a trillion dollars. That's pleasing. The first ever private trillion dollar round. I remember in the mid-'90s, I remember the first above a billion dollar round and we were like, whoa, that's kind of crazy. Yeah. Moving right along with three orders of magnitude.
Rory, congratulations for becoming the new head of CalPERS pension fund and you are now in charge of directs and you have the chance to do OpenAI at 1.4 or Anthropic as they go public at two. What would you like to do?
Oh, I think you lean to liquidity. It's a lot better to have liquidity than not have liquidity. So let's just say that, two, I mean, so you separately have to value the two companies, which is worth more, but then separate from that, you really asked the question, liquidity versus illiquidity. I would much prefer to be in the liquid stock when the amount of information is as unknown as it is. You either decide you either want the liquid stock or you accept that you're holding for three or four years because that's what the risk of being private is. Yeah, you think it can go public next year and it should be, damn it, it's big enough, but you never know with the world. So, I probably would lean to liquidity at the margin.
We can go to Anthropic's IPO or we can go to the slightly salacious one, which is Factory and Cognition. And I think we should go there. I'm going to just throw my hat in the ring there. So, I'm going to provide some context. Chris Degnan, sales leader, legendary from Snowflake for many years, was a board observer at Factory, spoke to Matan every day according to the founder and was very close to him. And whilst he was very close to him, was interviewing for a role at Cognition according to the founder, Matan, this is, and then took the role at Cognition as CRO and that is when Matan came out and said, 'You're dead to me. You know, this is wrong and how bad of you.' Let's start on that before we discuss Vinod, that Vinod's a whole new topic. How did we analyze this?
Yeah. So I just want to be clear, I know Matan personally, I'm an angel investor, Sequoia's an investor in Factory as well, and MongoDB is a partner with Cognition. The fact that an adviser has access to a founder's confidential plans, you know, there should be clarity on, you know, with the founder that they're thinking about going a different direction, especially if that's a potentially competitive situation. But if I was the CEO, I'd be pretty annoyed to see someone who had access to my plans go to a competitor. Again, I don't know if that had happened. But that would be quite upsetting to me if it did.
On the other side, like as a CEO, it's going to be profoundly frustrating, right? Your adviser goes off and joins a competitor, right? It's happened to us all at some level, right? I would think 95% of the cos we would talk to would say, this is fine, man, this is the way it works. Like CRO is a risky job, there's a lot of turnover, jobs come in fast and a lot of CROs, if it's a hot job, they move very quickly, right? And so I just think that a lot of CEOs would be frustrated. I think 95% of CROs say, this is how I found my last job. Maybe not literally the same facts.
Jason, I'm going to respectfully disagree because remember that CRO has also hired other people to work for them, selling them on the company, selling them on the vision, selling them on how they're going to make money and how it's going to be transformative for their family. If that person were to leave, they basically burnt their reputation with everyone they recruited as well. So I've never seen a CRO flip from one competitor to another. I've seen a CRO go from one company to another company in a different space because they think that's a better opportunity. But I've never seen a CRO, at least that I can remember, that's ever flipped from, you know, one company to a direct competitor of another.
No, I just don't think most CROs, if they were just an adviser, not a, he's not a full-time employee. He's an adviser. He's not even on the board. A board observer. What? And that's important, right? But there is a distinction between a board observer for a little while and a board member. A CRO, not a CTO, not a founder, not a CEO, gets some, maybe somewhat retired as far as I understand it. You know, made a lot of money at Snowflake, kind of, and that's fine. Nothing wrong for many for retiring. Gets what seems to be the greatest job offer. It's going to happen, right? And then says to the other CEO, I got this offer, I'm going to take it. It'd be nice if there was a little more temporal separation. But I just think most CROs that find the job of a lifetime that are just an adviser to another, they wouldn't even see a conflict. The way we think about conflicts as founders, I think a lot of CROs think of it as a game. Like competitors are often really friends with each other, CROs. They see it all. My job is to beat whomever, right? And so any, I just thought founder is shocking, but I don't know if the CRO is an adviser. This is seen as crossing some sort of line.
So again, not knowing all the facts, what I will tell you is that different advisers have different roles. So you can call an adviser for a particular part of the business or particular function and saying, 'Hey, I want your advice on, hey, I'm looking to hire someone in Europe. Can you vet this person or give me some ideas of who I should talk to?' That's very different than someone who's inside the tent looking at your product plans, your board plans, your financial plans, and has real visibility in terms of where you're taking the business. I think if it's the former, then yes, there's no issue. Hey, this person was talking to you, was kind of a, you know, a light relationship and they decided to move on to a competitor. Okay, it happens. You're not thrilled about it, but it happens. Again, not knowing the facts here, but if someone had real insight into the business, real insight in terms of what the product roadmap was like, real insight on how they're competing and winning and losing against their competition and then they suddenly jumped to a competitor, that would be very, very different to me.
The interesting thing is if you think about the employee relationship first and then the advisor relationship. At the employee level, you know, if I have a CRO working for me, right, at any point in time, it's entirely possible that he or she can get a compelling offer from a direct competitor and go over there and literally know everything about what I was doing up until the minute he gave his notice. Right. And Dev, you might think that sucks. That was a pretty shitty thing to do because you were all in for team A and now you didn't go to another sport. You went to team B in the same damn league. It's really shitty, but it happens, right? And, you know, the factory CEO voiced anger, but it's worth pointing out if I was a full-time employee. There's no way for me to tell CEO A, hey, I'm thinking of leaving until I got the job and then I have to give my notice. There's no halfway measure, right? So, to some extent, this is going to happen. If you have employees, they're going to know a lot about what you're doing. And at any point in time, they can go and do the same thing for a competitor. And at that point in time, it's going to suck. And right up until the minute before they sign that offer from your competitor, they had know all your plans, right? So, in other words, something like this can happen and if you could forward communicate. You could say, 'Matan, I've been approached by Scott and I would like to actually pursue it and so I wanted to let you know and be upfront about that because I'm aware I have a position of sensitivity.' Then he is able to know.
First of all, I'm dividing a full-time employee and then we'll talk about advisor because I think one thing Dev is correct. There was implicit, this is where you get explicit and implicit understanding. There was probably some implicit understanding about the advisor that wasn't fully fleshed out on either side. And that's where I think Dev is actually correct. But going back to the full-time employee, I don't think you can, Harry. If your number two said to you, come in and said, 'Hey, I got an offer from Accel. I'm thinking about it. I may go, I may not. If not, I still want you to love me as if I'd never kind of thought about leaving, but I may go.' Right? The truth, and I've done that. But it's generally in an operating role, it can at least be difficult at least half the time.
But that's not the case here, Rory. You know what you're doing.
I know what I'm saying. What my point is merely people are saying, 'Hey, you had a part-time adviser who left and brought all this information, right?' My point is you can lose a full-time direct report employee who brings exactly the same information and even more. So, it can happen, like it happens quite a lot in business. It's not like it's unusual. The thing that's unfortunate here is if the person's just doing kind of part-time advising, you kind of end up in an odd situation where you're disclosing everything. That person's not on the board, so has no fiduciary obligations. He's not employed by the company. There probably, to Dev's point, was some implied statement that said, 'Hey, you shouldn't do this.' But it was never made concrete. Harry, you were shaking your head.
I think you're absolutely wrong. I think as an employee you have very siloed information and you don't have a lot of privileges that a board member would do with the oversight on strategic discussion across the whole company, across all the different functions, and so it's completely different.
How would you, okay, CFO, COO?
I mean, my point is you can lose, if you're the CEO, you can lose one of your top two or three direct reports.
Sure, sure, but if you're the CFO who moves on, you have no real insight into the maybe product roadmap and function.
So do you think that an advisor... Okay. So you would say, okay.
Yeah. So I think, I mean, how is it so different? Half of the Frontier Labs folks are rotating back and forth, right? They go in and out. How they're taking tons of intellectual property and knowledge from each other. I mean, I know it's different, but I think, and Dev can challenge me here, and I wish this were not true, but I just think in the AI age, the definition of loyalty has changed. I think it's changed permanently. I think culturally it's changed. I see it with some of the top executives I've learned, and just one prong. I mean, people move more quickly. People are, the compensation is very different. And I think you just have to expect people will move from competitors. They will go from OpenAI to Anthropic to wherever. It's just that we've changed culturally and it can bother you, but the world changed, you know, we've got to accept it.
Yeah. So what I would say is the following. Like it's very rare for me to see people go to a direct competitor. It has happened and Jason, you're right. In this age, there seems to be more of that happening than not. But it's still quite rare. And I think you also have to understand that these people have reputations that if they're in a leadership position, they've recruited people who work for them, who believed in them, who believe the story. And then for this to get up and leave, and obviously I dealt with this last week, it kind of leaves a bad taste in people's mouth in terms of how really committed were you to this and did you sell me a bill of goods? And in the situation where a CRO or a senior executive is leaving, the question I'd ask is like, okay, I understand there's lots of opportunities out there, but do you have to go to a direct competitor? There's no other company who wanted you. There's no other company that values your skills as much as this company. Like I think the reputational damage that you can do, you know, long term is going to come back and bite you.
Do we even remember next week anymore in the age of...
I think you do. I think actually, Jason, I would say the more things change, in many ways a lot of it stays the same around people, around relationships, how you work with them. And I think people too often use the AI era as like, oh, everything's changed. I disagree. I think how you manage people, how you recruit them, how you develop them, how you hold them accountable, in fact, a lot of these young AI founders struggle with this all the time because some of them have never hired a go-to-market person, they've never hired a finance person, and so they're always asking for advice on what do I look for? And how do I evaluate someone with skills very different than mine? And they're all the same issues I had when I started my first company in 1998. Like it's the same thing all over again. It's just obviously happening maybe more visibly than it ever happened before.
Until recently, my experience was there was always, when you went somewhere else to another company, there was always the rule of two. It's kind of like the Sith but different, right? Wasn't the Sith like the rule of two? I don't know. You could take two people with you. So when you left to go do your own company, you could bring one and Dev would get kind of pissed, okay? And you'd talk about it and then you'd ask permission for the second and there'd be a lot of friction. But if you crossed two, the relationship was broken. Right now I'm seeing folks take like eight or ten with them the first week and like it's not a cultural issue anymore through people that I respect and think are high ethics, right? And some of that I think is the pressure to move so quickly in the age of, but I do think it's related to this. I think it's a symptom of change that the rule of two seems to have died. Was that your rule with your team over the years? Did you informally enforce the rule of two?
People who left and then carpet bombed the rest of the organization trying to follow them, and obviously you know that wasn't a great feeling and, you know, stern words. Let me ask an interesting question. Would you prefer someone, a senior executive, to leave and go to a direct competitor but recruit no one, or someone to leave and go to a totally different business, broadly endeavored developer tools, but say and recruit four or five of his top execs? Which would you prefer to defend, the talent or the knowledge? It's a tough one. I don't know.
It's a tough one. I think it depends on who that competitor was. And if it was like my mortal enemy, I'd be more upset about the former. If it was another company and four or five people left, it is what it is and I'd...
So I do just want to, I would just love to maybe slightly move the conversation along. Because we mentioned the word loyalty and, you know, when we hear the word loyalty, I just think of Vin or Kosla, who led Factory's Series C, led their most recent round and then goes on Twitter and the wording was like, struggling second tier competitor. Absolute demolition job of Matan. I don't know what to say. Was he asleep at the partnership decision? Like, was he aware that he led the last round? I'm genuinely not saying that stupidly. I don't know if he was. Candidly, I was shocked when I saw that. I was just totally flummoxed by that tweet because it basically handed a weapon to every other firm competing with Khosla Ventures. And Khosla got a great team with Keith and a bunch of other people there and basically handed every other firm a weapon when they're competing on deals saying, is this the partner you want when things go bad? I was just totally flummoxed on why he would do that publicly.
Yeah. I mean, there's nothing to add. Some days you just make a, it was a mistake. He will probably look back and say, gosh, I wish I hadn't done that. It's like a weapon in the hands of everyone who competes with them. Wish it hadn't happened. A lot of what's going on here is what are the legal obligations versus what are the implied rules and what are the consequences of breaking the implied rules and what are the non-legal but nonetheless implied rules. It is really hard for venture firms when they're in two directly competitive deals and the only way you get around it is you make all this argument about you love your children equally. You don't say anything in public about one or the other and then you have something like this that crashes through every wall and, you know, it just shows that those situations of two directly competitive investments are very fraught, especially if you're actively involved in the companies. It was very unhelpful for Khosla, very unhelpful for both companies. It's like ick, it was not venture's finest hour. Let's move on.
Maybe it's just, maybe being a Pollyanna, he meant it positively as tough love. Struggling sec, you know, there's a theory like be harsh on the most struggling founders. He thought struggling second tier competitor would be motivating. I yield obviously to Dev and the CEO of MongoDB side here, but I do vaguely remember something about praise in public and chastise in private. I don't think that was the mission here.
You never know. It seems baffling.
Yeah, we've been talking a lot about open models over the last few weeks. We saw Reflection ship. I thought that was a really interesting story. We also discussed, Rory, one of yours before, Intercom obviously selling to Salesforce. Well, we had Salesforce buying Listen Labs too for $2 billion. Of those two, where would you rather go?
I think both are worth some time. I mean, I'd love to hear your thoughts on Reflection because, look, they've been frankly talking about shipping for a long time. And from what we said earlier, it seems to me that being an open-weight US-sourced model is a super interesting strategic position right now. So, if someone can ship something that's near frontier quality, even if it's six, twelve months off, just from the dynamics of the market, that's a nice place. So what are your guys' thoughts?
Yeah, I actually tweeted out this morning. I actually thought that enterprises kind of can have their cake and eat it too. One, you get a US open source model that's near frontier intelligence that's, you know, three to four times more cost effective, i.e., cheaper, and that's music to the ears of any enterprise because there's obviously going to be some hesitation by using Chinese open source models, especially if you're in a regulated industry or industry that deals with a lot of sensitive information. So to me, this was a very noteworthy announcement. Again, I do want to disclose a Sequoia investment in Reflection, but I think it's great for the enterprise. I was at Dreamforce this year, right? And got to actually have a surprisingly large number of conversations with execs. Notwithstanding, you know, the incredible force of nature of Chinese-based open-weight models, when I, nobody I talked to really wanted to use a Chinese source model. Right or wrong, fair or not, everybody that was, they felt like it was under duress for cost, but no one really wanted to use one. Again, right or wrong. So, you know, if they nail this right, it could be a torrent of demand.
I agree. And I think that this also reinforces a point that you don't need the, you know, the frontier level intelligence for every workload. If it's close to the frontier, as Rory said, you know, within six months of the latest models and addresses 90% of your use cases, like especially around, you know, reasoning use cases like coding and agentic workloads, you know, I don't see why enterprises would not kind of want to talk to them immediately.
Yeah, and it's funny, it's, you know, because it ties back to the first topic at one level. So far the evidence has been, you know, when you actually talk, that there hasn't been a massive push because people are lazy and the bills haven't been astronomical and therefore they've just kind of used the frontier model and maybe, you know, downgraded to the Anthropic or OpenAI, you know, less expensive model levels. But I agree. I think the interesting thing now is because the numbers are getting so huge. I mean, if OpenAI and Anthropic reach making $70 billion in ARR, that's $140 billion. That's more than Microsoft taxed for the longest time across both Windows and Office. That's a chunk of US corporate profits, right? And at that point in time, if you're a CFO, you probably come into your CIO and say, 'I don't care that it's easier to just use Anthropic, dude. We need to shave $3 million off this token bill.' Right? So start using GPT for your 20% of your decisions and start using Reflection or something for anything that's not frontier and figure out a way how to use Anthropic for the hard stuff. I would imagine Dev's guys will start to have those conversations and all your customers will. That's why again these growth rates are so interesting right now for Anthropic. Can they still maintain that, you know, unparalleled growth rate in the context of this kind of chipping away? It's like remember The Old Man and the Sea, the guy's bringing back the shark and the fish and then the other fish are gnawing at bits of the body, you know, they're just taking away slugs of revenue from them.
Well, and then you've got Gemini and you've got a load of the other Gemini competitors also in a different way attacking them.
Yeah. Yeah. And again, the forward momentum has been such that all this is like, yeah, thanks Rory, that's super interesting. But did I mention we 10x'd last year and the year before? So, yeah.
The other thing I think it's important to understand, and Jason, I'm curious what you saw at Dreamforce, but I think there's a big gap between where the capabilities are and where actual usage is in the enterprise. I think there's a paucity of real experienced skill sets inside the enterprise. And so I think we're going to see Jevons' paradox kind of come into play where basically as these costs come down, people get more and more comfortable deploying it for more and more use cases across the enterprise. As they get more conversant and scaled around these technologies, they'll start deploying it for more use cases. I see this even at MongoDB. And so I think you're going to see that frankly this is a massive market and I think everyone can be successful. I won't comment on the growth rates of the Frontier Labs versus the open source model, but I don't see this as being a zero-sum game.
To answer this, the captain obvious answer to what I learned at Dreamforce. Everyone I talked to, everyone was just overloaded with the amount of internal demand there is for tokens and whatever it is, it's more than they can imagine. Whether it's being used properly, whether it's being burned for nothing, they're just overwhelmed with whatever the cap is, whatever it is, the demand is probably an order of magnitude more than they've figured out how to surface, right? That may lead to using suboptimal models. It may lead to a lot of waste in some ways, but everyone's got to manage it, which just that and the fear of using China-based models again, right or wrong, is just a huge opening if you can nail it. I'm just, the only thing I will just add just for fun, we can move on, is man, just personally I'm getting more and more skeptical of every published eval. Like of course you can make an eval look great, right? How fast was it? What were really the outputs? How well did it perform on a real workflow? How well did it perform on a mission-critical workflow versus telling you an item was in stock when it wasn't? Which is still kind of mission-critical workflow. I found that, you know, every single eval needs like three asterisks and four daggers next to it today. So I'm skeptical. Beam is as great as they say it is, but I just asked while we're on if I can get on OpenRouter and check before we end the podcast, but it's not there yet. But I would try. But it's not that I don't think these are all great. I just think all the evals are so biased or focused on chess championships that I just, you know, I'm slightly skeptical.
We have Neatron, we have Thinking Machines, we have Reflection. In 12 months' time, what percentage of tokens in open-source flow will go through US versus Chinese?
Well, look, just two things. As Dev made the point in the beginning, developers will change and especially here. This is a lot easier than swapping out your database, right? That could be done, but having recently done it, it's some work, man. I got to tell you, it's not even with an LLM, this is not that, I mean, listen, this is still work. You have to qualify your prompt. You have to redo your workflows. It's not nothing like the internet says, but I could see it being half in 12 months. 12 months is a lot of time. And I think sovereignty plus US-based is really important. I'm not saying it will happen because I don't know that, again, I'm skeptical of the, I'm skeptical the parity is really there versus pretend there. But if you're even three months to Dev's point, it could be half in a year. A year is a long time. But that gives me a lot of confidence. I'm an investor in Fireworks as is Sequoia. And I think the biggest concern I would have as a Fireworks or a Baseten or a Modal investor would be that actually big enterprises are concerned about Chinese models and that would prevent them from engaging with partners like Fireworks or Modal and they then go to Frontier. And so the fact that you say 50% shows me that the levels or quality you think is sufficient enough that it would be that rate that makes me excited.
Yeah, I think the other thing that enterprises are going to care a lot about above and beyond cost is safety. Like being able to trust these systems and I think we're just, you know, getting focused on this. Obviously there's that big conference last week in Washington and I think it's clear that, you know, there's going to be some self-regulating approach to figuring out how to address safety. The challenge is how do you define safety, right? Because if someone knows how to define it well, I'm really curious, but it's beauty is in the eye of the beholder and but we need some way or some mechanism for people to be able to determine, much like a benchmark, and yes people can game a benchmark, is like how safe are these models? And I think the, you know, the more safe these models are with whatever mechanisms have placed to prevent people from doing bad things, the more widely usable they'll become.
Agreed. And I wouldn't guess by the way 50%, Harry. But I think even 10 or 20% would be extremely significant in terms of tokens because if you think about it, the foundation models on OpenRouter I think get a, you know, 20-30% of tokens and get 90% of the dollars. If the US open-weight models are even 50-60% of that in terms of tokens, that's a lot of traction because it'll be by definition the more enterprise-centric, safety-conscious customers doing it. I think that would be a frankly a huge triumph for Reflection and as you point out a real de-risking for all the inference guys.
Ding, ding, ding. We got investment committee. Dev, welcome to the committee by the way. We're thrilled to have you as part of the investment team. You are, sorry to see on the, you know, new firm that is O'Driscoll, Lanin, Stebbings. Jason, we have ElevenLabs who doubled their valuation to $22 billion. Would we do $22 billion?
I think at $22 billion we should consider doing up to 10% of the fund. Time is short. We're at an 18-month deployment cycle. One of our most esteemed partners is going back to run a public company. So we're losing some of the folks on the team. The way I see it, ElevenLabs has done something quite different with actually a fair amount of intense competition. Its lead keeps widening. It is one of the 10 most important underpinnings of agentic applications is their voice applications where they're going and the rate of growth and surprisingly the margins are very impressive. This is a company that by the time we close the friend and family round, which we're hoping to do at $50 billion, right, this one could be approaching a billion in revenue and that's the sort of growth rate that I think it's the deal we've got to do and it's highly differentiated. I won't say they have the highest moat out there in the world, but this isn't GPT-level risk. ElevenLabs has built an incredible team. They have sub-millisecond response rates for voice which has become a key modality and it's tough to find a better asset today. Do I wish it was at 21 or 18? For sure. But I say we go all in.
Can I ask you, Mr. Lanin? You previously said on the show that this would be the year of substitution for voice models and that cost would drive that decision. Have you changed your mind given?
No, because I think ElevenLabs is for the most part, it's always a risk. For the most part, it is covered. It has multiple models. It has covered the low end and the high end. And quality here is in many ways more important than it is in your classic open weights model. We can tolerate a little bit of laxness in the reasoning for non-critical workflows. But when that flower dealer is picking up the phone, it has to work. It has to answer that phone in seconds and it has to be the right flowers. If it fails, if it misunderstands it, if we can't populate this, this product isn't worth it. So there's a limit to price, but I'm shocked at the deal size of ElevenLabs. Shocked at the number of folks that are relatively modest companies paying hundreds and hundreds of thousands of dollars a year happily and our reference checks said they love it. They feel like they're getting a good deal. So, I don't think all models are fungible despite some of the other discussions we've had at the fund. The fact that I did recommend putting maybe $500 million into Reflection and others, this is not a fungible use case today. 12 months, it's hard to predict, but I'm impressed with the team and I don't think anybody's going to catch them here today. It matters. This is where the LLM matters.
10% of the fund goes to ElevenLabs. Well,
I just don't want to do it over two rounds, you know? I don't need to see it at 50 or 100. Dev knows what I mean. Okay. I don't need to see it a second time. I just want to put all the chips in now, get the ownership I can, and move on. Right.
I agree. Plenty more things.
And I've got a friend of mine that's a CRO I want to put on as a board adviser. I think he'll be able to help out and give him some insight.
Okay. Moving from one Sequoia company to another Sequoia company. Sorry Dev, it's not intentional. Rory once said a quote which is like, what was the quote, like whatever game you're playing, just be Sequoia. Do you remember the one I'm talking about?
There's so many quotes. I mean, you know, look, I've been doing this 30 years and when I came in, Sequoia were top and give them huge credit, 30 years later they're still there, you know. So yes, I gave the Gary Lacker football quote if you recollect. So with great surprise, Sequoia made hundreds and hundreds of millions of dollars with Salesforce buying Listen Labs for $2 billion. Now apparently they had offers at $1.5 billion. A major win for Brian Schreier at Sequoia, about $850 million back to investors on $96 million invested.
Sequoia did the best. They led the seed, they led the A, making about a 25x. Ribbit led the B, which 4x'd in 8 months. So, well done, Mickey. What did we think of this, guys?
I think it was a great outcome for a very clear category. In a sense, it's pretty obvious that market research is a very large, fragmented, almost pre-technical industry. There's a bunch of consulting-type companies, and then there's three last-generation companies: Qualtrics, Medallia, and what was SurveyMonkey. All that got to kind of high single-digit, low double-digit billions of revenues in billions of dollars in terms of market cap outcome. So, it's a really good category. It's an existing category with a big-ass spend, and LLMs just do it so much better because, you know, what are you doing in market research? You're kind of asking questions of people, right? And sim- and aggregating and assimilating the answers, and the two things LLMs do well. We just talked about voice. Voice as a modality is a solved LLM problem. And then asking questions and knowing how to, and not just kind of these canned, stupid last-generation surveys where no matter what you say, the survey just asks the next dumb question. You know, it's like, did you like your stay? No, I didn't even check in. And then the next question is, was the bed comfy? Right? That's the last generation. Whereas now with LLMs, you can adjust the questions based on what the respondent is saying. All that means market research can be done so much better with LLMs, and I think Listen Labs was one of the early contenders that did it well and executed brilliantly and, you know, got the real revenue. And I think so, great outcome. I mean, it's kind of one of those when you make the list of use cases for AI and LLMs at the app layer that comes out like a top five, top 10 use case, and there it is.
But you know what, I would love to get Dev's thoughts on this. I don't love, you know, I am pretty much a fan of a lot of Salesforce's acquisitions and I've been champions of anyone. This one I don't really get. It's not that I think it's a bad idea, but I mean, Salesforce is coming up on 60 billion, their target 60 billion for, you know, they're almost at 50. I just don't, I don't know how 20 million of revenue of next-generation AI survey moves the needle that this is, and I'm not saying it doesn't. I mean, scale, you got to move, you either have to tuck in or move the needle, right? Maybe you think about it differently, but to move the needle, it's got to be big. And Salesforce even... So, this is the one where I get the theoretical appeal, but I wonder if this one will just be kind of forgotten in a couple of years.
What I will say is that I think we're going to probably see more of these deals, right? Because I think there's starting to be a difference between a feature and are you building a feature that's masking as a company or a franchise, right? And I think a lot of AI application companies will sell because a clever product on top of someone else's platform, you know, will basically, is frankly a feature. And if they can use the distribution of that larger company, they have the cash and distribution, they're going to take advantage of it. I think the companies that are more durable are the ones that, you know, essentially are creating a data loop where the, you know, the usage creates data that no one else has. The data makes the product better, the product attracts more usage, and you kind of create that virtuous cycle. So, and so if someone tries to copy you, they can copy the features, but they don't have all the interactions that you have. And so, I think the difference here is, you know, when you look at these kind of standalone AI app companies, are they really a feature masking as a company or are they really building a franchise? And I think the founders of Listen decided that this was an offer that made sense for them.
I mean, you'd be pretty pissed off if you're Intercom, wouldn't you? Sorry. He's just like, you've gone through like 15 years of struggle and, you know, just built this business, and then, you know, ABBA's answer to Justin Bieber comes along and 3 years later sells for a comparable number, and you're like... you know, so I know, I know they all made money, but it's like, oh bugger, it was an easier route, wasn't there.
Yes. Of course you think that as a founder. I mean, I don't mean, there's always someone that had the easier route and you do think about it.
Of course you think. Even at, I bet Dev thinks about it once in a while. God, those, I mean, we've done great, but those guys, they had an easier, easier way with it.
The lesson I tell my boys is comparison is a thief of joy. You start comparing yourself to everyone else and you're going to be, spend, be very miserable for the rest of your life.
