00:00Your latest investment has been in a company called Agora. We actually had them on the show
00:03last week. And that's on the AI legal sector and thinking about the applications of AI. Is that
00:08where you're thinking about there's still room to run in terms of valuation and value add?
00:14Well, if you look at the, well, first of all, it's great to be back. Thanks. Thanks for having
00:19me. If you look at the history of technology and computing, the application layer is actually
00:23where most of the value has been created. And if you look at the last three or four years,
00:28so much of the investment has gone into the foundational models and the infrastructure
00:32layer to now power the application layer in AI. And we're just on the cusp of that. And so I
00:38think
00:38you are going to see a lot of funding, but also really incredible revenue growth and opportunity
00:44at the application layer. What's so interesting is we had Gradient on, for example, who just raised
00:49funds and they're very much early stage investors, but they feel that there is a bubble at the moment
00:54in the foundational layer and maybe other parts of AI. But then when you speak to other
00:58key VCs and look, we've had them raising enormous funds, you've got Thrive, you've got talks about
01:02General Catalyst. They're saying, like, you're not getting it. Just the value that we're about to say,
01:08we are on the very beginning of this. Arun, where do you sit on the bubble mentality?
01:12Yeah, I think you kind of have to hold two things at once in your mind, which is a really
01:17difficult
01:18thing. I think there's an incredible opportunity and we're seeing companies in our portfolio that are
01:23growing faster than we've ever seen before. I mean, the numbers are truly astronomical.
01:27And I think oftentimes investors and other people think about this as the next evolution of software.
01:33I actually think it's a different paradigm. It's the reinvention of work. And I think legal is a
01:38really good example of this. If you look at legal software, that's a five to ten billion dollar revenue
01:44market globally. It's a good size, but it's not huge. But if you look at the amount of spend in
01:50legal that goes towards work and labor, that's a trillion dollars globally. And so you have to think
01:57about AI and particularly as we move towards agents that are executing work end to end as a function of
02:04the labor market. Because some of that will transform to agentic work. And so 5%, 10%, 15% of that,
02:12which is
02:12not an unreasonable percentage, is still a hundred billion dollar market. And so I think you're seeing
02:17this divergence in the market. People that think about this as productivity enhancers for humans, just
02:23an evolution of software, and others that think about this as a reinvention of work. And I think that's
02:29where you see a divergence. That said, there is going to be a fallout in the market. And there are
02:36bubble
02:36tendencies that are that are forming. I think there are going to be fewer companies that are actually
02:41winners in this next era. And we can talk about that. So many questions. I mean, like, I'm never going
02:46to turn
02:47down the opportunity to get into the impact of agentic on labor economics. We do that a lot. But the
02:54question that I have for you, right, if you take any of the companies that you've backed and this
03:00like application layer, the criticism or accusation is that they're just wrappers for something more
03:06substantive underneath, usually anthropic at this moment in time. What's the counter argument to that,
03:12Arun? Yeah, this is the big debate right now, whether the horizontal models are going to own
03:18everything. And I should note, we're investors in anthropic, but we're also investors at the
03:23application layer. If you look at companies like Cursor or Basis or Campfire or Lagora,
03:29these are really interesting applications that are in incredible verticals like legal and finance
03:35and health care. And our belief is there's room for both to be really, really successful.
03:41Obviously, the models are doing incredibly well, innovating at a rapid pace. Intelligence is
03:46increasing at an exponential curve even faster than what we anticipated. At the same time,
03:52the last mile of delivering real value to the end customer is really hard. There's security,
03:58there's compliance. We talked about deploying forward engineers into a lot of these organizations.
04:04There's change management. I think a lot of people underestimate how difficult that is,
04:09particularly if you have to produce a solution or provide agentic work that has to be accurate and
04:15thorough. It's really hard to get that last mile right. But I think that's where a lot of the value
04:21is created.
04:22On that last point, I really want to ask you about NemoClaw and what came out of NVIDIA GTC,
04:27because it's not straightforward, but it's basically an open source AI agent platform where
04:32if you just go on what Jensen Wang says, it could do anything you want. You invest in really interesting
04:38companies at different stages. But if NVIDIA comes out with NemoClaw and says,
04:42this is what it can do, what's the point?
04:46Well, I think it's an awakening moment for the entire industry. And this is when I go back to
04:53my earlier point about the reinvention of work. I think there's a world in which every person
04:58is going to have a set of agents for personal and professional use cases. And that's not going
05:03to be one company that owns it all soup to nuts. It's not going to be the model companies. It's
05:07not
05:07going to be the hardware players. It's not going to be one single application company.
05:10There are going to be a number, an ecosystem of players that's going to provide this agentic
05:16platform in the same way that we didn't imagine that every single person is going to have a PC
05:20on their desk. When Gates said that 30 or 40 years ago, that was so innovative. Now,
05:26you know, that makes sense. I think this is where the future is going.
Comments