00:00Joining us now is Samir Dallaki, a partner on the Growth Investment Practice team at Bessemer Venture Partners.
00:05There's so much in your portfolio that's interesting and highly relevant to this.
00:09But could I just start by asking what your net takeaway from everything Kimmy K3 spin?
00:15All consuming, Kimmy K3, in the news. And I think it's a great thing. I really do.
00:21I think we have a moment in time where I'm a free trade, competition is good for us kind of
00:28a guy.
00:29I believe that the right tool for the job is the right answer for our enterprises, for customers.
00:36And they will use Frontier Lab models like Anthropic for the stuff that they need to use that for.
00:42And they will use open source models from China and other places in the world over time.
00:46We need our open source U.S. and Western-based open source models, open weight models to get better and
00:53bigger.
00:53And that's great. And that will happen if we have free competition with the Chinese models.
00:57But that's my general take. I'm very positive on the development.
01:01Let me ask you one sort of final specific thing on Kimmy K3.
01:05Bessemer is an investor in Anthropic.
01:08One of the questions posed was, OK, if a startup of Moonshot's size and scale and resources can put out
01:16a $2.8 trillion parameter model,
01:18price it such that it did, what's the moat of an Anthropic or an open AI?
01:23Why do we value them such as we do?
01:27Your industry colleague, Son Monometti from Greylock, was on the show on Friday.
01:30And he basically said, well, Anthropic's got a real business with multiple streams of revenue.
01:35It's completely different.
01:36But where would you put your response to that question?
01:39Because we're big investors in Anthropic and they're in the quiet period, I can't comment specifically on them.
01:44But I will say about the frontier models generally, what I would say about their moats, they're still at the
01:50frontier.
01:51They're still extending the lead.
01:53We're in a stage of recursive learning improvements to the models.
01:57The models are being released faster and faster as we get ever better capabilities.
02:03I believe the frontier will continue to stay at the frontier.
02:06And you'll have others that will catch up.
02:09Now, there's also a lot of talk on X right now around whether there should be, Secretary Besson mentioned,
02:17should we be going after IP theft and distillation and how do we feel about that?
02:22I'm a free markets guy, but I don't believe in IP theft and distillation.
02:27So I do hope that we will figure out ways to better protect the IP of the closed models on
02:33the frontier.
02:34But that would be my take on that.
02:36Samir, it's been way too long since you've been on Bloomberg Tech.
02:39So take us inside the portfolio company's perspective on that.
02:44You have concern about IP theft on behalf.
02:47How would that impact if there were to be regulation or there were to be an action protecting those companies
02:53that you're backing and investing in?
02:55Yeah.
02:55I think at the frontier, it's important that they put up defenses.
03:01I think it's important that there are protections for IP.
03:05I think that we'll see those model providers are going to continue to explode in growth.
03:12I keep trying to explain to my colleagues in the market just to get a sense of the size of
03:18this market, Ed, is astonishing.
03:20This is like having the opportunity to invest in Rockefeller's Standard Oil in 1870.
03:26The market for intelligence is the largest market.
03:30I used to say in technology, it's the largest market in human history.
03:34And so the frontier model, the penetration in enterprises is so tiny relative to what it will be that the
03:42frontier models will continue to excel.
03:44But the open source ecosystem is going to continue to flourish.
03:49It should.
03:49Companies like Fireworks, we'll talk about, are going to benefit from that.
03:53The big news of late.
03:54So Fireworks took ARR from 100 million to 1 billion, but they did it in less than a year and
03:59a half.
04:00And you're, with respect, somebody with experience in software, you're basically saying we've not seen that growth trajectory before.
04:08It's extraordinary.
04:09And I used to be the CEO of a SaaS company and took it public in 2017.
04:12To build our SaaS company to IPO scale took us eight years.
04:17These guys are adding 100 million of ARR per month.
04:21But that took us years and years and years.
04:24Do you know what's interesting about that?
04:25Sorry to interrupt you.
04:26Those guys at startups like Fireworks, they're good at software, right?
04:30They're good at AI.
04:31Are they good at go-to-market?
04:33Do you know what I mean?
04:34So they're able to do that, 100 million of ARR per month.
04:38Have they been able to keep up with themselves on the sales channels and having good operational leaders in that
04:43business?
04:44Yeah.
04:44Well, I'll tell you, it's one of the biggest reasons we invested behind Fireworks.
04:47And we're so excited to back them.
04:50The CEO, Lynn, is an incredible strategist, a deep technologist, came from Meta,
04:54was leading the team from Meta that built PyTorch, which is sort of the leading open source framework for machine
05:01learning.
05:02But she's recently brought in George Hu as president.
05:07And George and I got to work together at Twilio after they acquired my last company, SendGrid.
05:12And I will tell you without hesitation, George is one of the absolute best go-to-market leaders, software executives
05:18in the industry.
05:20He was the COO at Salesforce, then the COO at Twilio.
05:23And so the companies that are scaling at this pace and have the wisdom to bring in some folks that
05:29have seen scale and growth
05:31and can put in process and operational rigor, we think are going to be real winners in this market.
05:37And that's why we think Fireworks will be the winner in their market.
05:40To lots of people, AI is still quite abstract.
05:42It can mean lots of different things.
05:43So, you know, at Bessemer, you guys basically say you invest across the stack, the model layer, the infrastructure layer,
05:49the vertical application layer.
05:51You know, but those are all quite different things.
05:54Like Lease AI is a high-flying portfolio company of yours.
05:57Indeed.
05:58But they all have slightly different factors behind them.
06:00They do.
06:01And you have to look at each of the markets.
06:03And every business is different.
06:05So, but what's extraordinary is just the demand and the pull from the market for all of them.
06:12You're just seeing exponential growth, growth curves like we have never seen.
06:16That's what they have in common.
06:17That's what they have in common.
06:18After, I've been in software now for 31 years.
06:20I've never seen, you know, a Lease AI selling AI to property management companies.
06:25You know, they're at, they posted publicly, they crossed 200 million of ARR and are still doubling.
06:31Like the companies in software history that reached what I would call triple-digit growth at triple-digit scale.
06:40Those triple-triples, that's extraordinary.
06:43Any company that did that in the prior era went on to become household names.
06:49And we believe we're investing in those kinds of companies.
06:51Lease is doing it in property management.
06:53Lagora and Evenup are doing it in the legal field.
06:57Abridge is doing it in healthcare.
06:58It's just an extraordinary group of businesses.
Comments