00:00It feels like the pace of change is so rapid.
00:03Can you just contextualize your portfolio companies, the industry writ large, just how fast things are moving right now?
00:10Incredibly fast.
00:11And, you know, our portfolio companies, like you mentioned, have a deep expertise in cyber,
00:17really across every facet of the cyber industry, which is a complex, very technical market to operate in.
00:24There's a lot of deep domain expertise there.
00:26And they have been reacting over the last several years to what we have seen coming.
00:31Now, the latest model releases have expedited the threat landscape.
00:36So, like Mithos and all that kind of thing?
00:38Mithos and every other model that's going to come next.
00:40And there's going to be a lot of them.
00:41And we're talking about Mithos today mainly because that's the one that people have been talking about in the news
00:46in terms of big cyber risk.
00:48But we see a lot more coming.
00:50And our companies are preparing for that.
00:52You have to operate today at a speed that is different than we've ever seen before.
00:59Companies are going to be exposed to threats faster than they've ever seen before.
01:04And so you need a layered security approach that, you know, our portfolio companies,
01:10which in combination produce about $8 billion of revenue, are providing to the market.
01:15But it's a really exciting time, but it's also one where our companies are needing to move very quickly to
01:21put the defenses up there for the industry
01:24and for our enterprise customers that really trust us to protect them.
01:28It seems like moving quickly isn't even enough, though.
01:30If you have these models that can find zero-day vulnerabilities in minutes and see things that have been overlooked
01:35for 20 years of human history coming through it,
01:38I mean, how prepared are we in this world for that type of technology to be released?
01:42It's so true.
01:43And we're sitting in a place today that, you know, companies have not had to operate this way before.
01:50And so we have companies like Proofpoint, for example.
01:53They have a massive network of 14,000 customers that every day they see all of the malicious emails that,
02:01you know,
02:02that are inbounded into those enterprises and all of the behavior, the way that their employees interact with those malicious
02:08emails.
02:09Those sorts of network effects give Proofpoint and their customers the ability to see zero-day threats,
02:16which is what we call them, very, very quickly, and then respond very quickly.
02:20It's really not as quickly as, like, a mythos, though, right?
02:22The way that we see the threats is very quick.
02:27And so, of course, mythos hasn't hit the market yet.
02:29True.
02:29But now we should all expect that it is out there and that perhaps people are already using it.
02:35By the way, do you think any of it is just marketing ahead of IPOs?
02:38Do you think some of it is just whipping up, you know, excitement and obviously fear alongside of it?
02:42I've heard that, but I'm going to take the optimistic view that this is much more about protecting everybody who
02:47might be interacting with it in the future.
02:49But I also think it's been a really good heads up for the world, for enterprises, for consumers to see
02:56what else is coming, right?
02:57Because, again, it might be the anthropic model today, but it's going to be someone else's model later.
03:03You know, the big thing that comes out of all of this also is that as these agents get deployed,
03:08and today, obviously, we're in a world where it's very minimal agentic deployment, although we're starting to see it pick
03:14up quite a bit,
03:15the governance around that is going to become critical.
03:17So what are the agents doing?
03:19What information do they have?
03:20Where is the data coming from?
03:22What are they doing after they get the information operating in the world?
03:25And that's what companies in our portfolio like SailPoint and Ping, ProofPoint and Darktrace are monitoring that environment
03:33to make sure that there's nothing malicious going on and then, you know, acting very quickly.
03:38And you're right that this is all new.
03:41It's happening very fast.
03:43But that's where you need incredible technologists, you know, behind the products that are in the market
03:47to react quickly.
03:49By the way, you've been one of the first to announce a partnership with one of the big LLMs.
03:53In your case, it was Google.
03:55We learned yesterday that Anthropics doing a partnership with Goldman Sachs, H&F.
04:00Before that, there was this like Bain, OpenAI, JV going on.
04:03What is going on in this industry right now?
04:05It feels like every single day we're getting some sort of announcement of some sort of partnership.
04:09Yeah, I mean, I can speak to our partnership with Google, which is, you know, they approached
04:14us to engage with us in our portfolio in terms of deploying their full stack technologies.
04:19We have a great relationship with them.
04:20We moved really quickly to take the portfolio there.
04:23And then on the cyber side, we're going to work with them to identify early any threats
04:27through our portfolio.
04:29You know, I can't speak to the other JVs that might be going on in the industry.
04:34I'm sure everybody has their own reason.
04:35We are model agnostic.
04:38So we have great relationships with OpenAI, with Anthropic.
04:42We know those companies very well.
04:45We, you know, we have ongoing discussions all the time.
04:47And certainly our portfolio companies are big consumers of theirs also.
04:51And I think it's really just our industry, you know, trying to understand and figure out
04:57more where this is going, how to deploy the technology, working with experts to help us
05:03do that.
05:03And then also, again, on the cyber side, making sure that we can get as far ahead as we can
05:08of these next models as they get out into the market.
05:10So I was having a conversation yesterday that was talking about how inference costs are really,
05:15really high for, you know, Anthropic, Googles of the world.
05:17But right now they're not passing that on.
05:19That led by Google, they're trying to offer something more cheap.
05:22Are you preparing for a world in which these, you know, Google, what have you, start to
05:27make their products more expensive and start to eat into more margins?
05:31How are you thinking about where the pricing world heads for AI, which is a really expensive
05:36project that we're undertaking?
05:37Yeah, it's a great question.
05:38I think today people don't quite understand the cost of rolling out a lot of these solutions.
05:44And there's some, you know, pretty interesting research out there today that says currently
05:48for most higher functioning roles, it's much more expensive to bring tokens into your organization
05:56to do it agentically than it is through a human.
06:00That's not true for everything.
06:01And of course, it's not always going to be there.
06:03But I think it's unclear today how much this will all cost.
06:06I think the reality is, especially as you sit in the enterprise and you have to budget for
06:10these costs and you have to re-engineer process, all of this takes longer actually than
06:16people realize.
06:18And that's before really absorbing, you know, what might be the true cost of deploying inference
06:23on a daily basis.
06:24Now, what you're seeing in reaction to that is very specific model use cases.
06:28So you see in our companies, both in our portfolio and their customers, they're deploying specific
06:34use case models that are much more efficient.
06:36I think we're going to end up in a place also where efficiency and power consumption end up
06:42being the focus and there's going to be a lot of innovation around that, right?
06:46So you don't need to use for every function, a general purpose model.
06:49That's where we are today.
06:50And you're right.
06:51The true cost of what this ultimately will be on a marginal basis for an enterprise customer,
06:57it's kind of unknown right now.
Comments