Check Point Software Technologies Ltd. (CHKP) Earnings Call Transcript & Summary

So obviously, traditional firewalls look at IP addresses, look at different traffic, different ports, et cetera. When you look at what the AI network firewall solves, and we've been working on this for a while. We also incorporated all this into our SASE product. We have AI-specific threat prevention engines. And all of that is utilized both from an outbound and an inbound. It looks at prompt injections. It looks at the traffic that travels through normal applications because that's how AI is going to be used. So you not only have to look at that traffic, but you have to look at the AI traffic within it. So you're looking at prompt injections. And this is where the protection and prevention approach that Check Point has always offered is manifesting itself in the AI world. When you look at it, we have a whole AI defense plane that spans from protection specifically for AI applications, and this is infused throughout [indiscernible] and our workspace apps also. And so the natural evolution is for our firewall also to participate in this. And so where that's different is now you're not offloading it to a separate application. You're not allowing it into your environment. If you're using a separate application, you've already allowed it into your environment. So you're allowing it one step further to be able to evade whatever security precautions you've taken. We believe that this is a superior approach, and we offered it in the form of an upgrade to all our customers so they can take existing hardware, upgrade their existing hardware with the software, have these capabilities. And your next question, I'll help you with it, Max. Yes, that could lead to earlier refresh for our customers as they realize they want to adopt that and deploy it in a more full manner. And so that might shorten the life of what they expect from their appliance by buying something that's new and robust to address their future needs. [indiscernible], you can eliminate a question.

Great. I'll move on to the next one...

I'll slip one in for the one that we just talked about...

No, here we go.

What do you think Anthropic and OpenAI are going to do in security? And maybe it's less guessing what their strategy is. What do you as an executive team think is the right positioning, the right Swimlane for a Frontier model to complement and potentially compete? How do you prepare potentially for...

So I'll talk in a very large focus other than focus on any one company. If you look at last year when there was the SaaS apocalypse I was telling people when they would talk to me that, no, it's complementary. You're not going to take a system of record and replace it with somebody who's vibe coding, right? And subsequently, that's what you've seen. You've seen Anthropic engage with Salesforce. You've seen other companies engage with software companies. On the security side, all of a sudden, we went from SaaS apocalypse to security apocalypse, the Mythos moment, as we say, greatest PR moment in AI history, right? So it's pretty phenomenal. When you look from an OpenAI standpoint, we're aligned much closer with OpenAI today. We also have a relationship with Anthropic also. But on the OpenAI side, I would say it's -- you can look at our latest announcement and how we're leveraging OpenAI and the -- God, I can never remember the name of the Glass...

Glasswing?

No, that's Anthropic -- trusted access, yes. So from that standpoint, I think complementary. You're not going to take 33 years of knowledge and approach to a market and recreate it just with AI.

No, I fully agree with that. The risk that I think is worth pulling the thread on is not can the frontier models replace security companies. It's -- could the incremental share of budget accrue to the frontier models because the incremental risk is tied to code and because Frontier models are really good potentially at securing code. such that the share of whatever budget incrementally a large financial organization is allocating towards security instead of going to Check Point and make...

So let me -- so [indiscernible], I think you're saying, let me put it in a house security moment. So what you're saying is leave the front door open but lock the cabinets.

No, I'm saying both, but the front door should theoretically already be locked because of 20-plus years of Check Point network security.

Yes. But the difference is now you're in an AI world. And so as we demonstrated with the delivery of the AI network -- AI network firewall, you have a different paradigm here. And what I would argue is when you talk about applications being exploited by AI and now protecting them with AI, the problem you have is the advancement in AI that we all know about you don't want to test the application. You want to make that as infrequent as possible. And the way you're going to do that is by having the best security at every level. You're also going to be protecting applications with prompt injection, run time, et cetera. You've got to protect from all those things because take, for instance, our game, it used to be called Gandalf until we got sued by the Lord of the Rings. And now it's called Gauntlet, 7 levels of hell, which every high-functioning individual with a low personality probably plays this game and no offense there, but it's amazing. I break my brain trying to play it. But we have 1.2 million users over the course of 4.5 years. The intelligence that this brings -- it's not what you know today, it's what you don't know tomorrow, and that's what you're protecting for. And so you don't want to give AI a chance to test what you've done with an application. You want to give every chance you can to prevent that from occurring. So I would say at every level, I don't -- I think security becomes more important from all stages because AI is a journey through an organization from the applications you have to be wary of your own agents doing things that are adverse to your company as we've seen in the frontier models themselves. So you not only have to protect from the outside, you have to protect from the inside because we're talking about something that's learning and does things that are adverse. If you don't govern what they're doing, you may end up with NVIDIA having to buy Hugging Face or is that what it's called? Yes, Hugging Face. There's dynamics that are occurring that you don't expect from an application you built. It's got a mind of its own. And we're talking about today. Imagine where it's going to be in 5 years. So I think from a security standpoint, it becomes even more important with the models getting closer to AGI.

On product revenue, that's been a bit softer than some of your firewall peers which you acknowledged have to do with the go-to-market changes. But how...

Go-to-market changes in something different, too. Look at my peers. They're coming off of '21 and '22 cohorts, and that's what you're seeing for their refresh. Remember, they didn't have the best refreshes last year. So they're coming off of '21 and '22. And in '23, they fell on their face. For us, we're a little earlier and you're seeing us refresh those. It usually goes hand-in-hand when you deliver new boxes. So you deliver new boxes that excites people. There's usually a little lag as they bench test them and then they start adopting. And that can last anywhere -- you see the bump for anywhere from, who knows, 18 months to a little over 2 years. And then it goes back to the normal. You see it as a nice normal cadence. Now I think that can all change with AI. I think you have a paradigm here that's going to lead to, one, a lot more need for protections, especially the ones that we have in our AI network firewall and the rest of our AI defense plane. But I also think the capacity needs are going to be increasing, too. whether it's in your data center that you're putting AI into or whether it be a neocloud, a hyperscaler who recognizes maybe they're not that great at security and maybe they need to use somebody else that's been doing it for a long time or your AI factories.

Do you think that unit share trends have changed because of this go-to-market disruption? Or do you think the strength is primarily from refresh and there's not a whole lot of change in unit share demand?

I mean, look, at any given time, you're going to see at the edges, people winning and people losing. Earlier this year, we saw some wins. Last year, we saw some wins. I don't know if we're seeing losses right now or not at the edges. But you're not seeing dramatic changes. What I think you're seeing or will see in the future is the advent of people recognizing the demand function related to AI and AI traffic, which is how we started this conversation.

So in a market where customers are consolidating vendors, does losing units and without saying that you are losing any share, but in a market where vendors -- or customers are consolidating vendors, does losing units make it harder to win the broader account over time?

I think when you look at security, I think there are paradigm changes, right? So if we go back to when we first started losing what I would call share back in the 2010 and beyond range, that was because our competitors convinced everybody that they didn't need the best security. They just needed detection and remediation. Because of our founder and our ethos, we always remain focused on protection and prevention. And lo and behold, look, we have an AI moment and everybody else has figured out that all those critical vulnerabilities and those CVEs and everything else are really not great for business. And you hear people complain about token cost. Well, when you have so much repair to do and off of so many different disparate code bases, I imagine you're paying a lot on those tokens. So I think the difference is in an AI world, you're going to recognize really quickly. And maybe when I say really quickly, we're talking over the next 1 to 5 years. And it will come at varying degrees for different corporations, but they're going to recognize what security is and what security isn't. And security isn't great marketing, great sales. Security is really definable in an AI world because you're either losing an intellectual property, you're either having your business disrupted or you're not. And I think those defining factors will decide who gets more uptake of their units and who doesn't. In the near term, I don't think it's really something that's going to distinguish the future. I think what's going to distinguish the future is who has the best security.

So the challenge with the industry is that it's always been a combination of both. It frustrates us to no end where the best product in security actually doesn't always win. You can have products that are on par, products that are not as good from a benchmarking standpoint. But when you pair it with the magic of marketing, you end up with market share gains. And so I guess my question for you is, do you think because things are getting more complicated, more sophisticated, the attacks are getting more sophisticated. Do you think it's easier or more straightforward for Check Point to lean on the technology excellence on prevention than it was before?

So let's start with this. Massive attacks that did big damage over the last 15 years. They took millions of dollars to accomplish, and they were very few and far between. So those who claim that detection and remediation is all you needed in a sense, probably were right during that paradigm. And even though people paid a price, they were willing to look past it, right? Because all they did was just buy the next thing they thought would solve that problem. The democratization of attacks has happened. AI, those millions and millions of dollars of attack now take hundreds of dollars. So now the guys on the defense have to be right every time. And you've heard this from everybody, it's cliche now. But the outside guys only have to be right once. Well, if you're doing detection and remediation, and that's prominently -- don't worry, anybody that was doing detection and remediation last year is doing detection and remediation today, and it hasn't changed. It's -- you don't evolve. It's an ethos, it's a discipline. And so there's going to be struggles for some time. And maybe marketing and all that will cover it up for a while. But eventually, customers are going to recognize what is real and what is not. And time will tell. Look, I'm not trying to be a prognosticator, but AI is the great equalizer because now the guy that was a script kiddie is now using a model that's capable of exploiting some of the most sophisticated infrastructure in the world. So when you look at it from that standpoint, I wouldn't want to be dealing with a bunch of fragmented products trying to cobble them together to deliver something that is going to provide protection and prevention. I'd want the guy that just spent the last 32, 33 years focused on doing that. So discipline is a big deal in life. And while we need to learn the discipline of sales and marketing, and we're working on that, I think it's easier to learn sales and marketing than it is to deliver a preeminent protection and prevention product.

Gabriela and I have done some work drawing comparison between the early cloud security cycle and the AI security cycle. And in cloud security, Check Point ultimately chose to be more selective rather than own the whole CNAPP stack. AI security may be a bigger architectural shift. So how do you decide where Check Point needs to own the technology versus where partnerships is the better answer?

So I think in cloud, you were dealing with where data was located and traffic and things like that. So for us, it was -- we never believed everything was going to the cloud. We always believed it would be hybrid. And I'm sure Gabriela can remember those discussions with me because everybody said it was going to the cloud. And I said, life is -- life is a balance, right? And while we always kept investing in the firewall, right? And so it's sad but true. All you have to do is wait and things get in vogue again. And right now, the firewall is really in vogue. When I think you look at how we approach the cloud, it had a lot to do with that. We saw a much more firmer approach in the data center. And we believe we didn't have to be everything to everybody. I think what's different with AI is this is something where you have to -- it's a whole journey, right? It's from applications, it's from code, it's from agents, and you have to address it every step of the way. Sometimes there's going to be partners that it makes more sense to snuggle up to and leverage. But for the majority of it, as you can see, what we're doing right now, we own that transition, that journey of our customers. And I think the last piece we added in the firewall kind of puts the stamp on it. And so you'll see more stuff coming from us. Look, Lakera is only in the beginning, right? When we bought Lakera, it was a design within a very large customer of ours, and that's proceeded to be in other customers and develop other products. So you're only touching just the very beginning of it. Look at this, 9, 8, 7, what's your last words, Max?

I would love to sneak one more in, but maybe we'll leave it at, though.

Please join me and Max in thanking Kip for his time. Kip, thank you for that.

Thank you. Thank you, guys.