All right, I can go ahead and get started. My name is Jason Celino. I am one of the software analysts here at KeyBanc. I have the great pleasure of welcoming back Ed, the CFO of JFrog, to our wonderful Park City conference. Maybe, I do not know if this is a softball or a warm-up question, or maybe it is a hard question, who knows. With the rise of AI and agentic-powered software development, it has been such a tailwind for JFrog's business. Shlomi has talked about a tsunami of binaries. Maybe in your words, can you explain why AI has led to this creation of more artifacts, more binaries, and how JFrog has benefited from it? Well, first of all, thank you for having us back. It is great to be here. I forgot 8,300 ft, how hard it is to run from the conference room here to the gallery, so I apologize. I seem a little breathy. I am telling you, I am actually in good shape. I am not going to pass out here. Yeah. We used to do this in Vail, and that was at a higher elevation. Wow. I feel like I am at more loss of words this time. Really, to frame this simply, what you are seeing today in the world of AI with new AI tools and AI coding, every organization is becoming this software factory and moving at the speed of machines. What you see is code is becoming cheap, and the primary asset is the binary. JFrog is the system of record for the universal control plane, the layer of trusted security, and is becoming incrementally more and more critical as you start to manage the binaries through the software supply chain. Humans would compile the binaries and then move the binaries through the software supply chain. Today, machines are doing that, and you are seeing an exponential number of binaries now going through the software supply chain. Then, in addition to that, you start to take a new category. You have large language models. You have MCP and Skills, something that did not even exist six months ago or eight months ago, and we are seeing a larger number of those type of binaries. Again, everything being the binary, the zeros and ones, in Artifactory, our product that manages the software supply chain for binaries, and this is what Shlomi talks about, the tsunami of binaries. The creation of the code, which is, again, very cheap to do through these AI tools, creates this tsunami of binaries and opportunity for JFrog. Okay, perfect. Last week was my favorite week of the year, second quarter earnings. Very busy night on Thursday. Another absolutely stellar quarter. Total revenue accelerated by three points to 29%. Cloud accelerated again to 53%. What would you say were the key underlying drivers that led to the strength? We are very happy with the results. I think probably one of our favorite days as well, this Q2. It ended up being a tremendous outcome for JFrog. As you mentioned, 29% top line, 53% in the cloud. There is three main drivers behind this. Number one is security as a cross-sell is becoming critically important. Two years ago, we talked about security as being the next vector of growth for JFrog, and as you add more assets around Artifactory, and you see customers expanding the ASPs and increasing their commitment to JFrog, security is a big piece of that. Second is the usage, what we just talked about, the tsunami of binaries. That drives two behaviors. One is a capture of commitment, a larger commitment to JFrog. Number two is those customers that continue to use and use above those minimum commitments. There are maybe some questions around that, which we can discuss, and what that usage over minimum commitment does. It is a bit of an insight into what customers are doing and the opportunity for JFrog, as long as we can capture that, but it does reflect in our revenues while customers use over their minimum commitments. That all translates to downstream motion with the full platform. We talk about adding more assets around Artifactory and taking the full platform. We had significant growth in our JFrog Enterprise+ platform. We grew 39% year-over-year for those customers. In addition to that, it represents 59% of our revenues now. It is a critical piece, and it demonstrates just the ability for JFrog not only to be a player around managing binaries, but to continue to add more critical assets around that product of Artifactory to further expand the opportunity and TAM for JFrog. Okay. I will try to keep this interactive, so I will ask a question, and then maybe I will open it up to the audience. You are a CFO, you are like a numbers guy. It perplexes me why customers would be willing to pay for overage if they are paying 30%-40% more. Maybe can you speak to maybe the dynamics of why? Yeah. Well, to be honest with you, as a CFO, it perplexes me a little bit as well, but we're also faced with the same issue, and that issue is the unknown. We don't know how much you're going to continue to consume in this new reality that we live in with AI. We know that we're going to set a budget aside, maybe a discretionary budget, to continue to use and to spend, and then we'll get to the point where we have better education and knowledge around that continuation of that usage, and we'll make a commitment. It makes a lot of sense when you think about it in that context. Customers today don't necessarily know what they're going to use. Maybe it's even going to be more over the long run. What they decide to do, it's almost like a hedge or an insurance policy. We'll go ahead and use over the minimum commitment. The beauty of that is we have the model to be able to support that. You make a minimum commitment with JFrog, you get unit economics because you take at a minimum commitment, which drives better pricing on a per gigabit. But we also give the flexibility to use over that minimum commitment until you're ready to make a full commitment with JFrog. Our sales organization and go-to-market team is working together with the customer in order to capture a commitment going forward. By the way, they're incentivized only on a commitment, not on usage over minimum commitment. But they also recognize pushing a customer into making a decision that they're not ready for could leave money on the table, number one, or put them into a situation that they're not comfortable doing. Again, I'm going to go back to the fact that the model allows for this flexibility, and as long as the customer continues to use over the minimum commitment, it will be reflected in our revenue, and we're actively working to convert that to a commitment. Okay, perfect. Let me pause here and see if there's any questions. Sure. Right here. Just on this topic of overages, is there any way to help us kind of quantify it? You obviously beat and raised the full year by more than the beat. I don't know, just any framework or sort of process that Let me paraphrase. Yeah. The question Sure. So we have the webcast. Maybe quantify or contextualize the overage performance in the quarter. Yeah. I think I want to point to the guide and what we do from a floor perspective. Last quarter, we guided our cloud at midpoint 34%, and we increased that guide to 42%. What does that do? That represents the commitment to JFrog. So there is some capture that is happening. It is not all being driven by usage over minimum commit. And there is also the aspect of security. Security, regardless of your deployment of Artifactory, majority of our customers take security in the cloud, and there is a lot of benefits to do that. So that is also captured. Now, when you look at the usage, we do not quantify necessarily the difference. And again, I am going to go back to what I stated in the previous question. Our team is actively working to capture that usage, but it is a benefit, and as long as that continues at the pace that we see today, it will be captured in revenue. So we continue to guide on a commitment level, and you should feel comfortable that even if the usage was to stop, and we do not see that happening today, you should feel comfortable with the guide that we provided from a commit perspective. Okay. We'll take one more question and then come back. Thank you. Boost, JFrog Boost. Yeah If you don't mind to share what's the strategy behind Boost. Boost is the token optimization and CI acceleration. Yeah. Boost is a product. The question was around Boost, something that we opened or that we created for the community and opened up to the community, and that's really around optimizing the usage of AI tools. Truthfully, this is evolving quickly. A quarter ago, it was around minimizing the number of lines that were being produced. Now it's about steering to the appropriate models. This is an evolving. Orchestration role, that? Yeah, orchestration. Boost is in this phase. We didn't release it to be a product for monetization because we know that this AI economy is evolving. We're working alongside the community to fine tune this model and help our customers that are using JFrog optimize their spend and make sure that they're utilizing AI tools appropriately. Okay. Perfect. You mentioned security. I mean, that's a good segue. Yeah. I think the one product that a lot of investors are excited about is Curation. Yeah. When we think about, call it the, not necessarily the guide or your pipeline, but when we think about how that's trending, maybe speak to some of the more anecdotal or numbers aspects of showing how much you've seen that increase. Well, nobody is surprised that there's a large number of open source attacks. Shai-Hulud is something that we've heard now over, I guess, maybe the last three or four quarters. The use of Curation and the growth in Curation, obviously, to protect organizations and what is being pulled into the organization becomes critically important in this environment. Security as a practice is not new, but AI and machines pulling packages has certainly elevated the importance of that, and Curation is step number one to protect that organization. Developers and machines want to move quickly. There is a hesitancy, particularly from the CISO, around what you bring into the organization, and Curation fills the need. We created that product several years ago, and we saw that that was going to be critical in the software, or I'm sorry, in the security landscape. But it really took this incident to elevate it from an awareness perspective. There is no competitive alternative. It was kind of managed within the organization. This is the first opportunity to operationalize and build it autonomously. Since the first Shai-Hulud event in September of 2025, we've seen a significant increase in our pipeline, and we've been able to execute on that pipeline and the results, where it was maybe a 50/50 split in security, has now started to weigh more heavily towards Curation as the value of Curation has become critically important for our customers. Okay. Interesting. Sure. Why don't we take one in the back? On the security front, I mean, a lot of the foundation labs have really pushed into that area, but your growth has accelerated at the same time. Do you think that's indicative of the market growth is accelerating, or do you think you're gaining shares despite their moves are pushing into that area? The question is around the foundational labs and I think what you're asking is, are they infringing on- Their growth just explodes- Yeah. Or are you gaining share? Yeah. We're gaining share. There's not a single customer in our portfolio that doesn't use security product. There's an opportunity as we look at the 3,000 or so customers that have JFrog Xray. JFrog Xray is kind of ground zero for security to be able to penetrate those customers. JFrog Curation, as Jason asked about JFrog Curation, there's no competitive alternative, so you don't have to displace, and it's easy to win right now in this environment with JFrog Curation. JFrog Advanced Security is a displacement of point solutions, and we do see as organizations want to consolidate the point solutions opportunity there as well. The sales cycles are much longer than JFrog Curation, but we do see really a continuation of winning in that space as well. It's further just, I think overall security's becoming critically important, and we're winning in that with JFrog being the system of record, this sits natively on top of JFrog Artifactory. Okay. Unfortunately, we're seeing all these supply chain attacks. Last week, Shlomi talked about another one. Yeah. Investors, they probably get too enamored by all these announcements, but one new attack won't drive incrementally new pipeline, I don't think. Maybe speak to your sales cycles or close rates or what are you doing internally with your sales to capture this big pipe that is coming down? Yeah. Well, it used to be that you would have an incident once every year or every other year. Now it is becoming more regular and frequent, unfortunately, but fortunately for JFrog, and it certainly creates an awareness. The pipeline is certainly connected with incidents. So when you have a big incident like Shai-Hulud and two weeks ago, Shlomi had talked about the situation, another Shai-Hulud event, which could be the largest. It drives awareness. It drives pipeline. It is an opportunity for the sales team to remind our customers the importance of having Curation and Advanced Security to protect the software supply chain and the binary management. So there is no doubt there is a correlation there. These events typically would increase the buying decisions. It is amazing when you have an incident, there is what we call an incident budget. When something like this happens, enterprises make decisions sometimes without including procurement and legal. They just need to protect the software supply chain, and they make a decision. We have had that happen in Q4, we had it happen in Q1, and again in Q2. When there is an incident, customers react relatively quickly and make decisions. That is not always the case. But certainly around Curation, the sales cycles are compressed. As I mentioned, Advanced Security is a replacement of existing products, so that typically takes longer. But we had a very strong quarter of new logo, and as we talked about, one of the disclosures we gave was around security. 40% of the new customer wins during the quarter landed with security. So that just tells you how critical security now is becoming as part of the platform. That asset is critically important to JFrog. Okay. It is actually a good segue. So, there was the OpenAI Hugging Face incident. Yeah. I think JFrog very transparent, published on your role within this. Yeah. Incident. My team, to their creative credit, looked at Reddit and Twitter to see what the forums were saying, and the feedback was not terrible. Yeah. Obviously not the greatest publicity. What is the risk to you as the CFO with something like that? Yeah. Let me start with the benefit because we actually saw a lot of benefit in this, and so I think it's important to highlight the benefit. You mentioned the quick reaction and the transparency that was important for us. In fact, the benefit here is we talked about having foundational labs in our portfolio. Now we have four of the foundational top five in the portfolio, but we can now openly talk about OpenAI because they came out and talked about JFrog, so we can speak about that. In addition to that, I think it highlights the need for having cloud and a SaaS environment because although we were very transparent in this situation, we worked very closely with the customer. Within light speed, we were able to find remediation and patch. You still have to send that to a self-hosted environment. Cloud customers were immediately covered, but self-hosted customers, they have to download and update their operating system in order to capture the patches for that CVE. A company like OpenAI does that quickly. We have 3,000 customers plus that still are in self-hosted, and we can't confidently say that all of those customers have updated their self-hosted environment. That's the positive of it, is there's some motion here that we see that, one, we can speak of a customer. Two, we can talk about maybe a migration opportunity. The downside of it is that you're right, there is a CVE, but that's not a JFrog issue. That is going to be widespread. Is this going to happen again? Yeah, obviously. As long as the protectors and the responsible vendors are working closely with companies like JFrog in order to patch in this environment, then it should be a positive. We're hardening the product. We're moving more towards a modern technology stack and be able to meet the demands of our customers in an AI world. Okay. Again, great with the segues, by the way. The fourth AI customer. Yeah. Congratulations. That's a big win. But I think some of the details were kind of interesting. It was a competitive displacement. Yeah. It wasn't just a pure self-hosted deployment, it was cloud as well. Maybe speak to why these are significant and why it creates additional opportunity. There's a new category, and that category is AI foundation model labs and customers. When you start to look at JFrog and the transformation of JFrog technology in automotive and financial services, now we're going into a new category, which is these foundation model labs and AI companies. This creates a blueprint, and it builds a lot of trust within the community. If you have some of the most sophisticated organizations in the world relying and trusting on JFrog, there's a lot of confidence there. Jason, you had mentioned kind of a hybrid environment. Many of the foundation model labs started with self-hosted. They're building data centers. They have deep pockets. It made a lot of sense for them to use self-hosted for good reasons. They're controlling what's coming in and out of their environment. They're doing it in their own data centers. Now you have a customer that is putting the mothership in the cloud because they recognize the ability to have to update continuously, and then pushing out to the edges where their data centers are in a self-hosted environment. This could be the blueprint going forward. It certainly brings a lot of, as I mentioned, confidence when you have this new category of a very sophisticated company building in a hybrid motion. By the way, JFrog was built on being hybrid. There's no other company in our space that's hybrid. You either cloud native or you're self-hosted. We're hybrid, and we can meet the demands of that sophisticated organization. I think that this is going to be good for JFrog going forward. Okay. Very cool. I'll ask a question, and then we'll open it up. Sure. Token rationalization has been a bigger topic. I'm trying to ask all my CFOs this. Yeah. Because you hold the purse, right? Yes. Or the wallet or however you want to describe it. I think for your internal AI use cases, What are some ways that you're keeping that budget honest? Yeah. Six months has seen a drastic change. From the time that we planned to the time we are today, so much has happened in the world of AI. We were using Copilot and Cursor. Copilot is replaced by Claude, and the productivity gains that we see out of these tools has increased dramatically. So has the cost. It's becoming its own cost line now. So it's a budgetary line, and what happened from January when we did the budget till today, you have to pivot, and we're a very disciplined company. You see it in our operating margins, even in this environment where budgets, and we're not immune to it, are going well above what was originally planned for token spend. We were able to pivot and continue to see growth in operating margins, but you have to do that in a way that's responsible. Number one is we looked at our growth headcount, in particular in the R&D. We saw productivity, and we had to make some decisions around pausing on maybe some of those growth objectives while continuing to invest in innovation and making sure that we're capturing this momentum. Unlike some other companies that had to make drastic changes in their headcount and reduce 20% or 40%, we're able to manage that headcount with discretionary budgets and other things that we set aside. We see the ability to manage that, and it's around the discipline and the nature of JFrog, which is continuing to drive profitable growth and durability. Okay. Are there any technologies that you can use? We've heard of routing and using cheaper models? Yeah. Are there other ways to do it? Yeah, certainly. We're looking at all of these different techniques. We're looking at model routing, and making sure that we're optimizing on the models. We're looking at ways to limit certain organizations by doing caps on spend. I don't think that's uncommon. I'm actually part of a CFO forum. I was asked by Cursor to join a CFO forum just to learn what other top CFOs are doing around very large enterprise companies and how we can shape maybe the return in the AI economy and what that means going forward and shape that for organizations. More to come. It's still very early. Okay. I think we're about out of time. You can follow up after. I always like to end it with a fun question. Ed, we know you as the CFO. We know you with JFrog. One of the things I like to do is cook. Tell us something that we won't know about you unless we ask. What do you do as a hobby? Okay. I just got back from a two-week vacation. I love to travel, and I've got two older daughters. One just graduated college, and one is getting ready to start her third year, moving to Spain, Barcelona, to study abroad. We love to travel, and when I do have free time, I'm a mountain biker. I enjoy mountain biking. That's how I keep myself in shape and try and detach myself from AI and binaries and get on a two-wheel bike and get out there and ride in the mountains. Okay. Very cool. Well, thanks, Ed. Thanks, everyone, for joining us.