Good morning, everyone. Thanks for joining our 46th annual Canaccord Genuity Growth Conference. Pleased to have with us today, once again, JFrog. We have Ed Grabscheid, CFO, and Jeff Schreiner, VP of IR. Thanks for joining.
Thank you for having us.
Thank you.
Let's kick it off with last week's quarter. We had a broad beat and raise, cloud up 53%. It's been a volatile market, but the reaction has been positive. Just what were the biggest takeaways in your view, and did anything surprise you to the upside?
Yeah, we were very pleased with the results of the quarter. As you mentioned, 29% growth on the top line, 53% growth in the cloud, and we see three contributing factors here. Not in particular order of magnitude, but security continues to be a significant growth driver for the company, and we had a great quarter in terms of security.
We gave some of the proof points around security wins in those $1 million cohort, as well as in the new customer lands. We continue to see strong usage with our cloud products. Some of those customers, and broadly across our portfolio, are going above minimum commitments. That's contributing, and then, of course, annual commitments.
We see those customers not only using above minimum commits, but we also see an expansion of those existing customers to a higher annual commitment. The combination of that leads to our E+ Platform, and continued strength in our full platform. Was there a surprise? No. We kind of expected this and I think the surprise really came from the Street and from investors on how strong JFrog is performing.
Good to see. You've described, and the team has described, a tsunami of binaries, and when a customer is adopting coding agents, can you just help the audience better understand how that flows into Artifactory, building more code faster? How does that then concretely turn into revenue for you?
Yeah. We love that the term tsunami of binaries is sticking and we saw that, and we strategically positioned JFrog to be able to capture that tsunami of binaries. It really comes down to a few factors here that are easy to describe. Organizations today are becoming software factories, every organization, and they're moving at the speed of machines. When you're moving at the speed of machines, this creates a surge in the amount of creation of code that, of course, turns into a binary.
Add on top of that large language models, MCP, and skills, which six, eight months ago wasn't even part of discussion. Those are also binary. You're creating an exponential amount of binaries, and we're seeing that increase through Artifactory, through software supply chain, and how critical JFrog is becoming to these large organizations, in particular the foundational labs and many of the AI native companies, and this is a tsunami that we're seeing.
You talked about how the idea of an artifact is changing. It's now models, it's agent skills, it's MCP servers that are first-class artifacts alongside binaries. Can you just double-click again on why that makes you so much more strategic to these customers that are figuring out how to develop with AI?
Yeah. Yeah. Thank you for the question, Kingsley. I think what you're describing is what we're seeing, is that JFrog was the system of record for your software artifacts, your software binaries. We are now positioned, and we would like to think of ourselves positioned, as now your system of record for your large language models, for your MCP connections, for your skills registries.
Those are all going to be critical in the world of AI, in which we've already seen some customers come to us and look at hosting their skills, their MCP connections in Artifactory, thus bringing more importance to Artifactory and having to treat those, as you stated, as almost first-class citizens in the same way that you would treat the way that the artifacts and the software artifacts have been treated previously. Because each of those skills are attached to an agent.
Is the agent, in fact, doing what it was programmed with those skills to do? That's being managed, and where you would go to find that out, Artifactory. Is the MCP connection that I have with JFrog or Atlassian or other companies, is that acting up or is it acting different than it's supposed to be in the way that it's registered in Artifactory? The models likewise. The models are becoming first-class binaries because they're, in fact, the largest form of a binary, typically in a container, and now they're becoming different in the way they impact our storage.
Yeah.
Because I no longer can say that the Git repository, I can no longer delete anything, and that was unique to Git, where the binary repository and software could be cleaned out of what I did not need.
Yeah.
In the new world of models, when I do update one and I'm now on update three, I can no longer delete update one and it has to be hosted. It's changed the dynamics of how companies in the world of AI are looking at JFrog binaries in Artifactory.
So in recent quarters, you've had success with customers consuming above committed spend, as well as then translating that into higher committed spend levels. I think, as you pointed out, you weren't as surprised by Q2, maybe investors were. Can you give us a better sense of insight into how durable some of that increase in consumption is and converting that?
Yeah. So the trends of what we see in terms of usage have materially changed, particularly in the cloud. So over the last, I'd say, three quarters, we saw usage trends going well above minimum commitments. It was a historical level in Q1. Q2 ended up being consistent in terms of usage over those minimum commits and broadly across our portfolio. What we see today is there's less friction between the developer and the budget holder, and partly due to the fact that two things are happening. Number one is that AI is impacting those organizations and organizations are learning how to utilize those tools.
Secondly, I think that budgets are shifting in order to support AI spend. So they're offsetting it with other areas and allowing this frictionless spend, so to speak, and giving them the autonomy to learn. Now, as you move forward, what does this mean? Our model is built to give that flexibility. So you have the minimum commitment, and you can certainly go above that minimum commitment. It comes with an overage rate, and we capture that as revenue.
So as long as the customer continues, and we don't see why there would be a reason for that not to continue, at least through the duration of this year, above those minimum commitments, it's going to be reflected in our revenue. We are actively, meaning our sales organization, is actively working with the customers to capture that usage into a higher annual commitment, but we're going to do it in a strategic way.
We're not going to force the customer to do that. At some point, there will have to be a budget that is brought to the office of the CFO to procurement. Once you have that budget aligned and you have that event of a renewal, we believe we'll capture a higher minimum commitment, which would be durability in revenue and predictability in that revenue going forward.
When you say through the end of this year, is that sort of a comment in CFO language about before we commit to a higher level of spend? Or maybe you could just reframe that in terms of experimentation that we are seeing, or durable increase and step change in consumption right now.
Well, I think it is a combination of both. Like every organization, JFrog is not immune to it. You start to see spend levels that at the beginning of the budget cycle in 2026, that has shifted more towards now using tokens and developer AI tools. You offset that in order to be able to deliver durable profitability, which JFrog has done.
You see many organizations doing that, either through reduction of headcount or shifts in the way that they manage expenses. This is why we believe that the durability through the rest of the year will continue and that the usage will continue because most organizations have made that change to be able to offset that spend level.
At some point, when you rebudget and you start to think about the plan for 2027, you have to capture that now. There will be less, what we believe, maybe forgiveness and overspend of those budgets in 2027. We believe that most customers will come back to JFrog and recommit at higher levels as they typically do. You would start to see a capture of that usage as a commitment.
You now have some Claude Code and Cursor integrations and access to over 1 million developers. Maybe you could just help the audience understand what those integrations do, what kind of opportunity it is for you. Then there was a quote recently, just thinking bigger picture about how the market is changing. Matthew Prince said that humans are going to be a rounding error in terms of traffic on the internet over the next decade. Thinking about code creation, are you moving towards more focusing on an agent than maybe a developer?
I'll take the second first. I would say that we're still very focused on working with developers and agents or machines. I mean, JFrog and Artifactory and binaries at their essence are in fact machines. But we've been working with developers and building binaries over time, and I think that we'll continue to see some form of human interaction, human development work alongside machines for some time. I don't think that that's changed the persona that we're selling to. It could be the change of persona of the user, right? Within the organization and who now is utilizing JFrog more heavily, the human or the machine.
In relation to the first part of the question, with the connections through Claude and Cursor, I think these are unique connections that allow us to integrate that kind of Switzerland of binaries, where we continually need to integrate with everyone to have that flexibility and ease of use for our customers. You now have native as it relates to, let's say, Claude, native integration with Artifactory. Through that native integration, I can now have Curation scan packages the agents may pull that are PyPI, Go, npm.
A lot of these type of normalized packages that those agents in Claude are typically utilizing in some of their builds. So it's allowing me to integrate more cohesively with Artifactory in my use of these coding agents and whatever that coding agent may be. You saw it's Cursor, it's Claude. We're certainly trying to work with others as well to be able to give the customer the choice of who they would like to use and then integrate the JFrog tool stack.
So one of the early use cases for JFrog is analyzing and securing third-party packages. Now we're sort of seeing this mimicked and mirrored with users going out and pulling down whatever model they'd like to work with, their third-party models. That's a new use case for JFrog. But in terms of security and Curation, how much is that going to be applied to this AI use case, or what are you seeing now?
Well, this is where Curation becomes critically important for organizations because it sits outside of the organization. It is outside of the firewall, and this allows really to curate what comes into the organization. If there are concerns around open-source packages or open-source models coming into the organization, or maybe it is around the versioning of those models, you can curate that and protect the organization and still move at the speed of machines and the speed, trust, and governance that is needed in order to develop new applications.
We see Curation as being a critical asset, and we are certainly seeing that in the numbers that we reported during the quarter around security with wins of 80% in that $1 million cohort, security contributing 80% to those customers. And then for even new customers that we brought into JFrog, 40% of the customers landed with security, and the primary asset is Curation.
Right. We have seen a significant step change in security, where it has gone from an attached product to a leader for you, 80% of the $1 million customers in the quarter adopting security, 40% of new business adopting security. Just how has that changed in terms of how you are targeting a buyer persona, or just how has it changed buyer conversations?
Mm-hmm. Yeah, there has been a shift over the, let us call it three years. Before, it was two separate budgets. You had the CIO budget, and then you had the CISO budget. Those two budgets are now coming together. We had an overlay security team. That has now been consolidated in one team that is going to market and speaking on behalf of a platform. It is not only Artifactory, the assets that are around Artifactory, which is security and, of course, DevSecOps, which is something we will talk about, but you are selling the p latform.
You can bridge the two teams together and go with one offering, and we see the budget also converge into one. Many of the decisions, and you talked about the numbers of these large $1 million wins that we had, 80% of those with security, those discussions are being led with security, no longer being led with Artifactory. It is being led with security. That is how you are seeing this shift. Security is becoming critically important to the software supply chain and protecting the software supply chain, and those discussions start with security.
The supply chain security market is somewhat fragmented and crowded. When you do win, why do you win? Is it the Platform approach?
Yeah. Why do I start?
Jeff, you can follow. There are two ways to look at it. We have two offerings in security. First is Curation, where there is very little, if at all, competitive landscape there. I think clear value proposition to protect the organization, and we believe that every company should have Curation to protect, especially in AI world. That is number one. Number two is the Advanced Security. Advanced Security is a natively integrated Platform into Artifactory, and that consolidates the best-of-breed point solutions. We see an opportunity there of a consolidation. Many organizations want to consolidate. The fact that it is natively integrated, and it does consolidate, we see that there is a big value proposition there, and this is what we are leading with.
Yeah. Ed talked about it, native integration-
Yeah.
...scalability. A lot of the competitive alternatives, the Johnny-come-latelies, they are point solutions. They are point solutions that if you read their product documentation, tell you to essentially hack Artifactory and give us all the information from Artifactory for us to work properly for you. Being the fact that we are the creators and give you native integration with Artifactory, we think that that's a strong selling point against these alternative tools.
Also, the scalability. A lot of these Johnny-come-latelies have tried to create solutions that would've worked, I think, well in a software world, as I described it. Where good enough, if I only programmed in four languages and I wasn't very sophisticated, a good enough solution could've got me across the finish line in software.
Yeah.
But in the AI world, a good enough solution ends up inevitably getting me hacked because I don't necessarily program in that language, but the agent says, "That's the best language to use for the task that you're asking me to do.
Yeah.
Brings in a language you're not even conceivably programming in, and thus, we have had people get hacked utilizing those alternative solutions. Those are two of the main drivers-
Yeah.
...when we're sitting down and talking with customers about why you want to utilize JFrog Security versus a point solution.
First, I want to point out that you've developed pretty deep relationships with the frontier labs, and one of those relationships is with OpenAI. They've been using you as part of this sandbox containment. We've all seen it. The agent broke out-
Yeah.
...with the Hugging Face, and there was this Artifactory zero-day, and I think the response was excellent, and you patched it quickly. But is there any more takeaways on what you learned from the incident, maybe how it made your relationship with either the frontier labs or other customers stronger?
Yeah. I'll start, Jeff-
Yeah.
...and you can jump in. We see, first of all, let's talk about the benefit and what we saw. We can now openly speak about OpenAI as our customer. For the last 18 months, we weren't able to talk about OpenAI as a customer, although many of you here in the room probably already assumed that. As you said, we have deep relationships now with the foundational labs, and that's very important, especially as these very sophisticated and intelligent developers are using JFrog as a critical asset for their software supply chain. That's number one. Number two is the environment, self-hosted. OpenAI used a self-hosted environment here. You talked about breaking out of the sandbox, and we're very responsible together with OpenAI to remediate, patch, and distribute the update.
When you're a self-hosted customer, even a most sophisticated company like OpenAI, it takes time, could be days, before you get that remediation updated on your system, unlike a cloud customer that immediately gets an update. We saw a big benefit there. We know, over the last three or four quarters, that cloud migrations have declined because as customers are considering what that cloud deployment looks like in an AI world, maybe there's an opportunity to reignite those discussions around migration because of the situation. I think there's a benefit there.
Kingsley, you brought up a great point that I think maybe hasn't been talked about enough. Who's allowed in the sandbox? What infrastructure tools are even allowed in the sandbox at Foundational Labs? We now know one of them, and that's JFrog, and that's because the critical nature that we bring and how they work and incorporate JFrog into their model architecture. I think that when we're working closely with them, the other thing I think that people are hopefully starting to understand, that I've been communicating and Ed's been working and communicating, is that we're having a very close relationship with these labs.
This particular customer in general came to us with a Moonshot opportunity that we still believe is inherent. An opportunity that could be very transformative, excuse me, for JFrog and how we're utilized in an AI world. I think that's one thing that maybe is underappreciated somewhat, is that we are very close with these labs, thus we are seeing where they want to go, the problems that they need solved as it relates to binaries, two, three leaps down the road of product reiterations. I think that bends to give that type of a customer interaction a great benefit to JFrog to be able to meet customer pain points for leading foundational labs that will tend to trickle down to other enterprises as well.
Right. So, it's a great point that if a customer like OpenAI is choosing you to operate in a sandbox, that that's a good win. If other customers want to skate where the puck is going, they could look at what some of the frontier labs are doing-
Yeah.
...and take note. I just want to think one step further about potential implications. Do you think that that will be universally understood by customers over the next few quarters? What's the cloud migration opportunity on the back of this as well?
Yeah. So we saw this over the last, let's call it two years. We're really big in technology, we're big in the financial services. Then we moved into automotive. Automotive became an industry where you had a car, but it's really software on wheels. There was a very large North American automotive company that adopted JFrog, and it became a blueprint for other automotive makers. We're creating a new category now, which is these AI natives, and I think you're going to start to see a blueprint of how these companies are using Artifactory, and the full platform, hopefully. Today, they don't have security.
I think there's an opportunity for them to adopt security, but that will become the blueprint for the next generation of companies, and they'll start to look to those foundational lab and AI native companies on how they will adopt JFrog and use JFrog going forward, especially in a model in AI world.
Yeah, I would just point curation. We're talking a lot about Curation. Curation was created how? It was created because a customer came to us with a pain point saying, "Well, I have two ways to handle my binaries. Either I allow everything in and I scan it with Xray, and then I feel comfortable and I'll use what is approved." Or in this particular customer's case, I allow nothing in. The developer petitions for the package and hope that it's approved by the time that the software is created. But hey, wouldn't it be great if I could have a centrally located security policy that I create that then allowed only certain packages into my organization?
Yeah..
So again, we've then taken that and that pain point that is created and look at what it has evolved and grown into as the opportunity driving security today. I think that's what you can take away then, Kingsley, in terms of what can be done with these foundational labs and products that can be derived and brought to market for other customers.
Yeah. So I just want to make sure it's not lost on the audience. So you grew high- 20s in the quarter, 33% free cash flow margins. That's elite territory. How are you thinking about that growth versus margin trade-off in such a big opportunity in front of you?
Yeah. We're really actually proud of the rule of balance. So between the growth and the profitability. We've never been a company that was driven by significant growth and maximizing growth and minimizing profitability. It's always been a strong balance, and going forward, I think we'll use that same philosophy. Although we'd be willing to give up a point in margin to accelerate growth. I think most investors would be very happy about that. And we certainly look at that at this stage. But the balance between the two is part of the DNA and part of who we are, and we'll continue to operate under that same philosophy.
I think that you've put on a master class in guidance over the past couple quarters and years. Maybe just help the audience understand your philosophy and how you go about guiding a business where increasingly customers are consuming above committed spend and navigating that.
Well, first of all, thank you for the kind words. I appreciate that. It's certainly been a learning experience sitting in the seat of the CFO and when you step in and changing the guidance philosophy, that strategically changed because of JFrog. I've been at the company over seven years now, and the company has changed quite a bit from developer-led purchases now to enterprise-led purchases, security, a new asset on top of Artifactory that is driving meaningful growth in the ASPs, large deals, and then the way that we've constructed the cloud growth. Now 53% of our revenue is coming from cloud.
When you take all of these factors into consideration, you have to be responsible on the guidance. There's variability in the outcome of the revenue that can materially change from one quarter to the next. We felt the most responsible way to do this was de-risk those large deals that are being driven by enterprise purchase decisions and the usage over the minimum commit. This gives the confidence as a floor, and this is what we did. We guided, which we think was exceptional, 34%- 42% midpoint in the cloud. We increased our net dollar retention rate of 120% on the floor. All of these represent what we believe could be some meaningful upside if the business continues to operate, because you de-risk that variability.
Helpful to see that de-risking. Close on time. Just want to check if the audience has a question.
Sure.
Yes. Your company has been doing extremely well in the land of the software. As one of you actually said, over time, agents are going to create a lot more traffic and use a lot more artifacts than human. First of all, how do you think about the pricing of new artifacts? Obviously you cannot price linearly because then everybody's money comes to you.
Yeah.
The secondary is my kind of hypothetical question. These days, you know that these AIs, they are very powerful. They can create their own version, their own software, their own database in a heartbeat. So if the cost of using your service becomes so significant, could a bunch of agents just get together and form their own JFrog? A JFrog version.
Yeah. I know we're running up on time, so I'm going to be very quick on my response, and I can catch you afterwards. From a pricing perspective, I want to remind you, we actually generate revenue off of data consumption and servers. The only thing that we do on a per seat or contributing developer is security. There may be some change in pricing, and we're not going to be the driver behind the pricing change. We'll keep an eye on the market.
Regarding your question on, can foundational labs and models replace JFrog, they're going to be signals in terms of maybe a change from a security perspective that's needed. But JFrog is the system of record, and they're the ones that are going to drive the governance. They're the ones that are going to do the remediation and push those out. The models are not intended to be a universal platform. It might be a signal going forward, maybe let's say a red team, but JFrog will still remain as a system of record.
I'll just add to that real quickly. You're not going to allow Claude to govern OpenAI or vice versa.
Yep.
Within governance, I'll jump the shark to where you were even going. When they don't need humans and they don't need Git anymore because that's just for us humans, and they're building just binaries, you still need JFrog because we govern what the agent's doing. That's what we do differently that the agents will never be allowed to do in an enterprise.
Ed and Jeff, thank you so much for the time. Thank you.
Thank you.
Thank you very much.
Thank you, guys. Thanks.