Confluent, Inc. (CFLT) Earnings Call Transcript & Summary

All right. Good morning, everyone. It's really great to see everyone at TMT 23 at Morgan Stanley. I'm Sanjit Singh, I'm the infrastructure software analyst at Morgan Stanley. We're super excited to have the management team from Confluent CEO, Jay Kreps; and Chief Financial Officer, Steffan Tomlinson. Before we get started, let me go through the disclosures real quick. For important disclosures, PC, the Morgan Stanley research disclosure website at www.morganstanley.com/research disclosures. If you have any questions, please reach out to your Morgan Stanley sales representative. And with that, let's talk some Confluent. So as the company, it started off or finished a 2022 that was successful, 50% plus growth. Cloud business grew 124%. I think the Cloud mix is approaching about 40% of the business. So a lot of real progress there.

Let's just take a step back, Jay, and sort of just lay out maybe just a very basic question, why can Confluent potentially become a big software company, a multibillion-dollar software company from your perspective?

Yes. Well, we're lucky to benefit from kind of 2 big waves, right? One, that everybody is aware of is the move to Cloud. And this is something that I think has continued to pace, right? Despite all the economic pressure on IT, et cetera, companies have found a better way to do stuff in the Cloud, and data systems are moving to the Cloud, being offered as a service. That's the first wave, I think most people are aware of. The second wave is data streaming. There's a big paradigm shift from thinking of data as something that is just stored here and there, sprinkled throughout the organization kind of in silos, to something that has to be able to come together in real time that drives the operation of the company. And that's a problem that's existed in some form for a while, but we haven't had a really good solution. And the rise of data streaming has really created a new paradigm for how to operate around this and one that's gotten significant traction. There are hundreds of thousands of companies using the open source Kafka that are offering is based around. That's an open source project that myself and my co-founders helped to create, and that's part of this larger move of data into continuous real-time processing and that's actually a really sizable shift. And one of the few examples where you're really witnessing kind of the emergence of a major new data platform in companies. And that's the other reason is, if you look at that, there's only so many of those. And empirically, you can kind of see it happening both in terms of the breadth of customer adoption as well as the depth where the companies that are doing this for real have hundreds or thousands of applications and are spending a significant chunk of their spend on data with us and so that's very exciting to see.

A popular question that I get from investors is, okay, I understand data streaming, there's value there. But what percentage of applications needs to be real-time? What percentage of your data pipeline needs to be batched versus real time? What's sort of your perspective on that, and how that may evolve? Like what does it look like today and how does that evolve?

Yes. Yes, it's a great question, right? It is the key question for this is, "Hey, is this streaming stuff some niche that you just kind of sprinkle on the edges? Or is it a big deal?" And I think a good way to address it is actually kind of flip it on its head. There's very few areas where you're like, "Hey, I wish this data was like slower and more out of date." It's actually not a good thing. And so it's not something where it's like sometimes I want A and sometimes I want B. You actually always want it to be fast in real time. That's always good. The question is; a, do you have a way to get that? And b, is there something kind of driving it along? And I would say, in both cases, there was kind of a gap there, if you were to go back 10 years. So there wasn't good infrastructure for this kind of real-time processing. And then the use cases were more a few and far between. So you would see this stuff maybe in financial services, some trading system or whatever that was very real time. But increasingly, what's changed is software and data has moved into the kind of drivetrain of the business, like how you interact with customers, how you produce and distribute your products, goods, services, all of that's now tied together by software systems. And that operation of business is real time. It is something that happens in the real world all the time. It's not like a batch thing that happens at the end of the day, reality is real time. And so modeling that in software kind of requires these capabilities. And so yes, what does that amount to? I think, in the near term, in the companies that have kind of done this at scale, maybe this is about 1/3 of their data footprint, which is quite substantial. The bet with Confluent is that many or all companies will become like that, right? And I think that's something where you're actually seeing good progress out there. Many technologists believe that there's kind of a clear path to that future. So I would say probably in the near term, a third, and then maybe try even more convergence between kind of action in real time.

Yes. That's very interesting. In Q4, you did see some impact from the macro, some longer times -- longer sales cycles and getting customers to come to an inclusion on purchasing decisions. Is there anything about Q4 and the weaker bookings you saw in Q4 that has, in any sort of way, shaken your confidence about the size of the opportunity or change your view on the market potential?

No, I don't think so, right? I mean we're part of this overall move to the Cloud to the extent that people are leaning into that further and further, it's happening faster to the extent that there's more scrutiny on spend, maybe it's happening a little bit slower. But I think there's a certain inevitability to this, right? It's fundamentally just a better deal to be building your applications, maintaining your infrastructure in a Cloud environment. And so even though that will control the pacing, I don't think it changes the outcome at all. And I doubt very much that, that's kind of a permanent state. If anything, I would see that as more a temporary slowdown.

That totally makes sense. And you sort of mentioned in your earlier comments just about the ubiquity of Kafka, you talked about 100,000-plus organizations using Kafka, 75% of the Fortune 500. If we square that against the customer base -- the paying customer base of Confluent, which is right around 4,500, how do you assess the ability of the company to more meaningfully convert the Kafka installed base to Confluent paid customers through this slower, more difficult-budget environment?

Yes. I think that's one of the exciting things in front of us. So in order to really convert the open source users in bulk, you really have to have a Cloud service, it's easy to use. So we started with an on-premise offering. That was an important part of capturing kind of the larger enterprise part of the market. Then we added a Cloud offering. In the last few years, that's become a really substantial part of our business. The interesting thing about Cloud when it relates to open source is it's actually a better deal, like it's actually cheaper than doing it yourself with the free software. And you would think, well, how can that be the case? And the reason is because when you think about your spend on one of these big distributed data systems, you're kind of hiring a team of engineers to run it, you're spending on Cloud infrastructure. The utilization of those things is not great, right? You're doing it just for yourself internally. It's probably not your core competency. And so you're kind of spending a lot of money, and you're probably not doing it all that well. And so we do this analysis with our customers. It's a very apples-to-apples comparison to say, "Hey, assume all I want is just Kafka, right? What would it cost me to kind of do it with open source? What would it cost me to do it with Confluent?" Assume there's no difference between the 2. Even in that assumption, Confluent is just a much better deal, substantially better. And then you layer on top of that, that the product is actually much more complete, much more substantial, data streaming platform with a bunch of capabilities around the connectors, the governance, stream processing. And you're actually getting something that is -- look better and cheaper. And that's compelling. And I think that's something that's really just started to come true for us as our Cloud offering came to maturity, and that's why I think we've shown the progress in the Cloud. And I think in many ways, that becomes more compelling when budgets are a little tighter. We're seeing that in our tech customers, and you probably hear it from some of the companies here. They're actually putting a lot of thought into how they deploy money internally. A lot of these companies had kind of big internal do-it-yourself efforts. A lot of those companies are looking at like, okay, is there a cheaper way of doing that? Do we need to have a small army of internal infrastructure, engineers across all these systems? Or can we just get a streaming service from Confluent and some stuff from the Cloud providers and some databases for Mongo, and would that be a better setup for us? And I think that calculus has changed quite substantially for a lot of these companies, and now they're looking for how do they make the shift. And of course, they have big teams and big spend in this area. So it's not an overnight thing for them. But I think that it's happening really across a lot of different industries where they're really looking at the build-versus-buy decision differently.

And looking at that total cost of ownership argument, so it's kind of an opportunity for the sales team as we look forward. Let's bring Steffan into the conversation. And so maybe to start with you. Confluent made the decision to accelerate its path to profitability by a full year. You made some difficult headcount decisions. Can you give us a sense of where the team is investing less where you're sort of doubling down? And as CFO, how are you coming to the decision as to what initiatives will get funded versus not today, versus, let's say, 12 months ago?

Well, let's start with the growth side of the equation. We're leaning into a $60 billion TAM, and we're able to continue to grow at a high growth rate, 30% per year while pulling in profitability by a full year. And while the decisions were difficult to make, the reality is we had doubled headcount over the last couple of years, there are places where we could have become more efficient, and we took this opportunity in this environment to do so. We are very much preserving our quota-carrying headcount capacity, and we're really leaning into ensuring that our unit economics around LTV to CAC, sales efficiency are all improving, and we're going to continue to invest in that area. On the R&D side of the house, it starts with Confluent Cloud. We're a cloud-first company. We're going to continue to make strategic investments in furthering many elements of the Cloud offering. They range from security to data governance. And we also -- we acquired a company called Immerock, which gets us into the Flink open source base. We'll be coming out with a commercially available product at the end of this year in the stream processing space, which is what Flink does. And we're doing that with the same design principles as we have with Confluent Cloud, which is it's going to be cloud-native, it's going to be complete and everywhere and that will be an area of focus for us. And we think that the Flink offering that we will be coming out with over time can be as big as the Kafka opportunity. So we're very much focused on building out a complete platform for not only data streaming, but stream processing. And in this environment, as far as us greenlighting incremental projects, it is very much a continuation of what we've done in the past, which is looking at the ROI, looking at payback, and we're going to be disciplined on that investment. And we believe that we can grow at very healthy growth rates over the long term with that philosophy of efficient growth.

Yes. That makes a lot of sense. And another popular question, and I guess the context here, Q4 2021, the operating margin was around negative 40%. By Q4 of next year, you're talking about getting breakeven. So that's a 40-point swing in operating margins. And so the question that I often get is, what are the levers? Like how are they getting there without sort of compromising what Jay sort of laid out as a pretty interesting opportunity. So you touched on it a bit, but I was wondering if you could just sort of go through some of the levers on the margin trajectory, not only in terms of breakeven, but ultimately to your target model?

Yes. When we look at the progression, it speaks to our growth and profitability framework that we laid out at the time of our IPO. Starting with our gross margins. Our gross margin profile of the business has been very resilient at roughly around 70%, even with a very large mix shift where Cloud has become a much bigger percentage of revenue. The Cloud unit economics that we have seen improve over the last 2 years has been very dramatic, and it's been a real bright spot for the company, and we're going to continue to make investments and operational decisions to improve that as Cloud becomes a bigger part. So we're going to see improvement in gross margins towards our longer-term target model of 72% to 75%. When you look at the other elements of the P&L, sales and marketing is the area that we have the most wood to chop relative to getting from where we are today to our targets. We want to get sales and marketing down to, call it, the 30% range. And we're doing that by really investing in the product side of the house, making sure that we have the right products for our sales teams to sell. We're looking at the product-led growth where customers can be onboarded without having any salesperson speak to them. And then we're really leaning into our customer growth go-to-market journey, where we matriculate those customers that start with the frictionless pay-as-you-go through the committed contracts, through the $1 million phase gate and then multiple million-dollar phase gate as we look at the $5 million to $10 million plus ARR customer cohorts, and those will be increasing over time. G&A as a percentage of revenue, will naturally come down over time. And then R&D, we're looking at having, I would say, modest improvement. R&D is the lifeblood of the company. And what Jay and Jun and team have done relative to building out the best-in-class R&D organization in the data streaming and data platform space is something that we're going to continue to lean into.

Makes a lot of sense. You mentioned, Jay sort of previewed my next question, which is, Steffan, my question next one for Jay, which is around the Flink acquisition or the Immerock acquisition. Jay, can you sort of frame us what does stream processing mean for your market opportunity? And then also, what are your sort of essentially new use cases that are going to be possible now that wasn't necessarily available for Immerock?

Yes. So I know this is like a new area, so it can be a little hard to make sense of all this data streaming, stream processing stuff. And the analogy that it's not perfect, but it's actually pretty good is if you think about kind of data at rest, how did that evolve? So early on, the focus was on storage, and you had storage systems, which would store your data. It's obviously useful. You got to keep it somewhere. Over time, you start to see the emergence of databases, maybe relational databases that can combine storage with processing. And that combination is really powerful, right? You can actually do a lot more with your data, if you can do those 2 things together, you typically end up with a lot more data if you can do those things together. And that became the basis for application development in kind of the data-at-rest paradigm. For streaming data, we see the sense that the first thing you need is you need to have data streams. And so the core of Kafka is about how do I read, write and store data streams? And how can I create like a hub for those that spans all the parts of my organization? You can think of that as kind of like a central nervous system that connects everything. Obviously, once you have those data streams, you're building these applications that use them. You're doing a lot of that building kind of from scratch. And so we've had some features that support building that kind of streaming application, the processing of streaming data, but there really has emerged these rich frameworks like Flink that make that much easier. And so you can think of that as kind of the query processing layer of a database and bringing that together with the stream itself. So we have a unique ability to do that because we're providing the stream. We can kind of bring in that processing. And we think that, that's a big part of the solution and that 2 things make each other more valuable, right? You're going to have more streams because they're easier to process. You're going to create new streams because of that, and because we have that stream, your processing is easier to take advantage of. So yes, we've seen this traction and the emergence of Flink. This is becoming kind of a popular de facto layer in the stream processing space that people want to build against. When we think about successful Cloud offerings, that's really the formula that works. It is something that is kind of an open de facto standard to build against with a really differentiated cloud-native offering of it. And so to do that, we knew that we would need to really pick the layer that people wanted to build against and then bring together the team that had the most expertise and knowledge, both about that community of people as well as about the internals. This is a very technologically sophisticated area, any query processing layer is, but a kind of horizontally scalable real-time query-processing layer is even more complex. And so to offer that as a Cloud service, you have to really put in the investment and build something that was kind of built for the Cloud, not just put it on some servers. And this was an opportunity to bring in those people and put them to work. It's proven immensely popular with our customers. We've been having this conversation since we have the announcement. Customers are extremely excited about it. Flink had a reputation as really delightful to build applications against but really horrific to operate. And so maybe even more so than Kafka. This is an area where you would have to hire this team of experts to run it. And so people really just want it done for them. And the value proposition of that managed service is, if anything, even stronger. So we're very excited about what it's going to allow us to do, and then what it's going to allow our customers to do, what the applications they can build.

Makes total sense. We've been talking a lot about kind of the mentality of customers in this environment. And so we've been getting the leads around stream processing, but we're also sort of talking about just big, higher-level questions. And one that I always continue to get back from the IPO and even to today is, why is this capability -- data streaming capability appropriately delivered from an independent provider like Confluent versus something I should just get from my primary Cloud provider?

Yes. I think to some extent, that's a question, companies ask about any vendor, right, that you usually you have some big vendors who will sell you lots of stuff, and then you have some specialists who go deep in an area. And so for an area of specialization to make sense, it has to be something deep and big where you're like, "Hey, this is going to be important for me. It's worth doing one more contract, right?" And that vendor has to bring something to the table. They have to do a better job that whatever your de facto, and the de facto, maybe it used to be IBM, maybe now it's AWS. But the -- it's the same thing of kind of there's some go-to-market contractual simplicity if you just bundle it all together, but there could be some advantage if you go deep. So that's the general trade-off. So in that general trade-off, you would ask, "Hey, is this data streaming thing a big deal?" Well, a lot of companies have adopted, this is one of their major data platforms. It's a really big deal and it's new. So getting it right and kind of plotting that path forward is something we have a unique advantage in. But -- and so that's how it would fall into that general trade-off, but there's something very specific to this area that's different that isn't true of other technologies. And that's, that ultimately, data streaming is something that connects together applications. Data flows with these streams across environments between different data stores, between SaaS layers and clouds. And so it doesn't just exist in 1 environment. It needs to exist across all your environments. It needs to actually knit them together. So one of the reasons that we didn't build just for the Cloud or just for 1 Cloud is really because of that. We knew that to properly serve customers at scale, you would have to have an offering that could exist on-premise. You would have to have an offering that could exist in each of the major Clouds, and then we need to be able to bridge between them. And that's something that's particularly hard for the Cloud providers to do for a variety of reasons. It's hard for them to offer services in their competitors' environments. There's been attempts at that, but they haven't been successful. It's been hard for them to bridge into the on-premise environment. There's been attempts at that, but kind of make success. And that's something that we were kind of built uniquely to do and is extremely valuable to our customers. When they think about the central nervous system, it has to kind of go across all the parts of the body, right? It can't just be in one area.

It's a point of step that makes on the earnings call quite often that your highest expanding customers are the hybrid ones, the ones you're using Confluent Cloud commiserate or in conjunction with Confluent Platform. I'm staying on the topic of like partnerships with the hyperscalers. Maybe give us an update on the AWS strategic partnership that you signed in 2021, and your ongoing relationships with Azure and GCP. In this tougher environment that we're operating in, do you see the hyperscalers potentially getting more competitive? In that continued cooperation competition, do you see any potential that they get more aggressive in this category?

We've seen the opposite. Actually when we were small and the Cloud overall was less mature. I think that there was an assumption in some of the Cloud providers that these third-parties would not really be a successful part of the ecosystem. And I think that, that's changed. So as companies have gotten to scale, and they're doing something very valuable for customers, ultimately, these cloud providers, their competition is not Confluent. I mean, there's like 1 team or 5 teams that need to compete with Confluent, but the bulk of that organization, they compete with each other. And one of the dimensions by which they compete is what's the ecosystem that I've got? And when customers are selecting, they're really thinking about that ecosystem I'm going to get. And if Confluent isn't good in that ecosystem, that's the problem. And beyond that, when they think even just purely economically, how am I going to make my next dollar? It really is how can I get a workload out of the on-premise environment into Cloud? And how can I kind of spin the meter on all the differentiated services that I have in the Cloud? And if you think about what makes that hard, what's the root thing that makes it hard to move workloads and take advantage of the different systems they provide, it is ultimately about the kind of liquidity and flow of data. Data tends to be locked up on-premise. A lot of the core systems running these big businesses are on-premise. The question for them is how can I chip out little chunks of that and move it into the Cloud without having to rewrite it all from scratch. Once it's in the Cloud, how can that flow and take advantage of these new systems? And so, in that respect, I think they're particularly excited about Confluent because we enable that. We enable the flow of data from on-premise environments. We deliver a huge amount of data, petabytes of data into S3 and other services that they have. That's obviously a huge thing for them just to get up and going, enable these workloads and take advantages of the services they have. And then, of course, we fiercely compete with a handful of small services that they have internally. But the reality is we've just gone much deeper in this space with a much more complete vision. I think because we're going after something new, and we have a vision for what that should be this year, but 3 years from now, 5 years from now, that I think it's hard to match unless you have something like that, that you're working towards.

Makes total sense. I'm going to go out to the audience and see if there's any questions. But before we get there, Steffan, would you like to speak in one just on stock-based comp and net share dilution? Your approach to managing that, and is there a share dilution percentage that we should think about on an annual basis in 2023 and going forward?

Well, first off, on the net dilution basis, in 2022, it was about 4.7%. And in 2023, we are saying that it's going to be between 3% and 4%, and we're going to continue to drive that down further over time. Stock-based compensation expense is a little bit of a different animal. It's reflective of grants that were done in the past. It has a 4-year kind of time line to amortize. So you're not going to see stock-based compensation expense as a percentage of revenue come down meaningfully until a couple of years from now. But once it does, it will get down into the mid- to high teens.

Great. I think we had a question right upfront.

Question about the edge. So you guys don't do really well in the core clouds, but there's sort of Cloudflare, Akamai, Fastly, [indiscernible] IO. And these guys stand usually between your data source and wherever you're processing it and could potentially step in and either take share or augment or otherwise act on that data flow. So how do you think about them? And how do you see them fitting in going forward?

Yes. The kind of edge environments and edge topologies are really important for, I would say, a subset of our customers, the ones especially that have some kind of real-world presence, IoT use case. And so the -- I don't see those organizations as competitive in any way. But yes, we do support running Confluent Platform in these kind of low-profile environment. And that's one of the reasons I talked about on-premise, but it works equally well at bridging into those environments. And that is the architecture we're seeing companies lean towards if they have that type of problem. It is kind of -- the hub of data is in the Cloud where they have all the rich capabilities, but they need effectively an outpost everywhere that they are that kind of collects and maybe can do some life preprocessing. And so we're serving a number of customers in manufacturing, automotive, some other use cases like that, that have that kind of topology.

With that, we're all out of time. Thank you, Steffan and Jay for giving us a really solid perspective on the opportunity. Thank you very much.

Thank you.