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

All right. Good morning, everyone. I'm Sanjit Singh. I do infrastructure software on the software research team at Morgan Stanley. We're going to be talking real-time streaming, super thrilled to have the Confluent management team, CEO, Jay Kreps; and Chief Financial Officer, Rohan Sivaram. Thank you both for joining us again at the TMT conference. Really appreciate you having here.

Thanks for having us.

Awesome. So let me get through the disclosures. For important disclosures, please see the Morgan Stanley Research Disclosure website at www.morganstanley.com/researchdisclosures. If any questions, please reach out to your Morgan Stanley sales representatives.

And so with that, sort of level set, Confluent had another great year last year, grew revenue 32%. Your Cloud business grew 65%. I think even almost as impressive operating margins improved by 2,300 basis points in a single year, super impressive. The spending environment has not been without its challenges. And so we've been -- if we look at the past several quarters, just broadly in software through a sort of tech downturn. Jay, from your perspective, going through the last several quarters, where do you think the company has shined. And what were some of the things that the downturn sort of surface that, okay, we need to get better at these couple of things?

Yes, yes, yes. I think you touched on one of the things I think we did well, which was sustaining high growth through a tighter environment and then driving efficiency. We came out -- as we came public, we were running relatively hot, and that was intentional. If I think about our growth trajectory, the company grew very quickly. We were able to, I think, really just outpace some of our earlier competitors to turn into kind of nonfactors. And I think that was a positive thing that we were really the only company deeply focused on streaming at large scale. I think that's been a tailwind. So I think our high investment was valuable. But coming into '23, we definitely wanted to really focus on efficiency. So what did we do well? I think first of all, we achieved that. Second of all, and maybe more importantly, we did it while sustaining our strategic investments in the product in the space. And so from the outside, this is always less apparent from just looking at the financials. But there's different ways to cut. You can chop off large initiatives. That could be good if those initiatives aren't going to pan out. But in our space, we felt this next wave of functionality for us going beyond our core Kafka service, which is just the kind of raw data stream to the connectors, they actually get that data and attach to the rest of the organization, the governance functionality to manage this at large scale and Flink the stream processing capabilities to really act and build applications around that. We felt like that was the next waves of functionality that really take us beyond that kind of core data stream to a complete data streaming platform. And so we wanted to make sure each of those stayed on a good trajectory. And that means heading into this year and beyond, we kind of have these waves where the connectors are on a very nice growth trajectory and can turn it into a large business and their own right. The governance functionality has grown very quickly in its early days. And in Q1 of this year, our Flink service will go GA, and so that was obviously a hard thing to balance where you're both investing in one area, while also really trimming and optimizing and driving efficiency in a lot of other areas of the business and trying to manage all that in ultimately a tighter environment. I think we did that well and that's something I'm definitely proud of. In terms of what did we get wrong, what could we do better? I do think one thing we were a little behind the ball on was what we ended up doing this year, which was this consumption transformation. We took the first steps on that change maybe to the tune of about 10%, 15% in '23. At the time, we felt like that was too big a change to make in coordination with everything else that was happening like you can't both trim 23 percentage points of operating margin and also make big go-to-market changes at the same time. But what this change means for those who don't follow us closely, is really aligning our internal go-to-market with the way customers work with us. So customers buy our product on a consumption basis. And they have the ability to lock in a commitment, but their bill month-to-month is determined by the actual usage and applications they've built. And this has been a trend in our portion of the industry overall. The companies are really aligning the go-to-market to that consumption, saying, "Hey, the sales team is going to be paid for the new applications, they bring in that drive consumption, not just the committed spend." And that actually turned out to be more important than we realized in '23, heading into an environment where there was more conservatism, having a model where the go-to-market team is focused on trying to lock-in big 3-year commits upfront before anything has been done. It might have a lot of risk in terms of exactly how you're going to use the product and exactly how much you're going to need to consume that turned out to be, I think, a little out of step with where the market was. And I think we felt that as we went through the year. And so we've corrected that. They say the best time to make the change would have been a year ago, but the second best time is now. And so yes, we did -- we've corrected that. We've kind of made a more complete transition heading into this year. That's gone well so far, but obviously, a lot focus for this year is around that change.

Yes. It makes total sense. So if we talk about the founders of this data and motion market, the streaming market, starting with Kafka way back in 2010 and now Confluent itself being a platform. Can you give us a sense, Jay, of where we stand sort of mark-to-market, if you will? Where we are on the journey in terms of these real-time data platforms becoming a fixture of the modern data infrastructure. And maybe in terms of just like the complexity of the use cases, the sophistication of use cases, where do we stand in 2024 now sort of 15 years or just sort of 15 years into the journey?

Yes. Yes, a lot has happened. As we were starting Confluent is really based on the hypothesis that the infrastructure around data would go from really just focusing on data at rest, like how it sits, how it's stored, how you kind of look at bits out of a store data store database. A platform that's really about how all the different data systems in an organization connect and how data flows in real time. So that was obviously at the time, it's a bit of a conjecture as to how the world is going to change. What's happened since then is a whole set of things that I think take a lot of risk out of that transition. It doesn't mean the transition is finished, but it just means some of the risk is gone. So what changed? In 2014, I think most technologists, if you asked them, would have said that there was kind of fundamental limitations in real-time streaming that meant that it would never really be able to do the kind of batch things and this was certainly the experience. If you looked at a lot of financial services firms, they had big sophisticated real-time platforms, but it was very limited in what it did, and it was very hard to build that stuff. And so the idea that, that would become an easy, powerful primitive that every company would use all that was not obvious, right? Secondly, I think the use case is advanced pretty significantly, like just the pressure to have customer experiences that bring together all the ways that you interact with a customer across the business that are up to date and rich and contextual, that became much more over the last 10 years and not pushed on this. So we went from something that was -- it was unclear if it was possible, and it was unclear if everybody wanted to do it to something where suddenly, we've checked off a lot of these kind of technical areas. And we know, yes, we can make streaming as efficient. We can make it as easy to use. We can make it as complete. Now it's kind of a question of just maturity of that technology stack. And then on the demand side, a similar thing where we've seen this now take root in virtually every industry, large companies to small companies, use cases that range from the side of the business to things that are right at the core of payment systems, transaction processing, the kind of highest risk, most mission-critical systems in the world. And so I think a lot has happened there. It doesn't mean we're done. If we think about this transition, when you think about infrastructure stacks, they don't move overnight. This happens kind of application by application as things turn over. But I think it actually presents a really interesting point in time where there's some technology change, there's a lot of evidence for how it's happening. The kind of end state or stopping point, you could debate. We've said a number of times that, like, hey, when we look at companies that are further along, it's more like 1/3 or more of their applications are kind of in the streaming domain. We think that's more the norm. Logically, that makes sense. If you think about how business works, it's a very real-time thing. As these software systems become more connected, they have to exchange data that way, they have to react that way. So that's where we think the stopping state is, but we're far from that. But it's kind of clear that it's going to be a lot bigger than it currently is. We just don't know how far it goes. And so I think it's a very interesting time for this whole area of streaming. You're seeing streaming show up in all the different parts of the data stack to integrate with this, that's a huge driver for us because now we can connect into all the other data platforms companies have, be able to actually power Confluent off of that. So yes, I think it's a really exciting time. And I think a lot of the kind of more foundational risk of how this will turn out has gone away, and that puts us in that kind of deployment phase of something large going across the economy, which is cool.

Yes, it makes a lot of sense. And the pick up on a point that you made about customers that get it, roughly about 1/3 of their applications that are embedding real-time capabilities. But overall, we're not there yet. And a lot of the questions that I get from investors is like, are there any sort of analogs or paradigm to think about the penetration or how this category will evolve? I sort of point to the mix between sort of no SQL relational databases 10 years ago, that was like 95% relational, today it's 75% relational. And in a category that big, that's multi-billions of dollars, when I know SQL goes from 5% to 25%. Is that like a decent way to think about this? Or are there any other sort of paradigms to think about the penetration of streaming into...

Yes. I think that's not a bad example. One of the things that I think is maybe a little easier to reason about with streaming is there's kind of a more fundamental logical reason why things have to move to streaming. So if I said, is it the case that logically you have to move from relational to non-relational data storage, will say, I don't know. Maybe it kind of depends on what developers like, maybe they like it, maybe they don't like it, maybe they like it, and then they stop liking it. Like it's hard to say. And it turned out they have liked it, so there's been more. With streaming, I think the argument is actually really straightforward, which is businesses run in real time. A lot of the customer experiences that are now powered by software, a very real time and cut across many parts of the business. They have to hook together many pieces of software to function. The software is really entering a lot more of the kind of the drivetrain of the business, how products and services are produced, manufactured delivered, logistics, orchestration. To do that, fundamentally, it's a real-time thing. Like reality happens continuously throughout the day. And yet a lot of the technology stack around data came out of this paradigm of batch processing, meaning the job kicks off at midnight and it computes through yesterday's data, and it spits out some results at 5 a.m. or whatever it is. That's just very hard to integrate with the real-time business. If you look at what tech companies have done, they've kind of moved out of that batch into a more real-time stack, you look at the mainstream economy that's happening. So I think the nice thing about this area is you can get there just kind of watching the line go up, just watching adoption empirically. But maybe unlike some of the move-out of relational data management, you can actually get there just logically, like logically, if you can have something which is continuous in real time and if that's not more expensive and not harder to use, then you're going to see a very significant portion of data processing moved to that over time. And I think that logical pressure is actually exactly what gave us confidence as we were starting the company. We were like, look, as we started Confluent, maybe we fail, but if we felt somebody is going to succeed. And if we have to watch them be successful, it's going to be really aggravated. So we better get out there and start this company. And I think, indeed, we were close enough to the right time that we were not too early or vastly too late that we've been able to really kind of help push this forward. But there's no question that this is something that's moving in the world now. So yes, I think the relational analogy is not bad. I think the rise of databases themselves, if you're willing to go back in technology history a little further is actually reasonable as well. This is a time where, hey, there used to be many hacky ways of storing data and building around data that ultimately coalesced to the relational database. As you've got a layer that made that type of application much easier, you got a lot more usage and value that was created on top of it. Similar thing in the streaming world where you've had -- depending on how you count, a half of dozen, a dozen little micro categories that solve some part of data movement, real-time data integration, application integration, ETL, Massachusetts, just all these crapping micro categories that are kind of being displaced by something much more general, much more powerful and much more easy to use. And so I think if you see that analogy, I think you can see it as kind of like what databases did for data at rest, streaming can do for data in motion. And that's another way to think about it.

Yes, that's a great way to great paradigm. I wanted to go back to the sales transition to consumption. And you guys have been clear, this is not a pricing change. This is around how you're orienting the go-to-market organization. I was wondering if you could take us behind like looking through the eyes of a salesperson who's been used to getting a quota, hopefully exceeding that quota, getting a big fat check at the end of the year. How does their job change? Or what is your strategy change to make the same amount of money or money or more money in this new sales model?

Yes. Yes. Well, a lot stays the same. You still have a quota, you still get a big fat check as you land new use cases. But if you think about maybe the best reps that we had, I actually don't think their behavior will change all that much. They were already out there driving adoption and getting new applications and working with customers on rightsizing the commit as that usage grew. But there is a place where maybe the commit oriented compensation was a little bit misaligned with ultimately what the customer wanted, and what Confluent wanted in terms of our revenue and growth. And that would be this pattern. So maybe the rep comes into account and they say, "Hey, there's a lot of interest in Kafka here. There's a lot of use cases they could be pursuing that they're talking about doing." And you kind of go around the account and you collect everything that could happen. And you put together this massive deal. So you say, okay, we're going to do Confluent for all these things and you kind of drop this big 3-year thing on the counter and say, "Hey, sign here." And the challenge with that is it can just turn into a very long process where you, the company now have to derisk this. You're like, well, are we really going to consume that much? We haven't even built the application? How much does it need? Go do a 6-month study to determine the exact usage in this case. If you have dozens of applications, in some cases, it's more than that, that are kind of being pursued that suddenly turns into a massive project where all the risk is on the customer side. If they commit to too much, that's a big problem for them. And yet the incentive of the go-to-market is to kind of maximize that upfront commit. So that's the bad case. That's where you have misalignment of what the company wants because it's not like our revenue changes at all. When you make that consumption, we actually need the -- sorry, when you make that commit, we actually need the consumption to drive the revenue growth. So the model now really incentivizes the behavior of our best reps, which is go in, if the customer wants to commit upfront to lock in pricing, we're happy to do that. If they want to wait and get some stuff out into production and then do it then, we're happy to do that. If they want to commit to a smaller amount, launch some things and then take it up later, we're happy to do that. The focus of our team should be all around finding the new use cases making sure that we're part of that stack, accelerating unblocking their deployment and taking it out there. And that's what we want. That's what drives Confluent's revenue. That's what now drives the sales compensation. And that's ultimately what's valuable for the customer. If you just kind of look at the history of software business models, I think things that create alignment between the customer and the providers are the things that have actually driven a lot of value when going from perpetual license to subscription licensing to SaaS. And I think consumption is like one more step on that where we're bearing some of the -- we have skin in the game with them in making sure that we're creating value, and that ultimately means they are willing to move faster and it becomes stickier because there's less chance of shelfware or unused commit. So I think all of that is positive. We're kind of going through a lot of changes to make sure we do that right. Naturally, this changes not just the sales compensation, but your notion of pipeline, you have to be much more detailed about what are the applications that are kind of going out to production. We're kind of working through all that. We're very optimistic about how this plays out in the business. Both -- because we had taken the initial steps on this already last year, but also because we watched a whole set of peers that leaned into this very aggressively, and saw really good results over time as I really mastered that new system, that became a significant accelerant for them.

Yes. It looks like -- still it's like a adopt a similar approach to it. So that definitely seems to be on the way sort of interest software sales is going. Jay, how long do you think these changes are going to take to bear fruit in terms of the behavior you want to see out of the sales force? And then Rohan to bring you into the conversation, what does this do to -- what does this transition do due to some of the financial KPIs. Any highlights there that you want to call out?

Yes. There's two answers to the question you asked. I mean we've modeled this as more impactful in the first 2 quarters of the year because that's when the most changes are happening. We've rolled out the kind of cut over to the new comp, kind of the new motion, but that kind of ramp-up of adoption, we've really modeled at that time. That said, I think the companies that have really leaned into this have continued to evolve it and tune it. And so I think that's where we've modeled this as an impact or headwind or change, I think this is something that through next year, we'll see as an accelerant, where there's opportunities to accelerate further based on the alignment we've created.

Awesome.

Yes, Jay touched on things that are changing as part of this consumption transformation, it's important to understand what's not changing. What's not changing is our business model. It's the same. What's not changing is how we recognize revenue. What's not changing is anything related to the Confluent platform business. And the reason I'm saying this is because a large amount of KPIs that we currently run the business with will continue to be the same. Three things I'll call out, which say from an investor perspective, it will be important to look into. The first is, we've called out in our earnings prepared remarks that subscription revenue will be a focus. Why? Well, it does a really good job of capturing the ACV for our Confluent platform and consumption for Confluent Cloud. So that will be a good, I would say, visibility into the organic momentum of the business. That's one. The second net retention rate has been a focus for us. And obviously, that's a good way to track the retention and expansion of our existing customers. So we call that out. And the third piece that Jay briefly touched on our focus with the consumption transformation is to make sure that our sales reps are having the next new use case conversation, which means that we are focused on driving consumption, and not getting the commitment from our customers. As a result of that RPO as a metric, where we've started to deemphasize that. We'll continue to report it as part of our accounting disclosures, but that's not going to be the only forward-looking indicator for our business. To summarize, like, I'd say, subscription revenue and NRR should be focused, and we should deemphasize RPO.

Yes, makes a little sense. To give Jay a bit of a break and to focus a little bit on your side of the house, Rohan, going back to sort of the trajectory of operating margins Q4 of 2021, the business was at negative 41% operating margins. I think last quarter, you guys were positive 5%. That's a massive amount of margin expansion in just 8 quarters. Do you feel that you got the sort of pace right? Did you improve margins too fast? And then broadly, how are you thinking from here the balance between growth on one side, but continue to get more efficient over time?

So when we think about resource allocation in general, like just philosophically speaking, it's never a 1-year exercise. It's a multiyear exercise. And the reason I say that is because when you invest in R&D, your returns typically come in 18 to 24 months. When you invest in sales and marketing, your returns come in 9 to 12 months. So any time we look at resource allocation, it's a multiyear process. And our true north is always the opportunity that's ahead of us, and how can we take advantage of that opportunity by driving durable growth over a long period of time. So that's the overall philosophy. To directly answer your question, no, I think we've been investing in the right areas. We did have a ruthless prioritization with respect to where we want to invest. And the proof is always in the pudding, if you look at our 2023 and 2024 product road map that Jay touched on, we've made this transition from a single product streaming company to a multiproduct data streaming platform. We have a series of unlocks, which start with the Flink GA in Q1, and the rest of our data streaming platform through the year. These are all, I'd say, proof points that show that we've been investing over the last couple of years and putting our money in the right places with respect to where we want to be. Long term, I mean, in 2024, our margin guidance is to be net neutral, margin neutral, which is another 7-point improvement from where we ended 2023. And we've also said that over the medium term, we want to be in the ZIP code of 5% to 10%, and that continues to be the case.

That's awesome. Let's talk a little bit about the 2024 revenue guidance, right? It no longer seems like with the strong results in Q4 and the sort of update to the guide, you're looking for more consistent growth throughout the year. You do have some easier compares starting to layer on. What would -- in -- if things sort of play out the way you guys hope. What are the factors that can go right that can deliver a potential upside to that outlook?

I mean our guidance philosophy has always been to make sure that we are setting guidance that's prudent and achievable. And as we are doing that, it's really important to provide the investor community with the right levers and what's going on in the business. And we've been fairly consistent with that. When I think about 2024, there are a few puts and takes. So number one is macro. The geopolitical situation, there's the Fed and the interest rates, there's elections going around, like 40-plus elections around the world. I mean that's a bucket that we do not control. So our thought process there is, let's assume it's going to be as is, and it's just going to be a continuation of where we are today. So that's number one. The second piece is around the product unlocks. We have our Flink GA happening in Q1 and the rest of the data streaming platform, we'll have unlocks. We're working on FedRAMP certification in partnership with NASA. So these are all, I'd say, things to monitor, but the real benefit is going to happen in fiscal year '25. And the last but not least, that Jay touched on is our go-to-market transition that we are making with respect to consumption transformation. And we've baked in the impact of that slightly more in the first half of the year versus the second half of the year. So our balance, we feel that all of these implications are baked into our guidance for 2024.

Awesome. Let's get back to talking about the core of the growth opportunity, Jay. Coming off that conversation we had on sort of the sales transformation that's sort of in flight right now. I think one of the fundamental aspects about your category of software is that you're competing for the workload of the use case versus, let's say, an application software or application SaaS, it's where more of a displacement market, right? Where you get 1 CRM usually, right? And so what initiatives does the team have in place for to make it easier for customers to go from, let's say, that initial use case, the use case 2, 3, 4 and 5 over time?

Yes. I would say there's a few things on the go-to-market side. And then a few things on the product side and then something that's a little bit inherent to the category. So on the go-to-market side, it's really about making sure that we have a clear picture of the use cases that are prevalent in each industry, and we can give people a very nice picture of, hey, what is it you could do, what's the art of the possible? How would you do it? What might competitors in this space be doing? And I think that, that helps you go from the most advanced companies who, of course, will figure all this out on their own to the kind of broad majority who like to see a clear road map of those who've gone ahead of them. That's something we've put significant energy into. We have continued investment in partners that help especially SIs kind of take you out into the broader set of initiatives and transformation the company made orchestrating. And I think that's an area that's just really gathering scheme for Confluent is contributing this year but will continue to grow in the years ahead. Both of those, I think, are important. On the product side, it's really about completing this data streaming platform, right? If you're just offering Kafka, that kind of low-level stream of data, of course companies can bake that into their applications, and they do, right? That has enormous traction in open source. But as you start to have connectors that just plug it in off the shelf as you have real-time processing capabilities in C language, which is kind of the universal language of data as well as programmatic capabilities. suddenly gets easier and easier and easier to build applications in this way. You can take something from a 9-month application cycle to being able to build very simple data pipelines in days, right? And that's a huge deal in terms of accelerating the adoption. The most mission-critical applications will always be carefully built and tested over a long period of time, but making the easy things fast is absolutely important for us. So that's on the product side. And then the last accelerant, I think is really foundational to the category, which is there is a kind of inherent network effect within companies for streaming, and that's because these streams of data often go between parts of the company. The goal is to connect or to share the real-time flow of data. So as a company spins up this area, the first use cases are kind of siloed applications. But over time, you have these critical data streams, which draw in those new applications organically. And that's what we've seen take hold in our largest customers. And it makes sense in a retailer, maybe the first use case would be, hey, we need to get the kind of real-time flow of sales for some kind of marketing, pricing, promotion use case. But sure enough, what's selling retailer is one of the most critical data sets they have. And it's very likely that the only place you can get the real-time view of that is Confluent. And that then becomes the basis for all kinds of applications, whether it's managing inventory, logistics, analysis, analytics, fraud, a whole set of things will ultimately feed off of that. The first one came for the capabilities of the platform, but the latter ones mostly came for the data. Of course, they benefit from the capabilities, but they came because that was the only place you could get that stuff. And that's what helps us get to scale so that the end application is a lot less work than the first one.

Awesome. This time last year, we were all talking about AI, GenAI and we're kind of think as a community as a tech community kind of guess where this is all going and probably still are to a large degree. How have you sort of -- is there any update to your views on the role of streaming, stream processing as it relates to enabling GenAI I mean, in views to applications and the building of those applications? What's the role here?

Yes. At least the short-term changes we've seen have played out as we'd hoped. And so I think it's been to solidify. And then, of course, over time, we may see quite a lot in this area, and that will be interesting. So what did we think the two opportunities were in this area? Well, the first one was, there's a bunch of AI companies that are building out their infrastructure stacks, and they need streaming the same as every other tech company. So we went -- sold to all those companies and we brought in awesome customers. Last earnings call, we talked about, OpenAI, which became a substantial customer, but really broadly across that set of next-gen companies, there's an opportunity for us to sell to them. So that's the smaller thing. The bigger thing is when we look at our larger customer base, we're seeing all kinds of enterprises bring together their data with these large language models to augment customer service, other customer interactions, really drive productivity in their employee base, trying to help people do what they do. And we're seeing that across all different types of departments, all different types of businesses. That type of application is really following an architecture, which has come to be called retrieval augmented generation or RAG. And it's really about, hey, how do I bring my data together with this language model. The language model kind of news about the world at large. My data is something that is more up-to-date with the current state of the world has to be tightly controlled in terms of who can see what and that's what I needed. If I have a customer service rep that needs to answer a question or if I want to directly answer a question with some kind of chatbot to you, I have to know about what you're doing with my business and product right now. Otherwise, the answer is not going to make any sense. And that really drives a lot of data flow and integration across companies. And that's the use case that we've seen most prevalent with our customers is that kind of data supply chain for these AI use cases. The reason we're so bullish about it is because these applications are not rocket science to build, we've been playing a similar role in the architecture from older kind of predictive machine learning and AI applications going back since the beginning of Kafka. That was actually one of the reasons I help to make it at LinkedIn was to power that type of thing. But the reason that this is so appealing now is both the power of these applications, but also just the fact that it's a very achievable thing for all types of businesses to actually put this to practice, whereas a lot of the predictive machine learning applications were pretty hard to build and operationalize. pretty hard to get good results for. And so I think the scope of what we're seeing is just much, much larger. I think people know that, but that's why we see it as such a promising driver in the business.

Awesome. I want to talk a little bit about Flink, but if anyone in the audience has a question, just raise your hand, we'll get the mic to you. But let's talk about Flink, you mentioned it's going GA in Q1. If we use Kafka in streaming as a baseline, how do you think the adoption of stream processing for Confluent will compare?

Yes. Yes. So there's two questions. What will the pace of that adoption be? And then what's the end state? Like how much value is there in this processing layer? And so I'll start with the end state because I think that's probably the most important is why we invested in this space. If you look at data platforms, in the data at rest world, it would kind of divide into storage and processing. And that's where a lot of the value is. Databases bring those two things together. And it can be hard to pull apart how much value is there in processing and how much value is there in storage. But indeed, you can look at some of these cloud applications that price them separately like Snowflake and you'd see that the majority of the money they make is on the processing. And logically, if you talk to customers, they would say, yes, that processing, that's our business logic. That's the intelligence. That's what we bring to the data that's quite important. And so if you take that analogy seriously, then in the streaming world, you would believe, hey, this processing opportunity is quite significant. Furthermore, when we look at how customers build around data streams, their spend on their custom applications that have that logic is significantly larger than the spend on the data stream itself, certainly on the order of 5x or so larger, right? So if you're able to build a platform that makes that easier for them, that makes it more cost effective because they don't have to stand up a bunch of custom servers that makes it kind of more elastic, more fault tolerant and reliable, easier to build, then you can capture a lot of that larger spend in addition to the data stream itself. So that's kind of the first principle of thinking. The other way to get there is what's that near term look like. So we've seen the adoption of the open source for Flink. That's really become a kind of de facto standard in the stream processing world. It offers a whole set of interfaces around real-time data across the popular languages that developers would use, really rich and thriving community around it. And so we feel like, hey, that can be on a very similar trajectory to Kafka itself in terms of the adoption. And with a bit of an accelerant because, of course, we can take this out to our existing customer base, whereas we had to go and land all those customers the first time. One of the wonderful things about a consumption model is how little friction there is in adopting the next piece of functionality. And so when I look at a business like AWS or maybe a little closer to us as a stand-alone company, Datadog. I think one of the things they did so well was that expansion from one thing to many, many more things. And if you think about what attaches to streaming data, the answer is like so much, right? There's so much that can kind of be pulled into that orbit that kind of data gravity, the functionality you can offer around it. And so we think that, that's a huge opportunity. Now I do want to kind of temper expectations. So a lot of people think, well, okay, if you've released it in Q1 with GA, and we should see Confluent revenue doubles in Q2, right? And the reality of cloud infrastructure services is the ramp is more gradual than that, right? People have to first see it be a solid target to build against. Then they have to build their applications, get them to production where they kind of run at scale and generate workloads. And then those workloads have to accumulate to add up to the revenue. So it's not an L-shaped curve. But we do think kind of coming into next year, this is a significant tailwind for the business. That's certainly what we've been kind of building towards and our expectation for it.

Great. Let see if there's any questions from the audience back there?

You mentioned Amazon and Datadog. I would add Microsoft as companies that reported and sounded quite good on the consumption patterns for this year. I think a lot of people in this room are probably confused because some of the January quarter companies that are reporting more recently, are singing a different tune, so to speak. So just any perspective you can give us on kind of the state of the world in software. Any changes would help?

Yes, yes. I would draw out two factors. So I do think we've seen at least some amount of stabilization in patterns. I don't know that, that means it's not like a recovery. It's just that if you think about '23. I think especially in tech, a lot of companies were in heavy optimization mode where they're literally all resources are going to that. I think as you get out into the larger economy, even some of that flows through to that, where a lot of IT departments were more focused on kind of getting the value out of the cloud spend they'd already made than they were on kind of the next new project. I think we're seeing a little bit of moderation of that, but I don't think it's like a complete reversion to 2021 spending patterns by any means. And I think that kind of plays out across the larger set of companies. So I think there's one other factor that you do see, which is I do think companies that serve kind of production use cases, like kind of mission-critical applications type use cases. I do think that they tend to see a more continuous build because there's less room typically to kind of optimize it away. Oftentimes, the things which are the easiest to add, that could be some kind of monitoring observability framework. It could be some kind of analytics thing on the side, those things are often also the easiest to take away. So they kind of -- they come the easiest, but they may go the easiest as well. And so I think you might see a little bit of dichotomy there as well, but I'm not a great prognosticator of results for other companies. So take that for a grain of salt, but that's certainly one of the things we've seen in our business is what drives us is the pace of new application development. There's typically less optimization. There's always some optimization in any business of how they use the product. But because these are production applications that kind of come out pretty well thought out and optimized. Whereas if I look at, say, our data warehouse usage, for sure, it's not that hard for us to be like take all the reports nobody looks at and get rid of them, or take the retention of data and shrink that down. There's lots of optimizations we can make, which are much harder in the kind of production application setting.

And with that, we've got to end it there. Thanks so much, Jay and Rohan...

Thank you so much.

Thank for having us.