Confluent, Inc. (NASDAQ:CFLT) Q2 2025 Earnings Call Transcript

Confluent, Inc. (NASDAQ:CFLT) Q2 2025 Earnings Call Transcript July 30, 2025

Confluent, Inc. beats earnings expectations. Reported EPS is $0.09, expectations were $0.08.

Shane Xie: Welcome to the Confluent Second Quarter 2025 Earnings Conference Call. I’m Shane Xie from Investor Relations, and I’m joined by Jay Kreps, Co-Founder and CEO; and Rohan Sivaram, CFO. During today’s call, management will make forward-looking statements regarding our business, operations, market and product positioning, growth strategies, financial performance and future prospects including statements regarding our financial guidance for the fiscal third quarter of 2025 and fiscal year 2025. These forward-looking statements are subject to risks and uncertainties, which could cause actual results to differ materially from those anticipated by these statements. Further information on risk factors that could cause actual results to differ is included in our most recent Form 10-Q filed with the SEC.

We assume no obligation to update these statements after today’s call, except as required by law. Unless stated otherwise, certain financial measures used on today’s call are expressed on a non-GAAP basis, and all comparisons are made on a year-over-year basis. We use these non-GAAP financial measures internally to facilitate analysis of our financial and business trends and for internal planning and forecasting purposes. These non-GAAP financial measures have limitations and should not be considered in isolation from or as a substitute for financial information prepared in accordance with GAAP. A reconciliation between these GAAP and non-GAAP financial measures is included in our earnings press release and supplemental financials, which can be found on our IR website at investors.confluent.io.

References to profitability on today’s call refer to non-GAAP operating margin unless stated otherwise. And with that, I’ll hand the call over to Jay.

Edward Kreps: Thanks, Shane. Good afternoon, everyone, and welcome to our second quarter earnings call. Confluent delivered a solid second quarter, highlighted by a 21% growth in subscription revenue, 28% growth in Confluent Cloud revenue and non-GAAP operating margin of 6% — up approximately 6 percentage points. Additionally, our DSP monetization continues to gain traction with Flink ARR growing approximately 3x over the past 2 quarters. This is a testament to our complete data streaming platform strategy and our strong positioning for the future shaped by agentic, real-time AI. Before getting into the broader business update, I’d like to start by sharing some observations on our cloud business. In Q2, our larger customers continued their optimization efforts and adopted new use cases in a more measured pace.

While we are confident that this elevated level of optimization will eventually subside, our outlook for the second half assumes consumption growth notably below what we’ve seen in the same period of prior years. Rohan will provide further details in his remarks. Encouragingly, we’ve seen some customers commit to larger multiyear deals following the optimization efforts they undertook last year. This helped accelerate our RPO growth to 31% in the quarter, reflecting the deepening of our customer relationships as they plan for long-term growth. To accelerate use case expansions and support the long-term growth trajectory of our cloud business, we’re driving operational enhancements across several areas in the business. This includes two key focus areas Ryan Mac Ban has identified following his first 90 days as Chief Revenue Officer.

Q&A Session

First, we’re improving coverage ratios between AEs, SEs and post sales roles to strengthen execution in the field. This higher-touch integrated approach enhances account ownership and provides tighter customer alignment in driving use cases into production across our enterprise customer base. This has shown early results in the second quarter as we’ve seen a sequential increase of more than 40% in late-stage pipeline progression. Second, we’re accelerating the build-out of our DSP specialist team to drive multiproduct selling. This team focuses on building repeatable high-impact sales plays that include pricing strategy, go-to-market messaging and streamlined migration offerings that combine tooling and professional services. We’ve also seen early signals of success with this specialization model with several customers accelerating their production go-live of DSP use cases in the quarter.

Together, these changes are designed to enable the field to move faster and unlock greater value from our platform selling strategy. In parallel, we’re doubling down on three areas where we’re already seeing strong traction. The first is replacing CSP streaming offerings with Confluent. We’ve had success displacing these CSP offerings with win rates well above 90%. This is an area where we feel our product capabilities and TCO story have improved enormously over the last year with differentiated offerings like Freight Clusters, Enterprise Clusters and WarpStream. Already in Q2, we saw more than two dozen displacements against a single CSP offering. We plan to amplify this success by intentionally targeting these offerings and increasing our number of at bats against these competitors.

Speaking of WarpStream, we’re seeing positive trends there as well. The large majority of our WarpStream business in Q2 is incremental. Even in existing customers, we’re seeing customers increase their spend with Confluent through WarpStream while actually lowering their overall cloud infrastructure costs. For example, two customers, a major retail investing platform and a leading prepaid mobile provider both deployed WarpStream for their high-volume logging and telemetry workloads in Q2. These customers increased their spend with Confluent by 30%, while decreasing overall CSP infrastructure costs roughly 50%. It’s a great example of how we’re helping customers scale efficiently while delivering meaningful cost savings. The second area that we’re doubling down on is our partner ecosystem.

Partners are instrumental in broadening our footprint and driving customer expansion, especially as we scale into a multiproduct platform company. We continue to see incredible traction in this area. In the past year alone, we’ve launched a new OEM program and partnered with leading AI vendors to launch a new AI accelerator program. At the same time, we’ve deepened key partnerships with Jio, SCCC, Databricks, EY and most recently, Infosys. This expanded collaboration with Infosys, a global leader in next-generation digital services and consulting is the first major partnership under this new investment. As a partner in our OEM program, Infosys has seen firsthand the growing demand for data streaming. This is a meaningful step forward in our broader strategy to deepen partnerships with leading system integrators.

To underscore the strategic value of our partner ecosystem, well over 20% of our business over the past year has been partner-sourced. Looking ahead, our partner ecosystem will be an important area of continued investment and co-innovation. We believe deepening partner engagement across Confluent Platform and Confluent Cloud will fuel our growth and accelerate our global market penetration. The value of our partner ecosystem can best be understood through our customers’ lens. A leading global financial market infrastructure provider that processes trillions of dollars of security transactions daily set out to create a shared Kafka service across its business. The goal was to enable real-time data streaming at scale with the kind of governance needed for a systemically important institution.

However, they face challenges with the specialized staffing required and the complexities of operating Kafka as a shared enterprise service. As the organization’s long-time transformation partner, EY went beyond pure technical guidance. They helped define a broader vision, positioning Confluent as the strategic foundation for enterprise-wide data streaming. After the deal closed, EY and Confluent partnered to launch a modern streaming center of excellence that helped the company evolve from siloed messaging to a unified, enterprise-wide streaming strategy. With EY’s trusted relationships, the focus has shifted to scaling high-impact streaming use cases across the business. Together, EY and Confluent are building a foundation for sustained innovation enabling this market leader to turn real-time data into a true competitive advantage.

The third area where we’re doubling down is Flink. While Flink is still a small part of our overall business, it has experienced exponential growth with the sequential dollar increase in ARR accelerating for 4 consecutive quarters. Our Flink business is approaching $10 million in ARR and nearly tripled over the first half of the year. This includes strong contributions from both Confluent Cloud and Confluent Platform with a fairly even ARR split between the two. We now have three customers with more than $1 million in Flink ARR and a diverse rapidly expanding base of customers well into their first set of use cases. Capturing the processing of real-time data is one of the most strategic elements of our DSP strategy. This allows us to make real-time use cases much easier to build and to capture the spend on these use cases.

The rapid growth of our Flink offering is evidence that this strategy is working. Wix is a great example of the power of our Flink offering. Wix is the global platform behind more than 100 million websites, serving 1 billion users every year. As they expanded into analytics and AI-driven personalization, it became clear that they needed a more scalable real-time data infrastructure. Their batch pipelines and self-managed Kafka setup simply couldn’t keep up. To support their next stage of growth, Wix turned to Confluent Cloud and our fully managed Flink offering. Today, they process over 30 billion events per day in real time across multiple regions and clouds. Flink is now central to Wix’ data architecture. It filters, enriches and joins data streams in real time, powering hyper-personalized web experiences, live A/B testing and Wix analytics, which gives users and developers immediate insight into site activity.

With Flink and Confluent’s governance tools, Wix delivers low-latency trustworthy data at global scale. That’s helped them increase developer velocity, improve customer experience and cut down on operational overhead. Confluent is now a key part of Wix’ long-term data platform strategy. And finally, we’ve been excited to see AI workloads beginning to move towards production in rapidly growing volumes. In 2024, much of the enterprise use of AI was early experimentation with only a few dozen production use cases. This year, we expect production AI use cases to grow 10x across a few hundred customers. A few of my favorite examples from Q2. A public sector organization in New Zealand is deploying AI agents to automate complex regulatory workflows and cut citizen response time from hours to minutes without operational overhead.

An astronomy institute is deploying AI agents to process telescope alerts in real time to filter noise and catch rare fast-fading cosmic events before they’re lost. A major Philippine power company is deploying AI agents to interpret real-time alerts, surface critical failures early and prevent million-dollar outages. An international sports network is generating real-time commentary that adapts to the flow of the game and player performance. Let me go a little deeper on one such use case. We’ve talked about Notion before when they turned to Confluent after it became clear their data infrastructure couldn’t scale or support their AI vision. Since then, they’ve made Confluent a much more strategic part of their business to accelerate the rollout of new AI capabilities.

With over 100 million users, Notion needed a scalable real-time data architecture to power AI-driven search, content generation and integrations. Their legacy messaging stack couldn’t keep up with the volume of product activity, slowing innovation. By adopting Confluent Cloud, Notion built a fully managed event-driven architecture that supports key use cases across their platform. Using our prebuilt connectors, they stream data into Snowflake and Amazon S3 to enable real-time analytics and AI workloads. Stream processing and schema registry ensures that every change in app is reflected instantly in their vector database, keeping Notion AI accurate and responsive. With Confluent, Notion has tripled platform team productivity, reduced operational overhead and accelerated time to market for AI-powered features.

Today, Confluent is the real-time backbone of Notion AI. In closing, while we’re continuing to see some near-term consumption headwinds, I remain highly confident in the strength of our business. With our differentiated and complete data streaming platform and strong partner ecosystem, we are well positioned to capture a meaningful share of the $100 billion plus data streaming market. With that, I’ll turn it over to Rohan.

Rohan Sivaram: Thanks, Jay. Good afternoon, everyone, and thanks for joining our earnings call. Our second quarter was highlighted by solid top line growth and continued margin expansion. These results underscore the strength and flexibility of our data streaming platform, helping customers unlock the full value of real-time data across cloud, on-premise and BYOC environments. Turning to the Q2 results. Q2 subscription revenue grew 21% to $270.8 million and represented 96% of total revenue. Confluent Platform revenue grew 12% to $120.3 million, reflecting solid performance in financial services and sustained momentum with our OEM partners. Cloud revenue grew 28% to $150.5 million, representing 56% of subscription revenue compared to 52% in the year- ago quarter.

As Jay mentioned earlier, consumption growth was impacted by continued optimization with month-over-month trends trailing the same period in prior years. Additionally, an AI-native customer has been making a broad-based move towards self-management of internal data platforms, reducing their Confluent Cloud usage as a result. We continue to support their data streaming needs and have now closed a Confluent Platform deal with them in Q3. This represents a significant reduction in total spending with Confluent starting in Q4 and is expected to dampen our Q4 cloud revenue growth rates by low single digits. Turning to the geographical mix of total revenue. Revenue from the U.S. grew 15% to $164.3 million. Revenue from outside the U.S. grew 29% to $117.9 million.

Moving on to rest of the income statement. I’ll be referring to non-GAAP results unless stated otherwise. While driving top line growth at scale, we continue to show significant operating leverage in our model. In Q2, subscription gross margin increased 70 basis points to 81.5%, above our long-term target threshold of 80%. Operating margin increased 570 basis points to 6.3%, exceeding our guidance of approximately 5% and reflecting our continued focus on driving efficiencies across the company. Adjusted free cash flow margin increased 270 basis points to 3.9%. Net income per share was $0.09, using 367.3 million diluted weighted average shares outstanding. Fully diluted share count under the treasury stock method was approximately 380 million.

We ended the second quarter with $1.94 billion in cash, cash equivalents and marketable securities. Turning now to customer metrics. On a year-over-year basis, total customer growth was in line with average growth rate of the previous 4 quarters. $20k-plus ARR customer count grew approximately 8% to 2,497 and represented more than 95% of ARR. $100k- plus ARR customers increased 10% to 1,439 and accounted for greater than 90% of ARR. $1 million-plus ARR customers grew approximately 24% to 219. New $1 million-plus ARR customers continued to come from a wide array of industries and include a conversational AI and automation company, a global food service distributor, a Fortune 500 insurance provider, a cloud-based video platform and a quality management software company for life sciences.

NRR for the quarter was 114%, reflecting ongoing consumption headwinds in our cloud business while GRR remained close to 90%. Turning now to guidance. Based on current consumption patterns, our outlook for Confluent Cloud assumes month-over-month growth rates for the remainder of the year, will remain notably below what we’ve seen in the same period of prior years. Given Confluent Platform’s pipeline visibility in the back half of the year, we are raising our full year growth expectations for Confluent Platform. This strength partially helps offset some of the consumption headwinds in our cloud business. For the fiscal third quarter of 2025, we expect subscription revenue to be in the range of $281 million to $282 million, representing growth of approximately 17%.

Non- GAAP operating margin to be approximately 7% and non-GAAP net income per diluted share to be in the range of $0.09 to $0.10. For fiscal year 2025, we are increasing the low end of our guidance range by $5 million, and we now expect subscription revenue to be in the range of $1.105 billion to $1.11 billion, representing growth of approximately 20%. Non-GAAP operating margin to be approximately 6%. Non-GAAP net income per diluted share to be approximately $0.36 and adjusted free cash flow margin to be approximately 6%. For modeling purpose, we expect cloud as a percentage of subscription revenue for Q3 to be approximately 56% and Q4 to be approximately 55%. Now I’d like to provide an update on the four strategic pillars of our growth: streaming, DSP, AI and our partner ecosystem.

First, we remain well positioned to lead the core streaming market across on-prem, BYOC and cloud. Confluent Platform’s continued strength has been driven by solid performance in financial services, early traction with partners and our team’s consistent execution. WarpStream consumption exhibited fast growth in Q2, benefiting from customers migrating latency relaxed workloads from open- source Kafka to drive cost savings while maintaining full control over their data. While consumption headwinds persist in our cloud business, we believe our two strategic focus areas along with three targeted double-down initiatives will begin delivering meaningful results in a few quarters, helping accelerate our land and expand momentum across customer acquisition, use case expansion and DSP monetization.

Second, we are encouraged by the growing traction of our DSP portfolio across both cloud and on-prem environments. As Jay discussed earlier, in just two quarters this year, Flink ARR grew approximately 3x, approaching $10 million with a fairly even split between cloud and on-prem versions of the product. This validates our strategy of building a complete platform for real-time data everywhere and our ability to take advantage of the shift left opportunity for stream processing. Third, Confluent’s strategic importance in AI is only getting stronger as the world expands from GenAI to agentic AI. Over the past year, we have seen firsthand AI use cases in production growing from chatbot, semantic search and content creation to code generation and iteration, multi-agent orchestration, agent recommendations and much more.

As Jay mentioned, this year, we expect the number of production AI use cases to grow 10x across a few hundred customers. And fourth, we are seeing sustained momentum in our partner ecosystem. In less than a year, we have expanded multiple strategic partnerships, including Jio, SCCC, EY, Databricks and Infosys, while continuing to build strong partnerships with Accenture, Deloitte, TCS and more. Partners have sourced well over 20% of our business, and we are capitalizing on this momentum by continuing to invest in our partners to unlock more revenue streams and to further expand our global reach and impact. In closing, we’re pleased with our solid top line growth and margin expansion at scale in the second quarter. While there’s still work to do in accelerating new use case expansion, we are encouraged by the traction we are seeing across core streaming, DSP, AI and the partner ecosystem.

We believe each of these areas represent a key driver of durable profitable growth as we look ahead. Now Jay and I will take your questions.

Shane Xie: Thanks, Rohan. [Operator Instructions] We ask that you limit the Q&A to one question and one follow-up. And today, our first question will come from Matt Hedberg with RBC followed by Deutsche.

Matthew George Hedberg: I guess I wanted to understand a little bit more of the consumption optimization trends you talked about? And if that’s more macro or company-specific and realizing you noted there was an AI customer that feels like they’re changing some of their consumption. But was there anything recurring in some of these conversations? And how prevalent were they? I think last quarter, you mentioned it was a handful of top 20 customers. But just trying to get a sense of how that trended sequentially.

Edward Kreps: Yes. I would say it’s a similar dynamic. I do think this is broadly of the same sort of what we’ve seen across other companies where customers are happy, they plan to be using more data streaming over time are putting effort into making sure what they bought, they’re getting the most value out of. This hit us a little bit later than some of the other consumption companies but it’s kind of persisted a few quarters longer. And I would put the AI native customer in sort of a different category where it’s not really an optimization thing at all. They’re broadly moving into a different way of kind of operating internally. And I think this is across a number of different vendors, including us. So we were happy to be able to support them with Confluent Platform deal, they continue to use our cloud product in more limited use cases, but there is a kind of overall reduction in spend, and it’s definitely a headwind for Q4 for cloud.

That’s a number of kind of positive forces at work here. So if you look at where these customers are going longer term, I think we called out the 31% growth in RPO. And I think that is, in large part, a bunch of customers upping their overall commitment. That tends to be a headwind to short-term consumption, bigger commitment means a little bit higher discount levels, but overall kind of gives an indication of the trajectory of their spend.

Matthew George Hedberg: Well, I guess on the other — that’s helpful, Jay. And I guess the other thing that you pointed out is just the growth in production AI workloads across a couple of hundred customers. I guess I’m wondering it feels to us like the relevancy of streaming and processing is elevated in an AI first world. How should we think about some of those production workloads eventually positively impacting subscription growth as we think forward over the next year or so?

Edward Kreps: Yes, yes, I think it’s very positive for us. Ultimately, this is turning into, I think, a really important ingredient in the architecture for AI applications. And it makes logical sense like if you want to have some kind of agent that’s taking action in the business, it has to have an up-to-date set of context data on what’s happening across the business. And so I think we’ve certainly found ourselves drawn into a set of use cases around that. I think the Notion story is a great one, but there’s a number that I called out across different industries and domains. So I think that’s a very positive for us. And not the only one. I think the Flink growth is a huge deal. It’s $10 million in ARR, which is very small, but it grew 3x over the last 6 months, and that’s actually very fast growth for something in the infrastructure space.

And in many ways, Flink is the crux of that DSP expansion for us where we feel like going from an important ingredient in the real-time architecture to a full platform. The hard part of that is capturing the application workloads like the real-time processing, we know what that’s about. But the question is, can you build a product that actually does it and that customers can use and benefit from. And I think that’s gone quite well where it’s now producing across both cloud and CP and very nice growth rate. So not something that determines the overall number at this point, but we’re very excited by that early progress, and we want to see it continue.

Shane Xie: Thanks, Matt. We’ll take our next question from Brad Zelnick with Deutsche Bank, followed by Morgan Stanley.

Brad Alan Zelnick: Thanks so much, Shane. And nice to see you guys. Jay, a lot of exciting things happening, a lot of good data points coming out of this quarter. But at the same time, you still continue to be surprised by this optimization activity that’s occurring with your large customers. You then talked about Ryan Mac Ban 90 days in, and the two key focus areas and operational enhancements that he’s making. I was just wondering, is that in relationship to what’s happening with large customers? Or is that — any optimization that you’re seeing or is that completely separate? And can you maybe just slow down and explain a little bit why you’re confident that these investments that you’re going to make in the coverage ratios across AEs, SEs and post-sale support as well as the build-out of the DSP specialist team is really going to make a difference and over what time line.

Edward Kreps: Yes, yes, it’s a good question. So, yes, I would think of it this way. There’s some amount of optimization customers are doing at any given time, right? I think that’s been exaggerated in recent quarters, right? And then when you think about what’s the balance, the balance of growth is new use case additions minus optimizations. And when we think about what we’re in control of, it is these new use case acquisitions. Are we going out and winning the new workloads? Are we making sure we’re connecting with those? Are we bringing in the right customers. And so that’s obviously where our focus is. The — I think we’ve seen very good early results from these changes that Ryan has made on some of these alignment things and the SE ratio, this was one of the changes we made, both to get go-to-market costs in line and through our consumption change.

I think just didn’t quite work the way we wanted. We were able to change to a better coverage model without negative cost impact. And I think that’s paid off in some of the progression of streaming projects that we’ve seen already. So I called this out just like we measure the early stages of this in terms of kind of late-stage pipeline, how is that trending? And this is consumption pipeline, so it’s actual customer workloads heading to production, that’s up quite substantially, I think I said greater than 40%, Q1 to Q2. And I think that’s a good result that came out of some of the operational improvements there. So that’s the early indicator. Now obviously, we don’t count our chickens till they’re hatched, but those are the things we look at when we think about what the kind of forward momentum is.

Brad Alan Zelnick: Great. That’s helpful, Jay. And maybe Rohan. For you, just as we think about your messaging coming out of last quarter, again, surprised coming out of Q2 as well. What can you tell us about the approach to guidance and whether it’s in the form of conversion rates or anything else to help us really appreciate what is expected in the back half and how much this may or may not be derisked at this point?

Rohan Sivaram: Yes. Brad, thanks for your question. Historically, as I’ve said in the prior call as well, we’ve typically seen a quick consumption rebound after, say, a quarter of optimization. Best example, recent example is Q2 to Q3 of last year. Specifically, what we saw in the quarter was the larger customers optimizing that continued and the adoption of new use cases was more measured. Given the dynamic of Q2 and Q1, what we are doing right now for cloud is we are primarily assuming the cloud outlook for month- over-month growth rates to be notably below what we’ve seen in the same period of prior years. So that’s what we’ve done for cloud. So despite this dynamic, I want to take a step back and just talk a little bit about the total guidance.

We’re actually raising our fiscal year ’25 subscription revenue guide at the midpoint. And what’s supporting this is the strength in our CP business. We have visibility into second half pipeline. And also on the cloud side, as Jay briefly touched on, we are seeing a bunch of green shoots. First of all, the Flink momentum. When you think about Flink, it has approximately tripled in the first 6 months of the year, closing in on $10 million in ARR. The late-stage pipeline projection is we saw a sequential improvement of greater than 40%. And finally, some of our customers committing to larger multiyear deals following periods of optimization. That shows up in our RPO numbers. So when you kind of take all of this together, the back half guidance is for cloud.

We’re not assuming month-over-month growth rates. It’s actually notably below historical averages for same periods in prior years. For platform, we have visibility into our pipeline. And for the cloud green shoots, it’s like a portfolio approach. There are upside levers in that, that we will realize. Like Jay said, we’re not counting our chickens before they hatch.

Shane Xie: Thanks, Brad. We’ll take our next question from Sanjit Singh with Morgan Stanley followed by Bernstein.

Sanjit Kumar Singh: Jay, I wanted to go back to some of the go-to-market changes that you guys have been rolling out over the past couple of years, particularly the move to compensating sales reps on incremental consumption. Just how has that been going? And to what extent is that still a friction or non-friction point when it comes to driving incremental cloud consumption growth?

Edward Kreps: Yes. I think overall, it’s gone well. I think it was a critical change for us just to be aligned to what the company is trying to drive and to actually be able to unlock the use cases, be able to have conversation on some of the DSP offerings that might get added on initially, even after your commit. So there are a whole set of motivators, I think it’s actually helping with — there’s certainly been adjustments we’ve made along the way, including the things I called out in the script to try and make sure we’re really optimizing for it. So I think that those will help us realize even more results from it.

Sanjit Kumar Singh: Awesome. And then on Flink, on that, we noticed in our customer conversations that Flink was sort of building the momentum as well. I know that Flink started as a part of Confluent Cloud. I think — was it second half of last year, middle of last year, you sort of introduced it into Confluent platform. Was that like the unlock? And can you sort of explain to us like when we think about the Flink opportunity, why is it such an even balance between cloud and on-prem?

Edward Kreps: Yes, it’s — both good questions. So yes, it’s been a buildup on Flink because we had to — because of German law announced one of the acquisitions we made. As we’re doing it, we had to tell people we’re doing something with Flink well before we had the product out. It did get into the market mid last year. Obviously, for us, then we have to ramp it across the different cloud providers, open it up for the kind of private networking types that certain customers have to really get the full unlock. And then we’ve seen, post that great monetization results. And I think that will increase there’s future unlocks coming there. This area of data processing, people know what it is. They know they watch it, they know they need to do it in real time.

It’s a question of really getting something that’s super solid that does everything a modern platform does, but does it continuously and in real time. So I think it’s been exciting to kind of get to that point with customers. That ramp has been quite steady. So if you look at cloud, just the growth quarter-over-quarter, but if you were to look into it month- over-month, it is just a very steady ramp. The reason for that is it is a serverless offering. So adopting Flink costs you nothing. It’s each incremental query or workload that you’re kind of adding, that’s building up that consumption. And that is the nature of that business. So kind of building that momentum then becomes a very powerful force even in customers that have adopted, they’re continuing to add queries and grow their consumption.

The CP Flink is a little different where the tendency is you’re kind of pre-deploying, right? So it will come in bigger chunks. Like CP, it will tend to appeal to some of these larger customers that have data centers, has a little bit more of an opportunity to kind of migrate [ in-place ] workloads. So for that reason, it tends to be the smaller number of customers and capturing a bit more in each chunk. So a little bit different between both, but the kind of net-net is we’re contributing — both are contributing. We’re serving customers across both and both are actually growing very nicely between the two.

Shane Xie: Thanks, Sanjit. We’ll take our next question from Peter Weed with Bernstein followed by William Blair.

Peter Weed: Appreciate all the detail you’ve been giving, particularly around some of the optimization that’s been going on. One of the things that struck me is we’ve been focused on these big customers. But I guess also when I’m taking a look at some of the customer bands that you report, the kind of $20,000 to $100,000-size customers, which I would think would be kind of your fastest-growing kind of future cohort that will hopefully graduate into those $100,000-plus, $1 million-plus customers over time, actually has been probably the weakest of any of those cohorts this last quarter. How should we think about the kind of incremental number of customers being added to that segment as kind of a future signal for growth and perhaps it’s some of the sales initiatives that are going on that are going to be trying to drive a lot more customers into that kind of early phase that turn into the very big customers over time.

Edward Kreps: Yes, yes, so if you look across the customer bands, as you say, kind of strength in $100,000-plus, strength in $1 million-plus, lighter on the $20,000-plus. And we think that’s a key metric. So that is a point of focus for us. Heading into this year, I do think we made some changes that lost some of the focus there. And some of what I described, we are trying to make sure that we’re nailing that. So in particular, I do think these CSP takeouts that I described are an awesome opportunity for that, that’s an area where we’re seeing a lot of early success. There are a lot of customers out there that have adopted one of the offerings from the cloud providers. And the reality is those offerings have never been great.

But increasingly, as our product portfolio has kind of filled out, we have something that’s not just a better offering, but it’s actually a better deal. And so you kind of have something that’s more complete, better performance and kind of better price point with freight and these enterprise clusters, WarpStream. And so early results from that are quite good. And we think that, that particular program and a focus on those lands is a great way of kind of getting out to more breadth.

Peter Weed: And is there anything that we should read into that around kind of increased churn levels? I noticed some of the commentary changed around gross revenue retention. I think historically, you’ve said, “Hey, it’s above 90%.” This quarter, you said it remains about 90%. I didn’t know if that was like an increased churn and that was being seen in that customer segment or I’m just reading too much into that commentary?

Edward Kreps: Yes. I wouldn’t attribute too much to that segment overall. I think it’s mostly about really driving the lands in that segment as the biggest contributor. I don’t think you’ve commented on that at all, Rohan.

Rohan Sivaram: I think [indiscernible], Jay.

Shane Xie: We’ll take our next question from Jason Ader with William Blair followed by Mizuho.

Jason Noah Ader: Can you hear me okay?

Edward Kreps: Yes, loud and clear.

Jason Noah Ader: Okay. Jay, I understand the challenges in predicting the customer consumption patterns and definitely appreciate transparency on what’s happening at the brand level. But I think we all expected some improvement in the business following the refinement in the sales comp model last year. The broadening of the product set — I mean pretty significant broadening of the product set over the last 1.5 years or so. And then greater amount of AI adoption, where we’re obviously getting closer to the tipping point. Now it feels like we’re kind of back to square one a little bit and waiting for things to get better. So why should investors believe that this time is different?

Edward Kreps: Yes, yes, it’s a fair question. I think a lot of the things we said would contribute are starting to do that. I think we shared a little bit of that, right? We’ve talked about some of the DSP offerings. I think we shared a bit about what’s happening with Flink. I think we’ve talked a little bit more quantitatively about some of these AI use cases and what we’re seeing there. There is a headwind with some of the existing large customers and optimization. I do think that these things are smaller but growing fast. They eventually do predominate as they grow, but to cancel each other out, they have to get to the scale that matters. So in those positive tailwinds, I would include those things I just mentioned. I would include the buildup of RPO and commitments in cloud, which I think is a positive sign in terms of what customers’ intentions are over time.

As well as that overall progression of pipeline, which I do think is a reflection of kind of the focus on this within our sales organization.

Jason Noah Ader: Got you. I mean it’s got to be frustrating for you…

Edward Kreps: Well, there’s different forces. I mean if you step back, right, I think Confluent has a fantastic position in the data landscape. If you ask, is there going to be more streaming in 3 years or less, there’s going to be a lot more and kind of our hands on the product side has gotten better. So yes, it’s frustrating when you have the dynamic with a subset of customers. But nonetheless, I do think the bigger picture I would say I’m as excited about where we’re at as ever. And I think there’s a lot of good things coming out of the business.

Jason Noah Ader: All right. One quick follow-up for Rohan. Rohan, can you help us on the NRR outlook. It dipped to the 124%, I guess that’s the consumption driving that. Do you think it will continue to dip just given some of your comments on the back half?

Rohan Sivaram: Yes. Jason, you rightly called it out, just for the broader group of folks here. When you think about our cloud business, NRR and GRR are essentially calculated based on the last 3-month consumption on an annualized basis. As a result, just — it has an outsized impact. The current quarter consumption typically has an outsized impact on both these metrics. So in line with our cloud second half outlook that we shared, we expect to see near-term pressure on both the metrics. Having said that, I’ll tell you that all the focus that Ryan is driving on the go-to-market side, be it the double-down initiatives, coupled with the green shoots mentioned earlier, there will be tailwinds. So there are a bunch of puts and takes to the NRR as we look at back half of the year and beyond.

Shane Xie: Thanks, Jason. We’ll take our next question from Gregg Moskowitz with Mizuho, followed by Truist Securities.

Gregg Steven Moskowitz: So Jay, I’m curious to hear your thoughts on your former company’s decision to move away from Kafka due to perceived scalability and operational challenges. Just given your history and your very deep knowledge of Kafka and the broader streaming space, your perspective on that would certainly be helpful.

Edward Kreps: Yes. This is kind of an internal system inside of LinkedIn, where I used to work 10 years ago. Yes, I think it’s kind of a nonissue. I mean, LinkedIn has very custom internal infrastructure. They actually made this change a long time ago. I think they only talked about it recently. Their internal thing is very tied to their infrastructure and not open source. So it doesn’t represent any kind of competitive threat to Confluent or anything like that. In terms of why they’ve done that, they’ve built their own custom database. They’ve built a lot of custom things at this point. They’re not even running in the cloud, even though they’re owned by Microsoft, so I think they have a bit of an in-house culture, which I think probably is a driver for some of this stuff.

Gregg Steven Moskowitz: Super helpful. And then just either for Jay or for Rohan. So when I look at Confluent Cloud, certainly, it was a shining star and has been a shining star of the Confluent growth story for quite some time. Obviously, the growth more recently for this segment is slowing down a fair amount. Any concerns that there may be competitive and/or pricing issues contributing to some of the incremental challenges that Confluent Cloud is experiencing right now? Or are you just not seeing that when it comes to win rates and when it comes to discounting?

Edward Kreps: Yes, it’s a great question. So yes, we haven’t seen a huge change in the competitive dynamic overall. If anything, I think our hand versus some of the cloud provider offerings has strengthened and that’s a focus area for us to go after those. On the pricing, we have introduced offerings that open up a set of workloads, the freight and enterprise clusters. We’re starting to see, I think, very strong early success with this. It always takes time for these new things to really go capture the opportunity. But I do think that’s a big opportunity for us to get out into that.

Shane Xie: Thanks. We will take our next question from Miller Jump with Truist Securities followed by TD Cowen. Miller, you are still on mute. All right, why don’t we come back to you, we’ll go to Derrick first.

James Derrick Wood: Great. I guess to start with you, Jay, could you give us a sense as to how much structural change we should be expecting from the field realignment efforts. And really, how long do you think that these realignments will take to implement and when we should be thinking about them driving some meaningful dividends?

Edward Kreps: Yes. I think we’re starting to see good forward momentum there already. There have been changes within the team. And I think those are ultimately positive things. Ultimately, when we bring in a new leader, we want them to change the things. It hasn’t been a complete reorg of everything. We’re broadly organized in the same way. But I do think coming into this year, we saw a few things that weren’t quite right and have made adjustments around them.

James Derrick Wood: And then kind of related, I mean — and maybe for Rohan, but — just looking at your net new customers, the $1 million net new customers have been really strong. The net new on the $100,000 has been pretty weak in the quarter and kind of directionally under pressure for the last year or two. But what should we take away from these numbers? Has there been a go-to-market shift more at the large deal front? Or are some of these realignment efforts designed to kind of drive better activity at that $100,000-plus? Just curious how you’re thinking about the direction of these numbers.

Edward Kreps: Yes. I can take some of that and Rohan can chime in as well. I do think kind of — as I said on the prior answer, I do think this is one of the points we want to address in terms of kind of the alignment between some of the SDRs, SEs, how they’re going after? What accounts they’re targeting. I do think we probably had more focus on some of the existing customers and so to really make sure we’re landing the right folks is critical. I think the CSP takeout opportunity is a key initiative there where we think there’s a good opportunity of things we haven’t paid as much attention to that are quite right. So we would like to see some results in that over the coming quarters. And Rohan, sorry, I took your — feel free to jump in there.

Rohan Sivaram: Yes. I’ll probably just add two points, Derrick. The first is one of the things that we focus on internally is from top of the funnel to $1 million, how that customer cohort is progressing. And some of the focused areas that Ryan is driving, that’s going to just fine tune that motion even more. So that’s one aspect of it. The second aspect of it on the larger customer front, our DSP penetration is a huge opportunity. And that’s something that we’re very focused on. What I can tell you is like with respect to our larger customers, we are having conversation and a substantial majority of them with respect to having some form of DSP continued usage. So in addition to what Jay said, those are two focus areas that we expect to drive even better performance, not only on the bottom of the funnel, but also on the top of the funnel.

Shane Xie: Thanks, Derrick. We’ll go back to Miller and try one more time.

William Miller Jump: All right. Can you hear me now?

Edward Kreps: Loud and clear.

William Miller Jump: All right. Maybe just starting with the large AI customer that you all mentioned in the prepared remarks. Moving back to Confluent Platform, I guess I’m curious like, is there an architectural advantage to supporting AI with platform? And do you anticipate realizing more of the AI opportunity on platform now after seeing this activity?

Edward Kreps: No, I don’t think it’s a structural thing. I would say it’s unique to the circumstances of this customer. Ultimately, the AI opportunity is kind of across both platform and cloud. The — I think it’s worth separating out, selling into AI companies. There, there’s obviously a set of start-ups that are very predominant in the cloud, and then you probably read some of these larger companies are making investments in on-premise data centers and more self-management and so on. So it’s more determined by the operations of the company than the industry or use case. And then when you think about the larger opportunity with AI, do you think it is the enterprise use cases deploying AI in their business. And that is again — it depends on where that is happening, the — much of it in the cloud, but there’s also large financial services organizations doing big interesting things in their own data centers. And so, yes, we’ll see it across both sides.

William Miller Jump: Okay. And maybe just — this is a follow-up on kind of Rohan’s commentary on the DSP traction that you’re seeing. Obviously, optimization headwinds are clouding this a little bit. But on the one hand, there’s really promising DSP uptake that you all are talking about. But at the same time, we continue to see cloud decelerate despite that contribution. So I’m wondering if you could unpack that a little bit more. Like are we seeing customers getting cost efficiencies on the streaming side that are then being applied towards the new capabilities? Or is it something where there are actually less streams being created and they’re just applying DSP to those?

Edward Kreps: Yes. I think this is ultimately an efficiency question, right? So we’ve been through this in the past. I would say it’s broadly the same pattern of things, right? There’s customers that are making kind of architectural changes to try and condense their workloads and squeeze more out of the infrastructure they’re using. That’s a natural thing you would do with almost any kind of cloud offering and that’s certainly what we’re seeing. The DSP growth, I think, is actually separate and super promising. What you need is for that number to get big enough that it outweighs effectively these very large Kafka installations, not exclusively, but certainly a number of them in some of these tech companies that are working very hard to cut cloud costs across the board.

And so you kind of have two forces in the business. I will say the optimization, you only optimized so much, right? So it kind of trickles off whereas the DSP growth, I think, is in the early days of a sustained run. So when I talk about just kind of overall a ton of optimism in this space, I think that’s one of the reasons why.

Shane Xie: Right. Thanks, Miller. We’ll take our next question from Mike Cikos with Needham followed by Barclays.

Michael Joseph Cikos: I know that you had spoken about the win rates versus CSPs and that being an area of doubling down for you. I guess one of the things, it seems like it’s a bit of a shift in message here given you guys have historically spoken about soaking up that Kafka opportunity. So to be going after the CSPs more directly, is that potentially a longer sales cycle because now you’re going into an organization that has been [ ingrained ] in some capacity with an existing vendor? And can you just kind of walk us through what are the mechanics been thus far that you’ve been able to evidence as far as those displacements you cited?

Edward Kreps: Yes, yes. It’s not — yes, first of all, it’s not exclusive. It’s not like we’re stepping back from the open-source Kafka in any way. I do think we probably paid less attention to these customers that had picked something from the cloud provider. It’s actually not a longer sales cycle. In some sense, it’s easier because there’s less of a team that does the self-management that has to be displaced and accounted for. You’re basically swapping in a better product for less money. And so yes, when we look at just kind of the overall sales stats against that. One of the things that motivated us here was what we felt we’ve done in the product portfolio, but part of it was just quantitative where we felt like, “Hey, these deals are winning at a high rate and progressing quickly.” And we know where it is. So it makes it easy to kind of package it up and teach the sales team exactly what to do in that competition.

Michael Joseph Cikos: Understood. And then just a quick follow-up for Rohan. I know, again, we’re talking about the changes to the cloud consumption here. If I go back a quarter ago, you guys were already assuming that we would not see a return to what normal behavior would be on that sawtooth pattern. And it sounds like, again, we’re adjusting our guide here. So did the consumption trends in Q2 actually deteriorate from where we were just in Q1. Can you provide some more granularity on that front?

Rohan Sivaram: Yes, Mike. From an overall consumption Q1 to Q2, I will put it in the category that the two dynamics we called out were fairly consistent that optimization of larger customers and slower use case adoption. So those two continued. If you look at month-over- month growth rates, they were flattish to slightly down from Q1 to Q2, and that’s obviously driving us as we think about the second half to make sure that our assumptions around month-over-month growth rates are notably below what we’ve seen for the same period of prior years. So that’s the dynamic on the cloud. However, I just want to remind around the green shoots that we called out in the call in different parts of our prepared remarks. First, I mean, Jay spoke about Flink, and the momentum in Flink, and we 3x this year, we are closing in on the $10 million run rate.

Second, we spoke about WarpStream. That business also showed really good growth in Q2 and the momentum is great. We spoke about the late-stage pipeline progression, and we also spoke about larger customers starting to commit more. So all of these are reissued. So at balance, obviously, the consumption is baked into it into our guidance for the back half but some of the green shoots are something that we’re excited to make sure we execute on and go and beat those numbers that we have.

Shane Xie: All right. Thanks, Mike. We’ll take our next question from Raimo Lenschow with Barclays followed by D.A Davidson.

Raimo Lenschow: Rohan, if you talk about the two drivers for the situation we have on the cloud side, like how much of that is driven by just one customer, the one that you kind of mentioned for Q4? So is that the majority of what’s going on? Or are we talking several accounts here?

Rohan Sivaram: No. What we said was, Raimo, like the larger customers, broad-based like for some of our larger customers, they are just optimizing and a slower use case adoption. If you look at guidance and what we spoke about the primary — the driver for guidance is our assumption for month-over-month growth rate. So that’s the primary driver. The dynamic that we saw in the first half of the year and in Q2, which is just the larger customers, optimizing for cost and slower use case adoption. That’s the driver that we spoke about. With respect to the AI native customer, that’s just us providing some color commentary as we look at the back half of the year and how we think about some of the puts and takes for the cloud business.

Raimo Lenschow: Yes. Okay. And then, Jay — one for Jay. Jay, if you think about the optimization and what’s going on there, is that people kind of doing more workloads in open-source Kafka? Or like — because like if you think about the overall volume seem to be going higher, so how do you optimize there? Can you just remind us there.

Edward Kreps: So yes, it’s not — we would characterize movements to open source. Yes, that’s effectively churn, right? It’s not optimization. Customers are moving off your product. No, what do we mean by optimization. You can try and take a lot of clusters that you have used across the company, combine them into bigger clusters and try and get some efficiency there. You can try and compress the data, you can try and optimize some of the usage patterns, so it’s more resource efficient. You can also do something more contractual where you kind of really model out your growth and commit to something bigger and return for a larger discount. Like all of those are the types of activities for customers that are looking to save. It was the case entering this year.

Last year had, had a very heavy focus on consumption. We were probably riding a little bit lower in terms of the forward commitments of customers to their future growth. So one of the things that we have done over this year is kind of take up that commit coverage to more of the forward growth. That’s basically a good thing, like you see it in the RPO growth, but it does mean somewhat higher discount, that’s kind of the normal discount schedule as you commit to more, you get a slightly better deal and so that would contribute to that as well, which is more of a contractual rather than technical optimization.

Shane Xie: Thanks, Raimo. We’ll take our final question today from Rudy Kessinger with D.A. Davidson.

Rudy Grayson Kessinger: Peter sort of asked it earlier, Rohan, close to 90% gross retention rate just to put the skepticism to rest, does that mean above or below 90%?

Rohan Sivaram: It was marginally below 90%, very marginally below 90%.

Rudy Grayson Kessinger: Got it. Okay. And then a lot of positive call-outs on the DSP products, in particular, Flink and some others. But with these — with what you’re kind of implying for the second half and kind of high-teens implied cloud revenue growth, what is core streaming cloud revenue growth growing at because you kind of back in and make some assumptions around Flink and some other products. And it would indicate it’s growing a good chunk lower than the overall cloud revenue. So any color you can share on just what is the core streaming, cloud revenue growing?

Edward Kreps: Yes, we haven’t broken it all out, but it is true the DSP portion of the business is outgrowing core streaming. I would look at that streaming growth rate as kind of a combination of additional use cases, growth and then optimization. And it is the case that certainly for the last few quarters, we’ve had more of that type of optimization, especially for some of the larger accounts. You obviously don’t see that on the DSP side where customers are building up new workloads. There’s not a big existing consumption base to go optimize.

Shane Xie: All right. Thanks, Rudy. This concludes our earnings call. Thanks again for joining us. Have a good evening, everyone.

Edward Kreps: Thanks, everyone.

Rohan Sivaram: Thank you.