Description

Thesis

Confluent (CFLT) offers a compelling combination of a secular product story around its new TAM-expansive Apache Flink product alongside a cyclical recovery in consumption spend aided by a change in sales comp model against the backdrop of heavily derisked guidance. With the stock ~70% off its ATHs / ~25% off its TTM highs and trading at a ~30% multiple discount to its peer group, coupled with relatively high short interest amounting to 10% of the float, there is significant room for the stock to rerate if we are right.

Introduction

CFLT is an infrastructure software company that provides managed data streaming and stream processing through its Apache Kafka and Flink offerings. Kafka is an open-source software project; its founders launched Confluent in 2012 and took it public in 2019. Data streaming allows data to flow in realtime between applications; Kafka offers realtime streaming which has significantly reduced latency over batching and micro-batching (less sophisticated ways for data to flow between apps in which data is sent in batches rather than continuously streamed). Use cases include FedEx and Uber using Confluent Kafka data streaming for realtime driver location updates and GenAI-powered chatbots using Confluent Flink stream processing to analyze and respond to queries with minimal latency.

Confluent offers managed Kafka in two environments – on-prem (48% of the business; growing ~45%) and cloud (52% of the business; growing ~20%).

New Apache Flink product will drive significant upside to Confluent Cloud #s this year, contrary to expectations that it only becomes material in ’25.

• On March 19, Confluent’s new Apache Flink product went GA. Flink is an open-source protocol for stream processing, i.e. running real-time computations and analytics on top of data streams between apps.
• Management estimates that Flink’s TAM is even larger than Kafka’s (which they’ve placed at $60b).
• Management has conservatively spoken to Flink only becoming material exiting FY24 and has not baked in a material Flink contribution into the FY24 guide. Sellside analysts similarly assume a de minimis revenue contribution from Flink in FY24.
• However, based on our detailed checks (several partner/customer interviews and a large N survey), Flink is expected to drive an uplift in Confluent Cloud spend anywhere from 15% to 35% in FY24, scaling further in FY25.
• Furthermore, we expect Flink to see a material near-term tailwind from GenAI-powered chatbot use cases.
• AI chatbots are a big use case for data streaming and stream processing due to reduced latency relative to batching or micro-batching. Running analysis on chatbot queries using Flink while data is being streamed offloads compute from other apps and allows customers to receive a quicker response.
• Adoption of AI chatbots to replace human customer service agents has been rapidly accelerating and will drive greater data streaming and stream processing volumes, directly flowing into revenue in Confluent’s consumption model. For example, Klarna announced on Feb 27 in a well-publicized press release that 2/3 of its customer service chats had moved from human reps to GenAI-powered chatbots.
• On their last earnings call, Confluent announced that they had landed OpenAI as a customer and launched a partnership with Anthropic.
• At the MS and JMP TMT conferences in March, Confluent’s CEO said that GenAI was a key realtime use case and specifically cited GenAI chatbots.

Cyclical recovery in consumption spend against highly derisked guide aided by new sales comp model

• On the 3Q23 call, management guided FY24 growth to 22%, meaningfully below consensus expectations. The weak guidance was attributed to a handful of customers moving their entire AWS deployments from a cloud environment back to on-prem, Confluent’s largest-ever customer (NEWR) being acquired by PE and ramping consumption slower than expected, and a change in their sales compensation model. We believe that the new CFO was also looking to sandbag the guide and reset numbers on his first quarter so as to set up a steady beat-and-raise cadence going forward.
• In 4Q24, Confluent executed well and delivered a beat-and-raise, with strong execution demonstrating that the issues driving the guide’s weakness the previous quarter had been one-offs. While they raised the effective $ amount embedded in the FY24 guide, the 22% growth rate was maintained. This assumes a -1100 bps deceleration from the 33% growth they put up in FY23, which appears very conservative as the 22% guide does not bake in a continued recovery in consumption revenue or material revenue contribution from Flink.
• Commentary on the most recent earnings call was positive on continued green shoots in consumption recovery. The CEO struck a similarly bullish tone at the MS and JMP TMT conferences.
• With the FY guide derisked, there is upside from a continued recovery in consumption spend post-optimizations, which IT budget surveys and partner/customer checks continue to support.
• Furthermore, at the start of this year, Confluent also moved to a new sales comp model in which sales reps are fully paid on consumption revenue rather than contract bookings. This directly aligns reps’ incentives with the interests of the company (and customers).
• Previously, due to 85% of sales comp coming from bookings, sales reps were incentivized to persuade customers to sign excessively large contracts well below the consumption run-rate that they could plausibly achieve. This resulted in friction between salespeople and customers, with customers we spoke to indicating they felt cajoled to sign big deals. As Confluent only recognizes cloud revenue upon customers consuming and has zero rev rec from bookings, this presented a rev headwind which is now reversing and becoming a y/y tailwind.
• While bears fear increased NT sales attrition from the comp change, management has indicated that they already derisked #s and baked this in when the new CFO sandbagged the guide two quarters ago. Furthermore, on the most recent EPS call, management indicated that they had not seen a meaningful pick up in sales attrition. Based on our interviews with former Confluent employees and LinkedIn profile checks, we believe that this is because management began communicating this change internally more than a year ago – i.e. the attrition has already been seen and the second derivative there is improving.

Confluent has been taking share from open-source Kafka and the hyperscalers’ managed Kafka offerings due to improvements in incremental feature/functionality and total cost of ownership (TCO) savings

• Confluent primarily lands new customers from selling into the existing open-source Kafka installed base.
• Confluent’s penetration of the open-source installed base remains very low (~3%), with significant runway ahead. There are more than 150k organizations using Kafka, including the majority of the Global 1000, but only ~5,000 Confluent customers.
• Open-source Kafka customers typically choose to pay for Confluent’s managed Kafka’s offering as they can save ~60% on the total cost of ownership (TCO) while enjoying 10x better performance. Running an open-source Kafka deployment requires hiring dedicated Kafka engineers; it is far cheaper to pay Confluent for managed Kafka and save on headcount spend. The payback period on moving from the open-source to Confluent managed Kafka is typically 6 months or less, so CIOs see a rapid ROI.
• While the hyperscalers offer their own managed Kafka offerings built on top of the open-source (e.g. AWS MSK), they are far less performant than Confluent’s best-in-class managed Kafka offering. Our checks and recent management commentary have indicated that Confluent’s winrates against the hyperscalers’ managed Kafka offerings have been improving and that Confluent has been taking share.

Multiple remains at a 3-turn discount to infrastructure software peers with consumption revenue models. With the stock ~70% off its ATHs and ~25% off its TTM highs, coupled with relatively high SI of 10%, there is significant room for re-rating if we are right.

• CFLT trades at ~7.5x CY25 revs, a 3-turn discount to the ~10.5x average multiple of the consumption model infrastructure software peer group (DDOG, ESTC, MDB, SNOW).
• There is significant room for multiple expansion as Confluent executes on topline reacceleration, with upside to buyside expectations likely from Flink contribution.
• There remains an under-appreciated margin story that is likely to further drive multiple expansion as Confluent scales margin from current breakeven levels to their long-term guide of 25%+. Over the past eight quarters, management has delivered on their objective of getting to breakeven margins exiting ’24, having improved margins from -40% to flat.
• SI is relatively high at 10%, making a squeeze possible if we are right (i.e. what happened in 4Q23 when the stock closed +35% T+1 post-earnings). With the stock still ~70% off its ATHs and ~25% off its TTM ATHs, coupled with the ~30% discount to its peer group’s average multiple, we think there is significant room for a re-rating if our thesis plays out.

Catalyst

Earnings