Roblox (RBLX.US) FY26 Q2 earnings call: Business volatility rises, Q3 guidance turns negative.

Roblox (RBLX.US) held its FY26Q2 earnings call. The company's revenue for the second quarter was $1.5 billion, a year-on-year increase of 36%; bookings reached $1.6 billion, an 8% year-on-year increase, falling at the lower end of the guidance range and below the company's internal targets. Daily Active Users (DAU) were 123 million, up 10% year-on-year; usage time totaled 29 billion hours, an increase of 5%. Operating cash flow was $318 million, a 60% year-on-year increase; free cash flow was $294 million, up 66% year-on-year. The main reason for the underperformance this quarter was that monetization (measured in bookings per hour) was below expectations, particularly for users under 13; user registrations, retention, and usage time were all in line with or better than expectations. For Q3, bookings are expected to be between $1.58 billion and $1.65 billion, representing a year-on-year decline of 14% to 18% against last years high base. In terms of profitability, the decrease in bookings will lead to a deleveraging of fixed costs; at the same time, new AI-driven initiatives such as Build, Roblox Reality, and Moments will increase infrastructure costs. According to Q3 profitability guidance, approximately half of the year-on-year compression in profit margins is attributed to the fixed cost deleveraging due to the decline in bookings, while the other half comes from the aforementioned AI-related investments (primarily incremental cloud GPU computing expenses). The companys view on the long-term profitability potential of the business remains unchanged, and the factors driving margin expansion are still present growth in bookings will bring fixed cost leverage, and the shift of the business mix towards lower-cost platforms will improve COGS. Notably, the nature of this round of infrastructure investments differs from the past: previously, infrastructure spending largely correlated with user numbers on the CKH HOLDINGS platform, while this round is primarily driven by model training, making it more like fixed costs, so the leverage effect will gradually become apparent with the growth in bookings. Additionally, the initial plan for infrastructure costs related to Morpheus is to be offset by subscription revenues. The company maintains its confidence in achieving long-term revenue growth rates exceeding 20% and its goal of capturing 10% of the global gaming market (with a significantly higher share in the U.S. market). Q&A Q: In the process of algorithm adjustments, Roblox Kids, and select accounts implementation, what metrics should investors monitor to determine when the headwinds from these changes will peak? A: There are several things worth watching. For users under 18, we clarified our direction in this earnings call, which includes our initial vision that we believe will eventually be realized You come to make games. Not every user can figure out how to make a game using Studio, and we've already seen in New Zealand that the daily users of Build outnumber those of Studio. We do believe that ultimately, AI creation of games will become a behavior on par with playing games. So first, please monitor the usage frequency of the under 18 group. Secondly, we expect that the use of AI in the Build process or tokens will be relatively lenient, but for those users who are using Build all day, we will introduce enhanced limits in the Roblox Plus subscription this limit might be referred to as tokens elsewhere, and we will have our own naming convention. So, please also pay attention to the subscription performance of the under 18 group. Thirdly, as communication platforms continue to roll out, please watch this line. For users over 18, we are highly focused on several specific demographics and their growth rates, while also paying close attention to the types of content produced by the creator ecosystem that we believe show stronger retention among these groups. So please monitor the growth rate for users over 18. I have already shared some current data in my presentation that points to that 80% of the market we believe we can replicate what weve accomplished with users under 18 within that group. Q: Is the expectation for sequential growth in user numbers in Q3 merely a seasonal factor, or have we already seen an improvement in trends by the end of Q2? A: It needs to be clarified that what you mentioned as "short-term participation friction" is not what we are seeing. We are quite satisfied with the participation trends we are observing; registrations, retention, and other aspects are quite healthy, and participation looks good. The weakness that does not align with our initial expectations when we set our quarterly plans is actually concentrated in monetization. This trend will also continue as we continue to adjust the discovery and recommendation mechanisms. Regarding the DAU trend itself, seasonality is undoubtedly a significant factor. Additionally, we will gain a full quarter's contribution from the re-launch in Russia. These two points are what I want to emphasize about the DAU trend. Q: With the updates to the discovery engine, when can we expect to see long-term retention benefits reflected in improvements in engagement and monetization, and what is the timeline and expectation? A: What any discovery algorithm is really meant to do is connect users with experiences, driving long-term business value while also signaling to creators what types of experiences will be rewarded by the discovery mechanism. This is a complex process involving feedback from both creators and users. Over the past few months as we've been implementing and improving the algorithm, community feedback has been overwhelmingly supportive. The sentiment is that we are shifting increasingly towards evergreen games games that can retain players over the long term, which are basically the opposite of so-called "quick cash grab" games (we're not saying we previously had this issue). We want to stand firmly on the evergreen side as much as possible. As we advance to the next generation of algorithms, we made two decisions: first, to extend the time window for signal measurement; second, to update the algorithm with this feedback loop, even as we are still running further tests. We have transformed the discovery system from a "fixed" entity into a system that is, in some sense, self-improving and self-calibrating, continuously estimating what kind of "user-game" pairings can lead to the best long-term retention. We are achieving a finer balance between optimal combinations of long-term retention and long-term monetization. In our experiments, we see a short-term decline in monetization; however, there tends to be a crossover point where improvements will emerge over the longer term. We will not disclose exactly where this crossover point is, but this is exactly why we decided to move forward with this it is a direct measure of retention, rather than a projected forecast; it directly measures both retention and monetization. Additionally, with the launch of Build and a significant increase in content creation on Roblox, we believe this algorithm is highly resilient: it can discover high-quality Build games that everyone can play while not drowning users in those "games that look like they are AI-generated" and have poor long-term retention. Q: As the company transitions to focus on users aged 18 and above and new content, what changes will occur in the cost structure? If everything goes smoothly over the next 12 to 24 months, what will the financial model look like compared to the past 12 to 18 months? A: Two points need to be emphasized. First, our view on the long-term profitability potential of the business indeed remains unchanged, meaning that the factors driving margin expansion in the future are still in place. Obviously, continued growth in bookings will generate fixed cost leverage; we still expect that the business mix will shift towards lower-cost platforms, improving COGS. Further enhancements will ultimately depend on achieving fixed cost leverage in other areas of the business. Second, in the near to mid-term, our significant investments in many AI-driven product enhancements will indeed keep infrastructure costs at a high level. But its important to recognize that this spending differs in nature from our historical expenditures; it is more like a fixed cost because past infrastructure investments were a function of the user numbers on the CKH HOLDINGS platform, while these infrastructure costs are primarily driven by model training. We are pushing aggressively in this direction. However, over time, due to its more fixed cost nature and growth in bookings, we expect to begin seeing leverage effects on these infrastructure investments. However, as shown in the Q3 guidance, there will be some pressure in the short to mid-term due to the need for incremental cloud GPU computing power. Q: How do you see the competitive landscape evolving not only within gaming but also in terms of time competition for younger users between "online vs offline," "games vs non-games," as well as how the company positions itself in the broader competition for time and attention? A: I do believe that a broader definition is something larger than Roblox, and our vision is that the world needs more "play." Play is about interpersonal interaction, not only just playing itself but also includes creating, being with others, communicating with others, being oneself, and observing others. Just as we redesigned the app and brought Build and Moments to the homepage, this is part of that vision to reflect the physical world using our digital world that supports "play." We believe, just as "play" in the real world sometimes includes creating the games themselves, I am optimistic that we will see the volume of creation brought by Build vastly exceed any levels during the Roblox Studio era; at the same time, those professional developers and teams creating extremely complex works will fully utilize Build's capabilities as they transition to large team collaboration in Studio to develop more complex content. So I truly believe that "what constitutes a game" is changing. I think we are entering an era similar to that of a decade ago when most of us did not edit videos, but now almost all young people and many others are editing videos on their phones. I think a similar transformation will occur in the gaming field: accelerated by AI, from 2D puzzle games to complex 3D multiplayer games, all are going to be created on mobile and enhanced in Roblox Studio, fundamentally changing the gaming landscape. Therefore, I do believe that todays youth are changing; I believe that "play" is universal. I also believe we are moving towards that initial vision where everyone on the platform is a creator and builder. Q: As free cash flow increases, how will capital allocation be prioritized between buybacks, internal investments, and acquisitions? Will stock performance affect the willingness to increase buybacks? A: Our capital allocation strategy is very clear: the priority is to ensure that we have the capacity to aggressively invest in internal growth, reflected primarily in expenditures such as R&D, personnel, tokens, and infrastructure. We strive to be diligent and efficient in all these investments but also recognize that the world is changing rapidly right now; we will keep sufficient "gunpowder" to ensure we are highly agile in internal investments. That being said, based on our balance sheet and trends in free cash flow, we have plenty of firepower for both the aforementioned internal investments, as well as supporting buybacks and making acquisitions if necessary. Our historical acquisitions have primarily focused on acquiring technical talent, and we still have ample capacity to continue doing that; in the future, we will not rule out larger transactions if theres a strategic or industry rationale. Regarding how stock performance influences our considerations about buybacks, there are two points. First, the structure of our buyback program is mainly designed to offset dilution from employee stock incentives, which results in us actually buying more shares when the stock price drops and fewer shares when the stock price rises so there is a sort of natural adjustment built into the mechanism. Second, we do indeed have the capability to accelerate buybacks or add to buybacks, which are obviously matters we will continue to assess. Q: The company no longer provides full-year guidance; can you give some boundary references for Q4 and the full year? Will 2026 still be a growth year, and will full-year bookings exceed those of 2025? And how will the factors contributing to the weak Q3 guidance improve in Q4? A: As I mentioned earlier, we will not be providing Q4 guidance now, quite simply because there are too many variable factors mentioned in our shareholder letter and earnings call commentary. We feel good about long-term trends. We are simultaneously addressing the normalization following last years high base and a series of platform changes in discovery, safety, and so on, and launching significant new products like Build and Moments. The reality is that many of these new initiatives are still in very early stages. We do not have a single model that can tell us what exact results they will yield in the next three to six months. We want to ensure we are pushing as quickly as possible on all these initiatives. So you can assume that what we are doing in Q3 and Q4 will align with the long-term goals weve spoken about, but other than what weve already shared, we cannot provide more specific guidance. Q: The shareholder letter mentioned that based on internal tests, the company is confident that longer retention will ultimately compensate for the decline in hourly monetization. Can you elaborate on what has been observed in the tests? A: To elaborate on this: we have been running a lot of experiments and continuously iterating on these algorithms. When we AB tested the new algorithm against users who were still using the old discovery algorithm, we observed that user retention changes quite rapidly, which creates a compounding effect for our business, as it ultimately drives more users to power Roblox's flywheel; this will also be reflected relatively quickly in incremental participation time. However, the hourly revenue (dollars per hour, the monetization metric we focus on) will be immediately impacted quite noticeably. Based on the curves we see from several weeks of data formed in each experiment, we indeed believe these two lines will intersect, meaning that the benefits from increased retention will exceed the short-term shock to bookings. We will not provide a specific timeline at this time, partly because we are still continuously iterating on these algorithms. Our current focus is actually on finding ways to improve the algorithm that would retain retention benefits while reducing the recent impact on bookings. There are some promising projects being pushed in this area, but it is too early to say how significant the impact might be. To add one point: the longer we observe the signals and the more directly we measure them, the more than just early revenue we capture; it is also the long-term revenue associated with retention. We are trying to optimize the direct measurement combination of long-term retention and long-term monetization, rather than overly extrapolating those potentially noisy and quickly decaying short-term monetization signals. Q: Can you specify the financial scale of investment in Morpheus or help quantify the potential related cost headwinds we might see this year? A: Morpheus is just a small part of our many AI initiatives, so I will answer from a larger perspective. The simplest way to understand this is to look at our Q3 profit margin guidance, and the extent of margin compression compared to the same period last year I would say about half of it comes from fixed cost deleveraging due to the decline in bookings, while the other half is related to these AI investments. These investments support new features such as Build, Moments, a significant amount of improvements in safety, and model training for things like Roblox Reality. So these are very important and exciting things, but they do require incremental investment. I also want to emphasize the significance of what Morpheus is pointing towards: our goal is to provide photo-realistic multiplayer gaming, which does not exist in the world today. At one end of the spectrum, we are starting to have offline video models approaching 4K movie quality; at the other end, 3D gaming technology is improving continuously, but it hasn't reached photo-realistic standards. We believe the mixed-pathway approach is the ultimate solution cloud doing 3D synchronization, coupled with high sampling upscaling on the client side to achieve photo-realism. The infrastructure costs here will initially be offset by subscription revenue. Games will run in both the regular Roblox mode and the supersampling Morpheus video realism mode, and we initially expect to charge a subscription fee for access to the latter. Q: Will the 2D content announced this time become a customer acquisition funnel for players outside the platform, or is it addressing existing demand within the platform? How does 2D content on Roblox differentiate from typical 2D mobile game experiences? A: Returning to the original vision of "you come to make games," we believe that what Build is doing aligns well with our statement of "removing the boundaries of content types on the platform." When we put Build in front of users, they dont limit themselves to saying, "I only want to make a 3D multiplayer obstacle course," or "I only want to make a puzzle game like this." So, there is a very good intersection in supporting people to build anything they want in Build. Another point is that I believe many people do not consciously differentiate between 2D single-player games, 2D multiplayer games, 2.5D isometric games, or 3D multiplayer games they just see them all as "play." And as we tap into older demographics, some of these groups have a strong demand for 2D. But we believe that any experience built on Roblox has tremendous advantages: it supports multilingual operation, runs globally, has social capabilities, is supported by our economic system, backed by our infrastructure, and can connect with each other. So our real understanding is that this is widening the range of game types on the platform. Ultimately, this also resonates with Build when people imagine creating a game and showcasing it to friends, we wont limit what they can do. Q: What are the differences in monetization models between users under 13 and other demographics that resulted in this quarter's particular impact? Is it specifically related to last year's viral blockbusters, or is the monetization of younger users inherently more volatile? A: It leans more towards the former. If you look at the hours contributed by those enormous, highly monetized viral blockbusters on the platform last year, you will find that there was very high concentration, and these games especially attract younger users. So when we see this transition in the composition those games being replaced by games with more "normal" levels of monetization the impact is most pronounced among younger users. I do not believe there has been any structural change in the behavior of younger users themselves; fundamentally, it's just that the currently popular games do not possess the same extremely high monetization levels as last year. Q: The Incubator and the initiatives related to users over 18 seem to be progressing well. What have we learned in this process, how will this project evolve in the future, and what other initiatives are underway? A: We have a large amount of content not only from the incubator but also from studio partners and the existing community. We ultimately see ourselves as a UGC platform and value the power of this, but we also see another capability in some sense like sales engineering or its extension which is establishing very close relationships with top creators on the platform, maintaining deep collaboration to guide them in understanding what content performs well on our platform. One of our focuses is on multiple technologies recently introduced in collaboration with incubator partners. Weve introduced high-performance avatars that significantly enhance performance, as well as worlds that run smoothly on low-end Android and high-end PCs. We have also introduced a synthesis technology called Slim, which allows very complex avatars to perform well on mobile. All of this collaborative work we are advancing with partners is believed to be a huge technical opportunity for them to build games for us. Of course, they are also participating in our 50% premium mechanism for DevEx for content aimed at users over 18. So the reality is that this is not just the incubator; it includes many corporate studios we are working with and the existing developer community, where we see the quality of their output is very high.