RBLX (Trans): Q3 guide turns negative; higher biz volatility; no FY guide

Below is Dolphin Research's compiled transcript of $Roblox(RBLX.US) FY26 Q2 earnings call

I. Key takeaways

1. Shareholder returns: Following last quarter's print, the company announced a buyback. The program is designed primarily to offset dilution from employee equity awards, with an automatic counter-cyclical feature — more shares repurchased when the stock falls, fewer when it rises.Management retains flexibility to accelerate or upsize the program and will evaluate on an ongoing basis.

2. Q3 guidance: Q3 bookings are guided to $1.58–1.65bn, implying -14% to -18% YoY on a tough comp. On margins, lower bookings drive fixed-cost deleverage, while AI-driven initiatives such as Build, Roblox Reality, and Moments lift infra costs. On management's Q3 margin framework, roughly half of the YoY margin compression stems from fixed-cost deleverage on lower bookings, and the other half from AI-related spend (primarily incremental cloud GPU).

3. Q4 and full-year outlook: With higher volatility and ongoing platform iteration, the potential outcome range for Q4 is too wide. The company is not issuing updated full-year guidance and is not providing Q4 specifics. Management said expectations for the remainder of the year have shifted meaningfully, but believes the current trade-offs are right.

4. Q2 key metrics: Revenue was $1.5bn (+36% YoY). Bookings were $1.6bn (+8% YoY), at the low end of guidance and below internal targets.DAUs were 123mn (+10% YoY). Hours engaged were 29bn (+5% YoY).Operating cash flow was $318mn (+60% YoY); FCF was $294mn (+66% YoY). The primary shortfall vs. expectations was monetization (bookings per hour), especially among users under 13, while sign-ups, retention, and time spent were in line or better.

5. Long-term financial framework unchanged: Management's view on long-term margin potential is intact, driven by the same factors — bookings growth creating fixed-cost leverage and mix shift to lower-cost platforms improving COGS. Note this infra cycle differs from history: previous infra spend scaled with hours and users, whereas this round is driven by model training, more akin to fixed costs, and should show leverage as bookings grow.In addition, early-stage Morpheus infra costs are intended to be offset by subscription revenue. The company maintains confidence in 20%+ long-term revenue growth and a 10% share of the global gaming market (materially higher in the U.S.).

II. Call details

2.1 Management remarks

1. Safety and Age Check

a. In H1, the company decided to age-verify all users and limit chat to verified users. Verification penetration reached 57% globally this quarter.Early-launch markets are higher: Australia near 80%, the U.S. and U.K. at 70%. Verification among users under 18 is particularly strong, with the U.S. under-18 penetration at 75%.

b. Beyond filters and a leading safety stack, the platform disallows image sharing in chat and introduced age-banded communication aligned with verification. In Jun., Roblox Kids and select accounts launched successfully based on age.Management believes accurate age data strengthens confidence in the platform's long-term growth and will power more precise content recommendations by region and age.

2. Discovery pivot to long-term retention

a. The company has made clear to the community that discovery is shifting to target observed long-term retention rather than short-term proxy signals. This has impacted monetization, concentrated in U.S. under-13 users, as titles that monetize quickly at the expense of long-term retention receive less exposure.Data already shows higher-quality, stickier titles are being lifted. Management expects the benefits from higher retention to outweigh near-term headwinds, as better retention and engagement compound over time.

b. Two key changes underpin the algorithm: lengthening the measurement window; and using a live feedback loop to update the model in real time while tests run. Discovery has moved from a static system to one of continuous self-improvement and self-calibration.The algorithm is resilient to the surge in Build-created experiences — it can surface high-quality Build games while avoiding flooding users with titles that look 'AI-generated' but lack long-term retention.

3. Content and user mix

a. Content diversity is rising while new-launch velocity remains steady. Three years ago, the top-10 titles accounted for ~30% of time spent; today that has fallen to ~20%.Intl growth is strong: Japan DAUs +67% YoY, India +64% YoY; hours +61% and +59%, respectively. Russia reopened at quarter-end with users returning.

b. A secondary factor in weaker monetization was the decision to disable cross-experience game pass sales.

4. Expanding 18+ (three initiatives)

a. The 18+ segment represents ~80% of the $200bn global gaming market and is central to replicating under-18 success. In the U.S., 18+ users are nearly one-third of age-verified DAUs, with U.S. 18–34 DAUs and hours up ~40% YoY, and 13+ globally growing strongly.Management estimates ~20% of U.S. under-18 gaming spend flows through Roblox today. As product-market fit improves, paid acquisition will complement viral growth for 18+.

b. Most product and tech investments targeting 18+ — beyond safety — also benefit younger users. - Initiative 1: Remove self-imposed genre limits. The company believes it is close to supporting most genres people play online, from near-physically-real multiplayer to previously shelved high-performance 2D tools and gameplay.- Initiative 2: Redefine the app to mirror 'play' in the physical world, including discovery. - Initiative 3: Return to the core promise that anyone can be a creator.

5. Roblox Reality / Morpheus (photoreal multiplayer)

a. The goal is to deliver photoreal multiplayer experiences that do not exist today. Offline video models are approaching 4K film quality, while 3D game tech keeps improving but is not yet photoreal.The path is hybrid: cloud-based 3D synchronization plus client-side super upsampling to achieve photorealism. Management believes for some genres the line between games and films will blur, and this will arrive in real time sooner than expected.

b. Business model: titles can run in standard Roblox mode or Morpheus' super-upsampled photoreal mode. The latter is initially planned as a subscription to offset infra costs.

6. 2D content support

a. Games are also spontaneous, single-player, and 2D. The company had prioritized 3D and shelved high-performance 2D tools and gameplay, but will now expand support across 2D puzzle to strategy.Together with Roblox Reality and 2D optimizations, the platform is widening both ends of the content spectrum. Many users do not consciously separate 2D single, 2D multi, 2.5D orthographic, or 3D multiplayer — they simply 'play', with older cohorts showing ample 2D demand.

b. The differentiation: any Roblox-built experience is natively multilingual, global, social, backed by platform economics and infra, and interoperable across experiences.

7. Moments (video discovery and creation)

a. The company announced a unified app vision to surface video discovery and creation to all users. Moments is integrated into the 16+ homepage and is in testing for 16+ users in Canada, New Zealand, and Singapore.The feed is expanding from gameplay clips to short-form creator content and will ultimately open to everyone, with appropriate, safe educational content for younger players.

b. Next, cards will let users watch creators or see friends' live gameplay, and eventually bring full 3D games into the Moments tab, connecting to Build-created titles. Users can jump in with one tap, creating a unique path from discovery to next-gen content.

c. Early tests show significant retention uplift from Moments, and management is very optimistic.

8. Comms

a. The company aims to deliver best-in-class integrated voice, video, and text on a safety-first stack. Global chat and QuickWords are live.A meaningful share of users still use off-platform tools. The roadmap includes video chat with friends or live streaming as a Roblox avatar, providing a safer, richer, more immersive experience that follows users across games.

b. In Q3, the first step is 'in-experience friend chat', bringing real-time text messaging directly into any Roblox title.

9. Build (AI creation)

a. Build extends the vision that 'everyone can be a game creator'. Powered by in-house gen-AI models and a conversational interface, anyone can describe an idea, co-create with others, and generate a playable experience in minutes.For existing creators, Build is a practical lever — faster construction and better audience reach. Build fully integrates with Roblox Studio: draft in Build, finalize in Studio.

b. Build is live in New Zealand with heavy inflows of creations, and in New Zealand daily Build users already exceed Studio users.The key is not just the model, but the harness, tools, and skills; the vision is multiple models unified under one harness, closed-loop with first-party cloud back end and infra. Build also aligns with removing content boundaries, supporting 2D, 3D, and everything in between.

c. Monetization will be relatively permissive on AI usage/tokens within Build, while heavy all-day users can extend limits via Roblox Plus subscriptions.

10. Creator ecosystem and tech support

a. Content sources include incubator partners, studios, and the existing community. The company is engaging top creators in a sales-engineering-like motion, collaborating closely to guide what works best on-platform.New tech opened to incubator partners includes higher-performance avatars and worlds running smoothly on low-end Android and high-end PC, plus a synthesis tech called Slim to run highly complex avatars on phones at high performance.

b. For 18+ consumer titles, DevEx revenue share includes a 50% premium.

2.2 Q&A

Q: With discovery changes and the rollout of Roblox Kids and select accounts, what KPIs should investors watch to gauge when headwinds peak?

A: Several things. For under-18s, we've outlined the direction, including our belief in the original vision — 'you make the games'. Not everyone can figure out Studio, and in New Zealand we already see daily Build users surpassing Studio users.We do believe AI game creation will become a mainstream behavior alongside play. First, track usage frequency among under-18s.

Second, we expect to be lenient on AI usage or tokens within Build, but for all-day creators we will add higher limits within Roblox Plus — that quota might be called tokens elsewhere, but we'll use our own naming. So also watch subscriptions among under-18s.Third, follow the communications roadmap as we roll it out. For 18+, we are focused on specific cohorts and their growth and on creator output tailored to those cohorts with stronger retention. So please track 18+ growth rates.

Q: Is the expected sequential DAU growth in Q3 merely seasonal, or did trends improve exiting Q2?

A: To clarify, the 'short-term engagement friction' framing doesn't reflect what we see. We are quite pleased with engagement trends — sign-ups, retention, and related metrics look healthy, and engagement looks solid.Where we fell short vs. the original quarterly plan was monetization. As we continue to adjust discovery and recommendations, that will persist.On DAUs, seasonality is a major factor. We will also see a full-quarter contribution from Russia's reopening, and those are the two points I would emphasize for DAU trends.

Q: When should we expect discovery's focus on long-term retention to translate into better engagement and monetization?

A: Discovery's job is to connect users with experiences, drive long-term enterprise value, and signal to creators which experiences are rewarded. It's a complex loop involving feedback to both creators and users.In recent months, as we implemented and refined the algorithm, community feedback has been overwhelmingly supportive. The sentiment is we are favoring evergreen titles — those that keep players long term — the opposite of quick-cash designs (not implying we had that issue). We aim to be as far on the evergreen side as possible.

We made two choices for the next-gen algorithm: lengthen the measurement window and update the model via the feedback loop even as more tests run. We've turned discovery from a 'fixed' system into one that continuously self-improves and self-calibrates, estimating which user–experience pairs maximize long-term retention.We are increasingly optimizing the trade-off between long-term retention and long-term monetization. In experiments, we see a near-term dip in monetization followed by a crossover and longer-term improvement.

We won't disclose the exact crossover point, but that's why we proceeded — it's a direct measure of retention, not a projection, and a more direct measure for both retention and monetization. With Build accelerating creation on Roblox, we believe the algorithm is resilient — surfacing high-quality Build titles for everyone while avoiding flooding users with 'AI-sounding' games that lack long-term retention.

Q: As you pivot to 18+ and new content, how does the cost structure evolve, and what could the model look like over the next 12–24 months vs. the past 12–18 months?

A: Two points. First, our view on long-term margin potential is unchanged — the drivers remain in place. Sustained bookings growth yields fixed-cost leverage, and we still expect mix shift to lower-cost platforms to improve COGS.Further gains ultimately depend on realizing fixed-cost leverage across other parts of the business. Second, in the near to medium term, AI-driven product enhancements will keep infra costs elevated.

Importantly, this differs from historical spend — it's more like fixed cost, as prior infra spend was a function of hours and users, while this is driven by model training. We are leaning in aggressively.Over time, given the quasi-fixed nature and bookings growth, we expect leverage on these infra investments, though as Q3 guidance indicates, incremental cloud GPU will pressure margins near to medium term.

Q: How do you view competition — not just within games, but for time across online vs. offline and game vs. non-game, especially for younger users, and where does Roblox position itself?

A: The broader frame extends beyond Roblox, and our vision is the world needs more 'play'. Play is social — creating, being with others, communicating, being yourself, and watching others.Our redesigned app bringing Build and Moments to the home screen reflects that vision — mirroring the physical world's play with our digital platform. As in the real world, play sometimes includes creating the game itself.

I'm optimistic Build will drive creation volumes far beyond any period in the Studio era. Professional teams delivering complex works will also leverage Build fully as they collaborate in Studio to produce more sophisticated content.What constitutes a 'game' is changing. A decade ago most of us didn't edit video; now nearly all young people do on their phones. With AI acceleration, from 2D puzzles to complex 3D multiplayer, creation will happen on mobile and be enhanced in Studio, reshaping gaming.

I believe young people are changing, and play is universal. We are moving toward the original vision where everyone on the platform is a creator and builder.

Q: With rising FCF, how will you prioritize buybacks, organic investment, and M&A? Does share performance affect your willingness to accelerate repurchases?

A: Our capital allocation is clear: first, ensure we can invest aggressively in organic growth — primarily R&D, headcount, tokens, and infra. We aim to be disciplined and efficient across these, while keeping ample dry powder given the pace of change to stay agile.That said, our balance sheet and FCF trajectory provide capacity to fund both organic investments and buybacks and to pursue M&A as needed. Historically, M&A focused on acqui-hiring, and we retain that capability; larger strategic deals are not ruled out if warranted.

As for share performance and buybacks, two points. First, the structure is mainly to offset equity comp dilution, leading us to buy more when shares decline and less when they rise — an inherent stabilizer.Second, we do have the ability to accelerate or add capacity, which we will continuously assess.

Q: With no full-year guide, can you frame Q4 and FY boundaries? Is 2026 still a growth year with bookings above 2025, and how do the Q3 headwinds ease in Q4?

A: As noted, we are not guiding Q4 given the number of moving pieces discussed in the shareholder letter and on this call. We feel good about the long-term trajectory.We are digesting normalization off last year's tough comp while rolling out platform changes in discovery and safety and launching significant new products like Build and Moments. Many of these are very early.

We don't have a model that pinpoints outcomes over the next three to six months, and we want to move as fast as possible on these initiatives. You can assume our actions in Q3 and Q4 align with our long-term goals, but beyond what we've shared, we won't be more specific.

Q: Your letter says internal tests give confidence that longer retention offsets the drop in dollars per hour. What are you seeing?

A: To unpack this, we run extensive experiments and iterate continuously. In A/B tests comparing users on the new discovery algorithm vs. the old, we see retention move quickly, which compounds for our business as more users drive the flywheel and hours follow relatively soon.However, dollars per hour — our monetization focus — sees an immediate, visible hit. Based on curves from a few weeks of data in each test, we believe the lines cross, i.e., the uplift from incremental retention outweighs the short-term bookings impact.

We are not providing a specific timeline, partly because we are still iterating the algorithms. Our near-term focus is improving the model to keep retention gains while reducing the near-term bookings drag.There are promising efforts underway, but it's too early to size them. The longer and more directly we measure signals, the more we capture not just early revenue but longer-term revenue tied to retention. We are optimizing a direct combo of long-term retention and long-term monetization rather than over-extrapolating noisy, quickly decaying short-term monetization signals.

Q: Can you quantify Morpheus spend or the 2024 cost headwinds?

A: Morpheus is just one part of our AI efforts, so let me address the bigger picture. The simplest way is to look at our Q3 margin guide and the YoY margin compression — roughly half is fixed-cost deleverage from lower bookings, and half is AI spend.These investments support Build and Moments, major safety upgrades, and model training for Roblox Reality — all critical and exciting, but requiring incremental spend. Importantly, titles will run in standard Roblox mode and in Morpheus' super-upsampled photoreal mode, and we initially plan to charge a subscription for the latter to offset infra.

Q: Will newly supported 2D content serve as a funnel for off-platform players, or is it meeting existing on-platform demand? How is Roblox's 2D differentiated from typical 2D mobile games?

A: Returning to 'you make the game', Build aligns well with removing content-type boundaries. When users open Build, they don't self-limit to 'a 3D multiplayer obby' or 'a puzzle played this way'.There's strong overlap in supporting whatever people want to make in Build. Many users don't draw hard lines between 2D single-player, 2D multiplayer, 2.5D orthographic, or 3D multiplayer — they view it all as play, and older cohorts show ample 2D demand.

We believe any experience built on Roblox has significant advantages: multilingual, global, social, underpinned by our economy and infra, and interoperable. This broadens the platform's genre coverage and aligns with Build — we won't limit what people can create when imagining a game to share with friends.

Q: Why was under-13 monetization hit harder — is it last year's viral hits rolling off, or is younger-user monetization inherently more volatile?

A: More the former. Last year, a few massive, high-monetizing viral hits had very high concentration of time spent, particularly among younger users.As that mix shifts — those titles being replaced by games with more 'normal' monetization — the impact shows up most in the younger cohort. We do not see a structural change in younger-user behavior; it's more that what's trending now doesn't monetize at last year's extreme levels.

Q: Incubator and 18+ initiatives seem to be progressing. What have you learned, how will the program evolve, and what else is in flight?

A: Content comes not only from incubators but also studio partners and the existing community. We ultimately see ourselves as a UGC platform and value that strength, while also leaning into a sales-engineering-style capability — building tight relationships with top creators and collaborating deeply to guide what performs on our platform.A focus area is the new tech introduced with incubator partners: significantly higher-performance avatars and worlds that run smoothly on low-end Android and high-end PC, and a synthesis tech called Slim that enables highly complex avatars to perform on phones. We believe these are major technical opportunities for partners building for us.

They also benefit from the DevEx 50% premium for 18+ consumer titles. In short, it's not just incubators — it's company-scale studios and our developer community, and the output quality we see is very strong.

<End>

Risk disclosure and statements:Dolphin Research Disclaimer and General Disclosure