Q3 Beat the Top of the Guide and 2026 Will Grow After All, but Q4 Still Guides Bookings Down 10-15%: Upgrading to Hold

Results vs. Consensus

Q3 2026 Scorecard

Two notes on the consensus figures. First, as in Q2, aggregators label a $1.62-1.64B "revenue" estimate that is in fact the bookings estimate; GAAP revenue was $1,572M and we found no clean GAAP revenue consensus. Second, published DAU estimates span 131M to 146M, so we show the line as mixed. The ~131M figure comes from the more detailed KPI-level build and sits close to what normal Q2-to-Q3 seasonality on 123M would produce; against the ~146M figure, the 138M print is a miss.

Year-over-Year Comparison

Quarter-over-Quarter Comparison

Income Statement Detail

Revenue and bookings assessment. Bookings of $1,697M rose 9.0% sequentially from $1,557M. That step is within the range of normal summer seasonality: Q2-to-Q3 bookings moved between −4% and +18% in 2021-2024 before 2025's viral summer produced +34%. What made it a beat is that the guide assumed only a 1% to 6% sequential increase. Against Q3 2025's extraordinary $1,922M, bookings fell 11.7%, but on a two-year basis they rose 50% from $1,129M. GAAP revenue grew 16% to $1,572M, driven, in the company's words, by "higher amortization of prior period deferred revenue." Underneath, consumable items fell to 10% of virtual-item revenue from 22% a year ago, and the average number of daily unique paying users fell to approximately 1.8 million from 2.3 million. The company says constant-currency growth rates were consistent with reported rates, so foreign exchange played no material role this quarter.

Margins assessment. The GAAP operating margin improved 376bps to (18.0)% and Adjusted EBITDA tripled to $138M, but both measure costs against a lagging revenue line. Against bookings, which is the honest denominator for costs incurred in the period, each of the three cost lines the company measures against bookings rose: developer exchange fees reached 24% of bookings (+140bps) on the September 2025 creator payout increase and the 42% DevEx rate increase for US adult in-game spend, certain infrastructure and trust and safety costs reached 15% (+440bps) on AI training and inference, and personnel excluding stock compensation reached 17% (+360bps) on headcount of roughly 3,400. Sales and marketing rose 66% to $111M, of which $31M was advertising and promotion. General and administrative rose 40%, almost entirely from $49M of state settlement accruals, which are again excluded from Adjusted EBITDA. Without that add-back, Adjusted EBITDA would have been $89M rather than $138M, and the nine-month figure $249M rather than $389M.

EPS and cash assessment. The $(0.32) loss beat a ~$(0.42) Street by about $0.10 and narrowed from $(0.37). Interest income of $60M on $6.2B of cash and investments narrowed the $(283)M operating loss to a $(230)M net loss. Weighted-average shares rose 2.7% despite $200M of repurchases. Operating cash flow fell 28% to $395M and free cash flow 18% to $364M, with capital expenditure of $31M against a $170M guide. Management was clear that the capex was deferred, not cancelled: Q4 is guided to $380M of capex and full-year capex to $468M, the low end of the initial 2026 plan. Net cash and investments reached $5.2B.

Key Operating Metrics

The seven-quarter series shows the two halves of the quarter in one table. Bookings, DAUs, hours and payers all rose sequentially for the first time since Q3 2025, which is the good news. All four are also below their year-ago levels, and DAUs and hours are down year over year for the first time since Roblox listed in 2021.

Hours, payers and per-DAU and per-payer bookings are the company's figures as restated in its Q3 2026 investor data sheet (one decimal), which revise prior rounded disclosures for Q4 2025 through Q2 2026 slightly.

Assessment. The sequential turn is real and it is what a recovering platform should show. DAUs rose 12% from Q2, in line with the 12-13% summer step of 2022 and 2024, and hours rose 17%, in line with the 14-19% of 2022-2024. The best number is one the table cannot show: management says seasonally adjusted retention is the highest since 2021, which means the people already on the platform are staying more than in any recent year. The weaker number is average bookings per DAU at $12.29, the lowest of the seven quarters, which says the recovery is in engagement and not yet in spend per user. Bookings per payer rose 5% year over year, so the payers who remain are spending more; the problem is that there are 16% fewer of them.

Geographic and Cohort Performance

Last quarter the regional story was a split: developed markets stalling while Japan and India compounded. This quarter the split closed from the wrong side. Every region now shows lower DAUs and hours than a year ago, against comparisons inflated by the summer 2025 viral titles, and the international markets that offset the US last quarter have lost most of their growth.

US & Canada: monetization stabilized at a lower level, users went backwards

US and Canada DAUs fell 10% to 24M and hours fell 15%, after +6% and +1% in Q2. Bookings fell 13% to $907M and ABPDAU of $38.52 was down 4%, a slightly smaller decline than Q2's 5%. Bookings per hour in the region grew 2% year over year, which management describes as an improvement on Q2. The younger cohorts remain the weak spot.

"Similar to Q2, monetization trends for younger U.S. cohorts remained challenged, due to difficult prior-year comparables, partially offset by realized improvements from updates to our RFY algorithm."
— Q3 2026 shareholder letter

Assessment: The monetization line in the platform's most valuable region stopped getting worse; it did not get better. ABPDAU has now been roughly $38.5-38.9 for three quarters, against $40-51 in Q2-Q4 2025. The region still generates 53% of bookings, and it is where the comparison against the 2025 viral titles is hardest. We read US and Canada as stabilized but not recovering.

Europe, Asia-Pacific and rest of world: the international offset has faded

Europe DAUs fell 5%, Asia-Pacific 9% and rest of world 11%. Within those regions, the named markets are still growing: Japan DAUs rose 33% and hours 22%, India DAUs 37% and hours 28%, Saudi Arabia 30%, France 12% and Germany 13%. But Japan and India have now decelerated for two quarters, from +96% and +84% in Q1 to +67% and +64% in Q2 and +33% and +37% in Q3. Asia-Pacific DAUs fell 9% even with Japan and India growing, so the rest of the region declined, and the letter does not identify where.

Assessment: International growth was the second bull pillar because it offset US saturation. In Q3 it offset nothing at the regional level: every region declined in DAUs, hours and bookings. The pillar is not broken, since the named markets still grow and less than 2% of the population outside the US uses Roblox daily. But the growth rates are halving each quarter against 2025 comparisons, and we move it to at-risk.

The over-18 cohort: still the fastest-growing, decelerating hard, and increasingly paid for

Over-18 users were 29% of global age-checked DAUs and 37% in the US, up from 27% and 33% at Q2. US over-18 DAUs grew 18% and hours 13%, and the 18-34 band grew 26% and 23%. Two quarters ago those figures were "over 40%" and "over 50%"; last quarter they were +32%/+27% and +42%/+37%. On a two-year basis US over-18 DAUs rose 89% and the 18-34 band 114%. US adults monetize more than 50% above under-18 users.

"We also plan to continue to increase marketing spend targeting over 18 users. We have been encouraged by what we have seen as we have been dialing up this type of spend. Retention, engagement, and monetization of paid users is outperforming organic sign ups."
— Naveen Chopra, Chief Financial Officer

Roblox ran more than 200 user acquisition campaigns this summer, with marketing spend aimed at adults up nearly three times year over year.

Assessment: The share gain is genuine and the cohort is the right one to chase, since US adults monetize more than 50% above under-18 users. But growth has slowed for two straight quarters, by 14-16 points in Q3 alone, and it is now supported by a marketing budget that nearly tripled. Paid users that outperform organic sign-ups are a good sign; a cohort whose growth rate nearly halves while its acquisition spend nearly triples is not yet evidence that the organic flywheel the bull case needs is turning. The company also still footnotes these cohort figures as extrapolated estimates. The pillar stays at risk.

Content velocity: the multi-hit ecosystem is getting faster, partly because of AI

Forty percent of the top 1,000 games by hours launched in the last 90 days, and about 40% of the weekly top 100 games on the recommended-for-you surface by impressions were created in the last three months. Two titles built almost entirely with AI, Steal an Egg and Huss Valley, reached the top 20 by concurrent players. Over the past month, more than 80% of the top 100 games by hours used AI tools in development, and daily asset uploads tripled month over month.

"Over the last month, more than 80% of the top 100 games when we rank by hours used AI in creation and we have seen 2 titles created almost exclusively with AI hit the top 20 by concurrent players during Q3."
— David Baszucki, CEO and Co-Founder

Assessment: This is the first bull pillar and it strengthened again. In Q2 we noted an uncomfortable corollary, that the long-tail content gaining share monetized worse than the viral titles it replaced. In Q3 the company says the opposite effect is starting to show: the refined algorithm surfaces new content faster and lifts both retention and monetization. If that holds, content health and near-term bookings stop being in tension, which was the core of the monetization bear point.

Key Topics & Management Commentary

Overall Management Tone: Confident and more expansive than in July, with the Q3 beat used as licence to spend most of the call on a multi-year AI platform vision rather than on the quarter. The financial discussion was brief and well-prepared, the Q4 guide was presented as deliberately cautious, and management did not engage with the bear arguments in circulation (the short-seller report, new-user acquisition, the regulatory notice disclosed in the 10-Q), none of which was raised in Q&A. Management was least concrete where investors most need numbers: the financial framework for the new initiatives and 2027 infrastructure spending.

1. Retention drove the beat, and the algorithm change turned from a cost into a lever

In July, the recommended-for-you algorithm change was half of the monetization problem: it traded near-term bookings for longer-term retention, and management would not date the payback. In Q3 the company refined the algorithm further, adding personalization and age-cohort weighting, and reports that it now improves both sides of the trade.

"With respect to Q3, as you have now seen, bookings came in above the high end of our guidance. The better than expected performance was driven largely by retention and engagement, In fact, seasonally adjusted retention reached its highest level since 2021."
— Naveen Chopra, Chief Financial Officer

"These optimizations drove meaningful improvements in overall platform retention and monetization in the quarter, which positively impacted bookings. Notably, year-over-year growth in bookings per Hour improved relative to last quarter."
— Q3 2026 shareholder letter

Bookings per hour grew about 4% year over year globally and 2% in US and Canada. The CFO was careful to say that monetization headwinds "did continue in the quarter, particularly with younger cohorts," and that the confidence is about lifetime value over time.

Assessment: This is the most important change from last quarter. The bear case we published in July rested on a monetization break that management called continuing and would not date. Q3 does not supply a date either, but it supplies evidence: retention at its best seasonally adjusted level since 2021, per-hour monetization growing faster than in Q2, and a $44M beat over the top of a guide set when the company was at its most cautious. The crossover is not proven, since ABPDAU still fell 3% and younger US cohorts are still weak. But it is no longer only asserted from "a few weeks of data."

2. The first year-over-year DAU decline as a public company

DAUs of 138M were 9% below Q3 2025's 151.5M, the first year-over-year decline in Roblox's quarterly series since it listed in 2021. Hours fell 15% and payers 16%. The letter's opening pages lead with the 12% sequential increase and with DAUs nearly doubling since Q3 2023; the 9% year-over-year decline first appears on page 11.

"In Q3, highlighting some tough compares last year with Q3, we achieved the highest seasonally adjusted overall platform retention since 2021 on the platform."
— David Baszucki, CEO and Co-Founder

Assessment: The comparison base genuinely explains most of this. Q3 2025 DAUs grew 70% on a handful of viral titles, and a 9% decline against that is consistent with a platform that kept most of its gains: DAUs are up 55% on two years and have nearly doubled since Q3 2023. What the comparison does not explain is acquisition. Retention is what the company chose to highlight, and retention is about users already on the platform. A short-seller report published two days before the print estimated that new active users fell about 50% year over year in Q3, using third-party data on a sample of accounts. That estimate is unverified, but the company's Q3 disclosure contains nothing that addresses it, and in Q2 it did discuss sign-ups.

3. The Q4 guide: below the Street, and conservative by construction

Q4 bookings are guided to $1,890-2,000M, down 10% to 15% against Q4 2025's $2,222M. The midpoint of $1,945M is about 4% below the ~$2.02B Street figure, and even the top of the range is below it. The company also widened the range "given the significant proportion of bookings that are typically generated in the last week of December."

"Now our bookings guidance does build in a modest headwind from the release of GTA in the quarter, Though candidly, we have limited data from which to size any potential impact."
— Naveen Chopra, Chief Financial Officer

In Q&A the CFO listed what the guide leaves out: the recent strength from fast-growing hit games, the Halloween update cycle, and any short-term monetization improvement. The guide assumes a seasonal DAU decline from Q3 to Q4 similar to recent years, when Q4 DAUs fell 4-5% from Q3.

Assessment: The guide is a miss against consensus and a conservative forecast at the same time. The midpoint implies Q4 bookings 14.6% above Q3. In each of the five years from 2021 to 2025, the Q3-to-Q4 step was larger: +21%, +28%, +34%, +21% and +16%. Even the top of the range (+17.9%) only clears 2025's step. A company that just beat its own guide's top by $44M, guiding to the weakest holiday step in six years with no monetization improvement assumed, is building in room. We think Q4 lands at or above the top of the range, but we would not argue that it reaches consensus, and the market will not pay for that argument until it does.

4. 2026 is a growth year after all

In July the company withdrew its annual guide and declined, when asked directly, to say whether 2026 bookings would exceed 2025. The Q4 guide answers the question by arithmetic. Nine-month bookings of $4,985M plus the Q4 range give $6,875-6,985M for the year, up 1.3% to 2.9% from $6,788M.

"And though bookings are expected to be down year over year given the comps from 2025, I would note that the midpoint of our guidance implies full year annualized growth of 26% over the last 2 years."
— Naveen Chopra, Chief Financial Officer

Assessment: Our July arithmetic put full-year bookings within a point of flat; the year now ends about 2% up at the midpoint, and above our $6.77B estimate. That is a small number with a large signal. The "is this lifecycle decline" debate needed a year that grew, and 2026 will, against the hardest comparison year in the company's history. The two-year framing (26% annualized growth) is fair on its own terms. It is also a reminder that the company now prefers two-year comparisons because one-year comparisons will be negative for one more quarter.

5. Cash flow: the zero-cash quarter that did not happen, and the negative one that will

Q3 was guided to free cash flow of $(60)M to $5M and produced $364M. Operating cash flow beat by about $250M and capex came in $139M below plan.

"That was a combination of the flow through from bookings outperformance and also the shift of $140 million CapEx from Q3 into Q4, which you will also see in our free cash flow guidance."
— Naveen Chopra, Chief Financial Officer

"We are currently planning for $380 million of CapEx in the fourth quarter, that implies $468 million for the full year. Which sits at the low end of the expectations that we shared at the beginning of the year."
— Naveen Chopra, Chief Financial Officer

Q4 free cash flow is guided to $(190)M to $(105)M, the first negative quarter in the company's 2024-2026 data. Full-year 2026 free cash flow therefore lands at $1,064-1,149M, down 15% to 21% from 2025's $1,353M but inside the $1,050-1,275M range withdrawn in July.

Assessment: We called the cash-flow pillar broken in July on a Q3 guide that implied near-zero free cash flow and an annual figure near $930M. Both premises were too pessimistic. The capex is real but its timing moved, and on a full-year view the company converts about 16% of bookings to free cash flow in a year when bookings barely grow. The pillar is repaired to at-risk, not restored: 2026 free cash flow still declines, Q4 is negative, and stock compensation of roughly $1.15B a year now exceeds free cash flow.

6. The AI creation push: Build, Studio Assistant and an "explosion of content"

Management framed the last three months as an inflection point for AI-driven game creation. Build, the prompt-to-game tool, has reached more than 500,000 users and 70,000 published games in its alpha markets (New Zealand, Singapore, Serbia and Thailand), and more than 75% of its users had never opened Roblox Studio. Daily users of the company's MCP server grew nearly five times since Q2, and games published using AI grew more than 50% sequentially.

"We are going to be doing a broader US release of build in December and a global rollout in 2027."
— David Baszucki, CEO and Co-Founder

Build will keep a free base tier with paid tiers for additional prompt capacity, and some Build titles already run ads and earn money.

Assessment: The strategic argument is coherent: if anyone can make a game, distribution, discovery, safety and an economy become the scarce assets, and Roblox owns all four. The evidence so far is about creation volume, not about users or bookings, and the risk is the one a heading in the letter names itself: "AI Slop." More content helps only if discovery can find the good part of it; the Q3 algorithm results suggest it can. We treat AI creation as a credible long-dated option with no financial framework yet, and as a direct driver of the infrastructure cost growth discussed below.

7. Roblox Everywhere and the Universal Player: Roblox content off Roblox

Two new initiatives extend the platform beyond its own app. Roblox Everywhere will let creators publish standalone games built on Roblox into third-party app stores, with Roblox curating select top titles first. The Universal Player aims to import "literally any existing PC, console, or mobile game" onto the platform using AI.

"In phase 1, which we estimate will launch in the first half of 2027. We, at Roblox, are going to directly curate and publish select top titles as stand alone apps directly into major app stores. In Phase 2, our vision is to allow any creator to generate a standalone binary directly within their creator portal and submit it on their own to the app Store of their choice."
— David Baszucki, CEO and Co-Founder

Assessment: These are the most consequential strategic announcements on the call and the least quantified. Roblox Everywhere reverses twenty years of a single-app strategy: it trades some of the app's network effect for reach into off-platform audiences, and every standalone title sold through a third-party store carries that store's fees and its own discovery economics. The Universal Player raises intellectual-property questions that management answered only at the level of principle. Neither has a revenue model, a cost estimate or a date beyond the first phase. We would not put value on either in a 12-month rating.

8. Monetization beyond Robux purchases: Plus and advertising

Roblox Plus, the subscription launched in April, has "millions" of active subscribers. In advertising, more than 70% of the top 100 games by Robux spent ran sponsored ads, more than 20% of games using rewarded video earned over 10% of their in-game revenue from ads, and the company plans to launch skippable pre-roll video ads.

"Plus subscribers average 2x the playtime, spend 7x more Robux and show materially higher adoption of social features like private servers compared to non-Plus users."
— Q3 2026 shareholder letter

Assessment: Both are sensible and both are small today; advertising remains an "insignificant" part of bookings by the company's own definition. The Plus statistic describes who subscribes more than what subscribing does, since the most engaged spenders are the ones most likely to sign up. The useful disclosure would be incremental spend after subscribing, and the company did not give it. We treat both as optionality that would matter in a 2027 framework, not as offsets to the Q4 bookings decline.

9. Infrastructure: costs keep rising, and they are now fixed

Certain infrastructure and trust and safety costs rose 24% to $258M on AI training and inference, reaching 15% of bookings from 11%. Q4 guidance includes further growth in infrastructure for Build, Roblox Reality and the Universal Player, plus higher marketing aimed at adults.

"From a modeling perspective, I think you should assume that, as we called out, in Q4, we do expect infra expense will grow. I expect there will be some additional year over year growth as we go into next year. Obviously, too early to get into any specifics about that."
— Naveen Chopra, Chief Financial Officer

Assessment: Q3's margin beat came from revenue, not from cost restraint. Q4 Adjusted EBITDA is guided to $(70)M to $(10)M, a swing of roughly $150-210M from Q3's $138M on similar revenue, which tells you how fast the AI cost base is growing. The CFO repeated July's point that the new spending is fixed rather than variable. That cuts both ways: it creates operating leverage if bookings resume growth, and it removes the cushion if they do not. Without a 2027 cost framework, investors cannot underwrite either case.

10. Safety execution improved, and the legal and regulatory cost kept rising

Age-check penetration reached 64% of global DAUs, including more than 80% of DAUs aged 13 to 17, and 80%, 75% and 75% of all DAUs in Australia, the US and the UK. Users who chat or use voice grew 24% sequentially, twice the rate of platform DAUs.

"In Q3’26, DAUs who sent a chat message or connected to voice grew 24% quarter-over-quarter, double our total platform DAU growth rate."
— Q3 2026 shareholder letter

The legal picture moved the other way. Q3 carried $49M of state settlement accruals ($140M year to date), with settlements now signed with Alabama, Mississippi, Nevada, South Dakota, West Virginia and Kansas. Courts in Tennessee and Los Angeles County rejected the company's early motions to dismiss in September. The EU designated Roblox a Very Large Online Platform in August, with obligations from January 2027, and the 10-Q adds a new item.

"In addition, in October 2026, the Federal Trade Commission (“FTC”) and the states of Pennsylvania and Utah notified the Company that as part of an ongoing investigation they are considering filing claims based on alleged violations of the Federal Trade Commission Act, the Children’s Online Privacy Protection Act (“COPPA”), and certain state consumer protection laws seeking monetary damages and declaratory and injunctive relief."
— Q3 2026 Form 10-Q

Assessment: The engagement cost of safety is contained: chat activity is recovering and the rollout no longer suppresses usage. The financial cost of safety is not contained, and it is growing in a way the adjusted metrics hide. The company keeps classifying settlement costs as outside the ordinary course while settling with a sixth state and receiving a federal notice of possible COPPA claims. We add a separate bear point for youth-safety legal and regulatory cost, rated emerging, because it is now a recurring line rather than a one-off.

11. Capital allocation: the buyback slowed and the M&A door opened

Roblox repurchased 4.4M shares for about $200M in Q3, down from $380M in Q2, leaving $2.4B on the $3B authorization. Fully diluted shares were 755M, up 2% year over year.

"As a reminder, this program is designed to partially offset dilution from employee equity grants while preserving strategic optionality to invest in future growth opportunities."
— Q3 2026 shareholder letter

Assessment: With $5.2B of net cash, the company bought about half as many shares as in Q2 at roughly the same average price, about $45 against about $46, even though the stock closed as low as $35.54 in August. In July management described a buyback that buys more when the price falls; Q3 did not show that. Asked about capital structure, the CFO put organic investment first and M&A optionality, including "bigger things," second. We read the slower buyback and the explicit M&A language together: management is keeping its cash for the AI program and possible acquisitions rather than for share count reduction. That is defensible, and it means the buyback should not be treated as a source of support for the shares.

Guidance & Outlook

What the Q4 guide implies for the full year

Management's framing paired caution on the next ten weeks with conviction on the longer arc. The CFO described the guide as excluding recent hit-game strength, the Halloween event cycle and any monetization improvement, with a modest GTA VI headwind built in; the CEO spent the call on the platform initiatives that are meant to drive 2027 and beyond. Margin guidance was framed explicitly as investment: "we expect fixed cost deleveraging relative to the year-ago period," and the company continues to fund the core app, the Universal Player and Roblox Everywhere ahead of their revenue.

Implied quarter-over-quarter ramp. The bookings range implies Q4 at 11.4% to 17.9% above Q3's $1,697M, with a midpoint of 14.6%. Q3-to-Q4 steps in 2021-2025 were +20.7%, +28.2%, +34.2%, +20.7% and +15.6%. Applying the smallest of those (2025) to Q3 gives about $1,962M, slightly above the midpoint; applying the 2021-2024 average of about 26% gives about $2,140M, above consensus. The guide therefore assumes a weaker holiday season than any of the last five, which is consistent with the stated exclusions and the GTA VI assumption.

Street at. Consensus sat near $2.02B for Q4 bookings before the print, above the top of the new range. Full-year bookings estimates near $6.90B were already close to the implied range, so the revision is concentrated in Q4 and in the 2027 starting point: a Q4 exit rate of −10% to −15% makes the first half of 2027 harder to model as growth, because Q1 2027 laps a Q1 2026 that still grew 43%.

Guidance style. Roblox's guidance has swung from reliably conservative (2025, two large raises) to unreliable on the downside (Q1 and Q2 2026, a cut and a withdrawal) and now back toward conservative: Q3 landed $44M above the top of a range set in July. The Q4 range is wider than usual and, on the seasonal history above, cautious. We expect Q4 at or above the top end, which would put full-year bookings near $6.99B, up about 3%.

Analyst Q&A Highlights

How much caution is built into the Q4 bookings guide

The first financial question asked what the Q4 guide assumes, how much conservatism it contains as comparisons ease late in the quarter, and how to think about the GTA VI launch. The answer listed three things the guide leaves out and conceded that the GTA VI assumption rests on very little data.

Q: "First, could you talk a little bit maybe more near term around the assumptions on the 4Q bookings guidance? And kind of how much conservatism you bake in as the comps ease, especially towards the end of the quarter?"
— Ken Gawrelski, Wells Fargo

A: "Number 1, some of the strength we have seen recently has been a function of some, you know, big hit games that have grown quickly on the platform. As well as a high volume of updates associated with Halloween and events that we have got planned around Halloween. And I mention that because for guidance purposes, we do not assume that those things will necessarily repeat in the coming weeks. Similarly, for guidance purposes, we do not assume short term improvement in monetization."
— Naveen Chopra, Chief Financial Officer

Assessment: This is the clearest statement on the call that the Q4 range is a floor-building exercise. Management also said the trends have been "trending very positively as we have gone into the beginning of Q4." A guide that excludes current momentum, excludes monetization improvement and adds a competitive headwind is one the company expects to beat; the question is by how much, and the seasonal history says the top of the range is attainable.

What the AI program costs in 2027

The same questioner asked how infrastructure spending develops into 2027 to support Build, Roblox Reality and the Universal Player. The answer confirmed growth and declined to size it.

Q: "Could you talk about the prepare your preparedness to enable all of the tools and experiences David detailed to open the call I know you talked about some 4Q expected deleverage, but can you touch upon 2027 and beyond infrastructure spending?"
— Ken Gawrelski, Wells Fargo

A: "From a modeling perspective, I think you should assume that, as we called out, in Q4, we do expect infra expense will grow. I expect there will be some additional year over year growth as we go into next year. Obviously, too early to get into any specifics about that. But I would remind you of something we talked about last quarter as well, that a lot of the investments we are making in-house now take more of the form of fixed expense as opposed to variable. A lot of it is related to model training."
— Naveen Chopra, Chief Financial Officer

Assessment: "Some additional year over year growth" in a fixed cost base, with no number attached, is the main reason we cannot build a confident 2027 margin case. The CEO added that the company will add GPU capacity for Build inference while keeping its record of running infrastructure efficiently. Both statements can be true; neither is a forecast.

The building blocks of a return to bookings growth

The opening question asked how the algorithm balances viral hits against long-term retention, and what the building blocks are for a return to bookings growth, particularly with adults. The answer described direction rather than a path.

Q: "And then conversely, as we look towards return to bookings growth over time, can you just talk about the building blocks of improvements across the algo AI content and distribution? Especially across the over 18 cohort."
— Cory Carpenter, JPMorgan

A: "What we have seen recently is both in that the algorithm is pushing more dynamic content to the surface well simultaneously bringing long term retention high. And you have also seen recently more and more new hits popping up on the platform, and our algorithm has been supporting that. As I look to what the home page looks like for an 18-and-up user, it is looking more and more like a platform for those age cohorts."
— David Baszucki, CEO and Co-Founder

Assessment: The question invited a timeline for bookings growth and received a description of the homepage. Nothing in the answer is inconsistent with the Q3 data, which does show the algorithm surfacing new hits while retention rises. But the company again declined to say when growth resumes, which leaves the 2027 shape to the model rather than to management.

When AI creation velocity turns into users and bookings

Questioning asked which AI creation milestones investors should track and when faster game creation should translate into user and bookings growth. Management pointed to milestones already reached and promised more statistics, without a timeline.

Q: "And then when should we expect that higher velocity to start to translate into user and bookings growth?"
— Matthew Cost, Morgan Stanley

A: "I think you are already seeing early signs of it. A year ago, I do not think we would have imagined that we would have 2 of the top 20 experiences on Roblox primarily AI generated."
— David Baszucki, CEO and Co-Founder

Assessment: Two AI-built titles in the top 20 is a real data point, and the US Build launch in December gives investors something observable within a quarter. The honest answer to the question, though, is that nobody knows yet whether more games produce more users or only more competition among the same users. The Q4 guide does not assume any lift from it.

Intellectual property and the Universal Player

A recurring line of questioning asked how the plan to import any existing game works when the games belong to other rights holders, and what role Roblox will play in connecting IP owners with its developers. The answer stated principles and a self-serve model.

Q: "Are you gonna need licenses with the IP owners whose games you are trying to port over onto Roblox?"
— Multiple analysts incl. James Heeney (Jefferies), Martin Yang (Oppenheimer)

A: "Yeah. So first, once again, we, you know, we really expect respect IP and respect DMC takedown. We want licensed IP on our platform"
— David Baszucki, CEO and Co-Founder

Assessment: Honoring takedown notices is a notice-and-removal model, which places the burden of policing imports on rights holders. For a platform that already carries youth-safety litigation, an AI system designed to import other people's games invites a second category of legal exposure. The company also described direct work with major studios through its developer-relations team, which is the safer route. We would want to see the licensed version before valuing the initiative.

Competition for developers as AI tools spread

Questioning asked whether new AI-powered creation platforms raise the risk of competition for developer time and talent. The answer listed the five structural advantages from the shareholder letter, led by distribution.

Q: "how do you think about the risk of increased competition for developer time and talent as these alternative platforms potentially emerge?"
— Chris Scholl, UBS

A: "Just the first is the size of our distribution platform, especially when it is powered by social and a safety infrastructure and an economy, of the reason we are seeing so much MCP usage and tagging into our ad system is when people build games, they are looking for a way to get them to distributed."
— David Baszucki, CEO and Co-Founder

Assessment: The distribution argument is the strongest one Roblox has, and the evidence cited (MCP adoption and developers buying traffic through the ad system) supports it. It is also an argument that Roblox Everywhere partly undercuts, since that initiative helps creators reach audiences off the platform. The answer did not address whether any competing platform is gaining developers.

Whether the chat damage from age checks is structural

Questioning returned to the communications density lost after the global age-check rollout and asked whether any of the shortfall against pre-age-check levels looks structural. The answer gave one new statistic and did not answer the structural question.

Q: "So can you give us an update on how that curve has been trending in 3Q as the age checks maybe sit in for longer. And to the extent that it is lower than pre age checks, is there any element of that you think might be structural at this point?"
— Andrew Marok, Raymond James

A: "On the comm side, the early stat I will give you is our chat DAUs are up 24% quarter-over-quarter. So as we age check more, we are seeing more chat activity."
— David Baszucki, CEO and Co-Founder

Assessment: A 24% sequential increase in chatting users is good news and twice platform DAU growth. In July the company said communications were still below pre-age-check levels; this quarter it neither repeated nor withdrew that statement. Silence on the comparison management itself introduced suggests the gap has narrowed but not closed.

What the cash pile is for

A question on capital structure noted that net investments are now a meaningful share of the market capitalization and asked for the company's philosophy. The answer ranked organic investment first and acquisitions second, with the buyback framed as dilution management.

Q: "Can you just remind us sort of what your philosophy is in terms of your cap structure? Because I only ask because it is not an insignificant part of your market cap now."
— Jason Bazinet, Citi

A: "We have actually had some great tuck in type acquisitions where we have been able to bring amazing talent into the company, particularly in the AI space. You will see us continue to do that. And similarly, having optionality to do bigger things there, I think, is important."
— Naveen Chopra, Chief Financial Officer

Assessment: "Bigger things" is new language, and it arrives alongside a slower buyback. With $5.2B of net cash, about a sixth of the market capitalization after today's decline, a large AI acquisition is now a plausible use of the balance sheet. We would treat that as a risk to the equity story rather than a support, because it would add cost and integration risk to a year that already carries a large internal investment program.

What They're NOT Saying

1. New-user acquisition: the Q2 letter said sign-ups improved through the quarter; the Q3 letter discusses retention at length and sign-ups not at all. That silence arrives two days after a short-seller report estimated a roughly 50% year-over-year fall in new active users, and no one on the call asked about it.
2. The federal regulatory notice: the FTC, Pennsylvania and Utah told the company in October that they are considering claims under the FTC Act and COPPA, seeking monetary damages. This appears only in the 10-Q. The company says it cannot estimate the loss.
3. The first year-over-year DAU decline: the letter's opening pages lead with the 12% sequential increase and with DAUs nearly doubling since Q3 2023. The 9% year-over-year decline, a first for the public company, first appears on page 11.
4. Younger-cohort monetization in numbers: described again as "challenged," with no figure. Bookings per hour is disclosed only as a growth rate. ABPDAU fell 3% and US and Canada ABPDAU 4%, so the per-hour improvement coexists with lower spend per user.
5. Any 2027 framework: no bookings, margin, infrastructure or capex outline for a year in which the company is asking investors to fund Build, Roblox Reality, the Universal Player and Roblox Everywhere. The CFO called it "too early to get into any specifics."
6. The economics of Roblox Everywhere: no disclosure of who pays third-party app-store fees on standalone titles, how revenue is shared with creators, or how much the Roblox app could lose to its own standalone games.
7. Return on adult-focused marketing: management cites a "high ROAS bar" and paid users that outperform organic users, but gives no acquisition cost, payback period or budget. Sales and marketing rose 66% to $111M.
8. The verified-account allegation: the short-seller report says age-verified child accounts are bought and sold on third-party marketplaces at scale. The company's safety section describes its age-check penetration and moderation systems but does not address account resale.
9. The full legal bill: $140M of settlement accruals year to date is excluded from Adjusted EBITDA. The 10-Q notes that non-cash settlement commitments (public service campaigns, law enforcement liaison headcount, platform changes) are expensed as incurred and are not accrued. The company has not said whether those costs will also be excluded.

Market Reaction

• Pre-print setup: closed at $52.46 on October 29 ahead of the after-the-close release. Down 35.3% year to date against the S&P 500's +12.8%, down 60.8% over twelve months, but up 27.4% over the trailing thirty days and up 47% from the $35.60 close that followed the Q2 print. The 52-week closing range entering the print was $35.54 to $133.74. Options were pricing a move of roughly 11-12%.
• After hours, October 29: approximately −17%, to about $43.55.
• Reaction session, October 30: opened at $45.20, a 13.8% gap down, traded a range of $41.39 to $47.68, and stood at $42.23, down 19.5% or $10.23, at 3:11 PM ET. The intraday low marked a 21.1% decline.
• Volume: 21.9 million shares by 3:11 PM ET against an 11.3 million thirty-day full-session average, already 1.9 times normal.
• Context: the S&P 500 was down 2.2% at the same time, so roughly 17 points of the decline were stock-specific. The price remains 19% above the $35.60 close that followed the Q2 print. Figures are intraday; the session had not closed at publication.

The stock fell further than options implied on a quarter that beat on bookings, EPS, users and cash flow, which tells you the market was trading the setup and the guide, not the quarter. The shares had risen 47% from the $35.60 post-Q2 close in three months, much of it in October on the expectation of a strong summer; a beat was priced, a guide above consensus was not. The Q4 range then came in below the ~$2.02B Street figure at every point, with negative Adjusted EBITDA and negative free cash flow attached. For a stock that had rallied on the belief that the bookings decline was ending, a guide showing a 10-15% decline for another quarter was the wrong headline.

Three secondary factors compounded the move. A short-seller report two days earlier had put child safety and new-user acquisition back in the debate, and management did not address either. The CFO's own reference to a GTA VI headwind gave the market a reason to doubt holiday engagement just as Q4 enters its heaviest weeks. And the decline came on a weak tape, with the S&P 500 down 2.2% on the day. We think the sell-off overstates what changed: the year now grows, the cash flow did not collapse, and the guide is conservative on its own seasonal history. But it correctly prices the absence of any evidence that bookings growth returns before the second quarter of 2027.

Street Perspective

Debate: Is Q3 the trough, or does the Q4 guide show the decline is not finished?

Bull view: Q3 beat the top of the guide, retention is the best since 2021, every volume metric rose sequentially, and the algorithm change that caused the Q2 monetization problem is now credited with fixing it. The Q4 guide excludes current momentum and any monetization improvement. Comparisons ease sharply from Q2 2027, when the platform laps a quarter that grew only 8%, and 2026 already grows.

Bear view: DAUs, hours and payers are all below year-ago levels and DAUs declined in every region. US adult growth nearly halved while marketing aimed at adults nearly tripled. The Q4 guide is below consensus even at the top, Adjusted EBITDA and free cash flow turn negative, and Q1 2027 laps a quarter that grew 43%, so year-over-year bookings growth may not return until the second quarter of 2027.

Our take: The bull view is closer to right on direction and the bear view on timing. Q3 is the first quarter since Q3 2025 in which the operating data improved, and the retention evidence is the best counter-argument yet to the lifecycle-decline thesis. But the next positive year-over-year bookings print is probably two quarters away, and nothing in the Q4 guide lets the market price it sooner.

Debate: Does AI-generated content strengthen Roblox or commoditize it?

Bull view: When anyone can make a game, distribution, discovery, safety and an economy become the scarce assets, and Roblox owns all four at a scale no new entrant can match. AI-built titles already reach the top 20, Build users are mostly new creators, and asset uploads tripled in a month. AI lowers Roblox's creators' costs faster than it creates competitors.

Bear view: The same tools that let a teenager build a Roblox game in a weekend let anyone build a game anywhere. Roblox's own response, Roblox Everywhere and the Universal Player, concedes that games will live outside its app, and it brings third-party app-store economics, IP exposure and higher fixed infrastructure costs with no disclosed financial model. A flood of content also strains discovery and moderation, the two systems the safety debate already questions.

Our take: We lean toward the bull reading on the platform and the bear reading on the financials. The distribution argument is strong and the Q3 discovery data supports it. But every AI initiative announced so far adds cost before revenue, and the company has given no framework for when the spending pays back. Until it does, AI is a reason to hold the stock for optionality, not a reason to pay more for it.

Debate: After the sell-off, is the stock cheap or still without a floor?

Bull view: At $42.23 the shares trade at about 3.9 times 2026 bookings on enterprise value, against 5.0 times before the print. Net cash of $5.2B covers about 16% of the market capitalization, 2026 bookings now grow, and the Q4 guide is conservative on six years of seasonal history. A Q4 print at the top of the range would arrive in February, just as comparisons start to ease.

Bear view: Free cash flow declines in 2026 and turns negative in Q4, stock compensation of roughly $1.15B a year exceeds free cash flow, the buyback is slowing, and youth-safety litigation now includes a federal notice of possible COPPA claims. A growth multiple on bookings that are still falling year over year has no floor until growth returns, and the next catalyst that could show it is the Q1 2027 print.

Our take: Neither side has a decisive argument, which is the case for Hold. The bear camp is right that the multiple has no fundamental floor until year-over-year bookings growth returns. The bull camp is right that the sell-off priced a guide that is cautious by construction. Our rebuilt base range of $39-45 brackets the $42.23 intraday price, and the next two catalysts, the Q4 print in February and any 2027 framework, are as likely to help as to hurt.

Model Update Needed

Valuation framework. On the 755 million fully diluted shares at September 30, the $42.23 intraday price equates to a market capitalization of about $31.9B. Net of $5.2B of net cash and investments, enterprise value is about $26.7B, or 3.9 times our $6.93B 2026 bookings estimate and 3.7 times our $7.3B 2027 estimate. Entering the print the same 2026 measure was 5.0 times. On cash, the shares trade at about 29 times our $1.1B 2026 free cash flow estimate, but stock compensation of roughly $1.15B a year means free cash flow after stock compensation is close to zero. That is why we value the business on bookings and treat the multiple as a judgment on when growth returns.

Valuation impact. Our base case of $39-45 implies −8% to +7% from $42.23, a range that brackets the current price. The bull case of $56-62 implies +33% to +47% and requires Q4 bookings above the top of the range, a 2027 outline with growth from the first half, and evidence that adult acquisition and AI creation show up in bookings. The bear case of $28-33 implies −34% to −22% and requires bookings to decline again in 2027 with free cash flow below $1B. The skew is modestly positive and the dispersion is wide, which is what a Hold means here.

Thesis Scorecard Post-Earnings

Against the tests we set in July

Overall: Thesis improved. The July downgrade rested on three claims: that monetization had broken in a way management could not date, that the company had lost the ability to forecast its own year, and that the cash-flow pillar was going to zero. Q3 weakened all three. Retention is at its best seasonally adjusted level since 2021, per-hour monetization is growing faster than in Q2, the year now grows on the company's own numbers, and full-year free cash flow lands near $1.1B. Two pillars still deteriorated: the international offset faded and adult growth slowed while its cost rose. We also add a new bear point for youth-safety legal cost, which is now recurring. On balance the business is in better shape than we argued in July, and the price we argued from was too low.

Action: Upgrade to Hold from Underperform. We were wrong on price after July: the specific claims behind the downgrade proved too pessimistic, and the stock rose 47% into this print. Today's decline returns the shares to $42.23, inside our rebuilt base range, so we neither chase the sell-off nor stay negative on a business whose operating data just turned. Holders can stay: the Q4 guide is cautious, comparisons ease from Q2 2027, and the balance sheet removes any solvency question. We would move to Outperform on two of three things: Q4 bookings above the $2,000M top of the range, a 2027 outline showing year-over-year bookings growth from the first half, or disclosure on new-user acquisition that answers the short-seller estimate. We would move back toward Underperform if Q4 bookings land in the lower half of the range (below $1,945M), or if the federal COPPA matter becomes a filed complaint with material monetary demands.