company:

Roblox Corporation (NYSE: RBLX)

sources used:

10-Ks FY2021–FY2025; Q2 2025 and Q2 2026 10-Qs; shareholder letters Q3 2025–Q2 2026 (8-K Ex. 99.1); Q2 2026 earnings call transcript; 2026 proxy (DEF 14A). Company sources only; no outside sources.

latest period:

Quarter ended 30 June 2026 (reported 30 July 2026); Q3 2026 guidance only.

missing or stale data:

Full-year 2026 guidance withdrawn. Q2 bookings per DAU and per payer are my calculations from reported bookings, DAUs and payers. Adult (O18) growth and monetization figures are management extrapolations from age-checked users. Advertising and subscription revenue never quantified. Q3 2026 not yet reported.

Roblox — Bull & Bear Memo

Prepared 11 September 2026

1. Business in one line

Roblox runs a free platform where users play games built by independent creators and buy Robux (virtual currency) to spend inside them. Roblox takes the cash up front, pays creators through the Developer Exchange (DevEx), and keeps what is left after app-store fees, servers and safety. It never pays to develop the content. FY2025 bookings were $6,788M from 127M average daily users (DAUs), of whom only about 1.8M paid on an average day FY25 10-K.

Fast grower — its bookings follow a hit cycle and currently look cyclical. Bookings rose 55% in FY2025 and are guided down 14–18% year on year for Q3 2026 Q2'26 letter. The memo turns on whether the fast-grower label survives the down leg.

2. Bull case — Peter Lynch pitch

Toggle to line up each bull assumption beside the bear risk that attacks it.

The simple reason this stock could work is that Roblox already takes nearly 4% of global gaming spend without paying to develop content. The 2026 bookings drop is mostly a comparison against 2025's viral hits, not users leaving. And the next leg is adults, who account for roughly 80% of gaming spend Q2'26 letter. Roblox can only now target them directly because it now knows its users' ages.

A1The 2026 decline is a hit comparison, not audience erosionattacked by B2

What must happen: Bookings return to year-on-year growth once the 2025 viral-hit quarters roll off (Q3 2025: $1,922M, +70% Q3'25 letter; Q4 2025: $2,222M), with DAUs and payers still growing.
Why Roblox specifically: Engagement is spread across millions of creator-made games, so creators replace a fading hit without Roblox funding it. The top 10 games fell from ~30% of hours three years ago to ~20% in Q2 2026 Q2'26 letter.
Evidence: Q2 2026 DAUs +10% to 123M, monthly unique payers +15% to 27M, existing-user retention "stable" Q2'26 letter. Management blames the miss on lower spend per hour: play shifted away from high-monetizing 2025 hits, and the recommendation algorithm was retuned to favour retention (28-day window, up from 7).
Monitor: Q4 2026 bookings vs $2,222M; bookings per payer.

A2Age-checking unlocks the older, higher-spending cohortattacked by B4

What must happen: Adults rise as a share of DAUs and lift bookings per DAU above the ~$53 a year it sat at in both FY2024 ($52.7) and FY2025 ($53.5; bookings ÷ average DAUs).
Why Roblox specifically: Roblox has verified age data at this scale and can now route content and creator pay by age. From 8 June 2026, creators earn a 42% higher DevEx rate on spend by age-checked US adults (26.6% → 37.8%) Q1'26 letter; Q2'26 10-Q.
Evidence: Adults are 27% of age-checked DAUs globally and about a third in the US. US adult DAUs +32% and 18–34 DAUs +42%; US adults spend over 50% more than under-18s Q2'26 letter. Both figures are extrapolations.
Monitor: Adult share of age-checked DAUs; bookings per DAU.

A3Safety becomes a moat rather than a permanent taxattacked by B1

What must happen: Age-check penetration reaches the 90% target, in-game chat recovers above pre-age-check levels, and state claims settle for sums that are small next to cash flow.
Why Roblox specifically: Age estimation, text and voice moderation and age-based accounts (Kids, Select) all run inside Roblox's own stack. Smaller rivals find that costly to copy.
Evidence: Penetration went from 45% (31 Jan) to 51% (Q1) to 57% (Q2); above 70% in the US, and 75% among US under-18s Q2'26 letter. Five states have settled (AL, MS, NV, SD, WV), with $91M accrued in 1H26 Q2'26 10-Q, against $6.1B of cash and investments.
Monitor: Penetration; chat recovery; new state suits and settlement terms.

A4Operating leverage reaches shareholders after stock compattacked by B3

What must happen: Servers, R&D and headcount grow slower than bookings, so free cash flow after stock comp grows and buybacks shrink the share count rather than just offsetting grants.
Why Roblox specifically: Cash arrives long before revenue is recognized (27-month deferral), and creator pay rises and falls with bookings. The fixed costs are servers and engineers.
Evidence: In FY2025, bookings +55% drove FCF +111% to $1,353M FY25 10-K. Fully diluted shares are 752M, up 2% YoY. A $3B buyback was authorized in May 2026, with $380M done in Q2 Q2'26 letter.
Monitor: FCF minus stock comp; DevEx % of bookings; diluted shares.

A1The 2026 decline is a hit comparison, not audience erosionattacked by B2

What must happen: Bookings return to year-on-year growth once the 2025 viral-hit quarters roll off (Q3 2025: $1,922M, +70% Q3'25 letter; Q4 2025: $2,222M), with DAUs and payers still growing.
Why Roblox specifically: Engagement is spread across millions of creator-made games, so creators replace a fading hit without Roblox funding it. The top 10 games fell from ~30% of hours three years ago to ~20% in Q2 2026 Q2'26 letter.
Evidence: Q2 2026 DAUs +10% to 123M, monthly unique payers +15% to 27M, existing-user retention "stable" Q2'26 letter. Management blames the miss on lower spend per hour: play shifted away from high-monetizing 2025 hits, and the recommendation algorithm was retuned to favour retention (28-day window, up from 7).
Monitor: Q4 2026 bookings vs $2,222M; bookings per payer.

B22025 was a hit peak; spend per user is structurally flatSemi-permanentattacks A1

How it could fail: Even in a viral year, bookings per DAU barely moved ($52.7 → $53.5); all the growth came from more users. Management concedes the retention-first algorithm causes "a pretty immediate hit to dollars per hour" Q2'26 call; the payoff is only asserted. Monthly payers are down from a 35.8M peak (Q3 2025) to 27M, and bookings per payer fell roughly 5% YoY in Q2 (≈$58 vs ≈$61). Credibility is also at stake: February's FY2026 guide ($8.28–8.55B) would now need ~$3.34B in Q4 (+50% YoY) to reach even the low end.
What would confirm failure: Q4 2026 bookings below $2,222M; bookings per payer still falling in 2027 once the comparison eases.
Damage to economics: Growth becomes a user-count story, and new users now come mostly from outside the US and Canada, which have 82% of DAUs but only 43% of revenue Q2'26 10-Q. Falling bookings leave fixed server costs spread over less revenue; Q3 FCF is guided at $(60)M to $5M.
Permanent or fixable: Partly fixable through adult mix, but not quickly.

A2Age-checking unlocks the older, higher-spending cohortattacked by B4

What must happen: Adults rise as a share of DAUs and lift bookings per DAU above the ~$53 a year it sat at in both FY2024 ($52.7) and FY2025 ($53.5; bookings ÷ average DAUs).
Why Roblox specifically: Roblox has verified age data at this scale and can now route content and creator pay by age. From 8 June 2026, creators earn a 42% higher DevEx rate on spend by age-checked US adults (26.6% → 37.8%) Q1'26 letter; Q2'26 10-Q.
Evidence: Adults are 27% of age-checked DAUs globally and about a third in the US. US adult DAUs +32% and 18–34 DAUs +42%; US adults spend over 50% more than under-18s Q2'26 letter. Both figures are extrapolations.
Monitor: Adult share of age-checked DAUs; bookings per DAU.

B4The adult push is expensive and unprovenFixableattacks A2

How it could fail: The adult case rests on extrapolation: the age mix of users who have not age-checked is inferred from those who have Q1'26 letter. Winning adults means a 42% higher creator payout, paid user acquisition Q2'26 call, and partner-studio titles that only arrive through 2027.
What would confirm failure: Adult share of age-checked DAUs stuck near 27%; bookings per DAU flat even as adult DAUs grow.
Damage to economics: Each adult booking carries a higher creator payout, so adult growth can lift bookings while lowering profit per booked dollar.
Permanent or fixable: Fixable: incentives can be retuned, but at the cost of time.

A3Safety becomes a moat rather than a permanent taxattacked by B1

What must happen: Age-check penetration reaches the 90% target, in-game chat recovers above pre-age-check levels, and state claims settle for sums that are small next to cash flow.
Why Roblox specifically: Age estimation, text and voice moderation and age-based accounts (Kids, Select) all run inside Roblox's own stack. Smaller rivals find that costly to copy.
Evidence: Penetration went from 45% (31 Jan) to 51% (Q1) to 57% (Q2); above 70% in the US, and 75% among US under-18s Q2'26 letter. Five states have settled (AL, MS, NV, SD, WV), with $91M accrued in 1H26 Q2'26 10-Q, against $6.1B of cash and investments.
Monitor: Penetration; chat recovery; new state suits and settlement terms.

B1Safety liability becomes structuralPermanentattacks A3

How it could fail: Suits and regulation spread faster than the fixes. Ten states (AR, FL, IN, IA, KY, LA, NE, OK, TN, TX) and Los Angeles County have sued; Texas's deceptive-trade-practices claim survived dismissal Q2'26 10-Q. The settlements also carry costs that have not been accrued: public-service campaigns, law-enforcement liaisons, platform changes and liquidated damages.
What would confirm failure: More states file; settlements move from cash to conduct remedies that restrict features; age-check penetration stalls well short of 90%; chat does not recover.
Damage to economics: Every conduct remedy permanently taxes engagement among the 73% of age-checked users who are under 18 Q2'26 letter. Restricting social features breaks the play-with-friends loop that drives hours. Adjusted EBITDA adds settlements back ($91M in 1H26), hiding the cost.
Permanent or fixable: Mostly permanent: conduct remedies and damage to parents' trust do not reverse.

A4Operating leverage reaches shareholders after stock compattacked by B3

What must happen: Servers, R&D and headcount grow slower than bookings, so free cash flow after stock comp grows and buybacks shrink the share count rather than just offsetting grants.
Why Roblox specifically: Cash arrives long before revenue is recognized (27-month deferral), and creator pay rises and falls with bookings. The fixed costs are servers and engineers.
Evidence: In FY2025, bookings +55% drove FCF +111% to $1,353M FY25 10-K. Fully diluted shares are 752M, up 2% YoY. A $3B buyback was authorized in May 2026, with $380M done in Q2 Q2'26 letter.
Monitor: FCF minus stock comp; DevEx % of bookings; diluted shares.

B3Creators, compute and stock comp eat the leverageFixable, by choiceattacks A4

How it could fail: DevEx has risen from 17.5% of bookings (FY2020) to 22.1% (FY2025) and 23% (Q2 2026), and management intends to keep raising it Q2'26 call. Infrastructure and trust & safety costs rose 54% YoY in Q2, and $845M of new hosting commitments were signed in June Q2'26 10-Q. FY2025 stock comp was $1,129M against FCF of $1,353M, leaving FCF after stock comp of just $224M, about 3¢ per booked dollar FY25 10-K.
What would confirm failure: FY2026 FCF after stock comp turns negative; the share count stops falling even with the buyback running.
Damage to economics: Owner earnings stay near zero at scale, and the buyback becomes a cash cost of the pay plan rather than a return of capital.
Permanent or fixable: Fixable in principle, but the founder holds 58.8% of votes 2026 proxy, so outside pressure is weak.

Why the market might be missing it: GAAP revenue echoes bookings over ~27 months, and Adjusted EBITDA strips out deferrals, so neither shows the real turn. Bookings, the metric that does, faces the toughest comparison in the company's history. If users and payers keep growing through that comparison, the down quarters are a timing effect, not a verdict.

3. Bear case — Munger invert

The most likely way I lose money is that 2025 was the peak of a children's hit cycle, and the safety rebuild permanently makes the under-18 core less social and less willing to spend. Creators, AI servers and stock comp then absorb the operating leverage. Bookings per user stays stuck near $53 a year, and free cash flow per share never compounds.

Bear risks are shown beside their bull assumptions in Section 2. Switch back to “Bull, then bear” to read them ordered by severity.

B1Safety liability becomes structuralPermanentattacks A3

How it could fail: Suits and regulation spread faster than the fixes. Ten states (AR, FL, IN, IA, KY, LA, NE, OK, TN, TX) and Los Angeles County have sued; Texas's deceptive-trade-practices claim survived dismissal Q2'26 10-Q. The settlements also carry costs that have not been accrued: public-service campaigns, law-enforcement liaisons, platform changes and liquidated damages.
What would confirm failure: More states file; settlements move from cash to conduct remedies that restrict features; age-check penetration stalls well short of 90%; chat does not recover.
Damage to economics: Every conduct remedy permanently taxes engagement among the 73% of age-checked users who are under 18 Q2'26 letter. Restricting social features breaks the play-with-friends loop that drives hours. Adjusted EBITDA adds settlements back ($91M in 1H26), hiding the cost.
Permanent or fixable: Mostly permanent: conduct remedies and damage to parents' trust do not reverse.

B22025 was a hit peak; spend per user is structurally flatSemi-permanentattacks A1

How it could fail: Even in a viral year, bookings per DAU barely moved ($52.7 → $53.5); all the growth came from more users. Management concedes the retention-first algorithm causes "a pretty immediate hit to dollars per hour" Q2'26 call; the payoff is only asserted. Monthly payers are down from a 35.8M peak (Q3 2025) to 27M, and bookings per payer fell roughly 5% YoY in Q2 (≈$58 vs ≈$61). Credibility is also at stake: February's FY2026 guide ($8.28–8.55B) would now need ~$3.34B in Q4 (+50% YoY) to reach even the low end.
What would confirm failure: Q4 2026 bookings below $2,222M; bookings per payer still falling in 2027 once the comparison eases.
Damage to economics: Growth becomes a user-count story, and new users now come mostly from outside the US and Canada, which have 82% of DAUs but only 43% of revenue Q2'26 10-Q. Falling bookings leave fixed server costs spread over less revenue; Q3 FCF is guided at $(60)M to $5M.
Permanent or fixable: Partly fixable through adult mix, but not quickly.

B3Creators, compute and stock comp eat the leverageFixable, by choiceattacks A4

How it could fail: DevEx has risen from 17.5% of bookings (FY2020) to 22.1% (FY2025) and 23% (Q2 2026), and management intends to keep raising it Q2'26 call. Infrastructure and trust & safety costs rose 54% YoY in Q2, and $845M of new hosting commitments were signed in June Q2'26 10-Q. FY2025 stock comp was $1,129M against FCF of $1,353M, leaving FCF after stock comp of just $224M, about 3¢ per booked dollar FY25 10-K.
What would confirm failure: FY2026 FCF after stock comp turns negative; the share count stops falling even with the buyback running.
Damage to economics: Owner earnings stay near zero at scale, and the buyback becomes a cash cost of the pay plan rather than a return of capital.
Permanent or fixable: Fixable in principle, but the founder holds 58.8% of votes 2026 proxy, so outside pressure is weak.

B4The adult push is expensive and unprovenFixableattacks A2

How it could fail: The adult case rests on extrapolation: the age mix of users who have not age-checked is inferred from those who have Q1'26 letter. Winning adults means a 42% higher creator payout, paid user acquisition Q2'26 call, and partner-studio titles that only arrive through 2027.
What would confirm failure: Adult share of age-checked DAUs stuck near 27%; bookings per DAU flat even as adult DAUs grow.
Damage to economics: Each adult booking carries a higher creator payout, so adult growth can lift bookings while lowering profit per booked dollar.
Permanent or fixable: Fixable: incentives can be retuned, but at the cost of time.

4. Signals to monitor

Bookings YoY

+8% (Q2'26); Q3 guide −14% to −18%

Direction that mattersBack to growth by 1H27 as comparisons ease

Q2'26 lettertracks A1 B2

Bookings per DAU (quarter)

$12.66 vs $12.86 (Q2'25)*

Direction that mattersRising, with adults as the driver

Q2'26 letter; Q2'25 10-Qtracks A2 B2

Bookings per monthly payer

≈$58 vs ≈$61 (Q2'25)*

Direction that mattersStops falling by Q2'27

Q2'26 lettertracks A1 B2

Age-check penetration

57% global; US >70%; target 90%

Direction that mattersUp, with chat back above pre-age-check levels

Q2'26 lettertracks A3 B1

Adult share of age-checked DAUs

27% global; ~1/3 US

Direction that mattersRising above 30%

Q2'26 lettertracks A2 B4

DevEx % of bookings

23% (Q2'26) vs 22% (Q2'25)

Direction that mattersFlat or lower

Q2'26 lettertracks A4 B3

FCF minus stock comp

FY25 $224M; 1H26 $333M*

Direction that mattersUp; diluted shares falling from 752M

FY25 10-K; Q2'26 lettertracks A4 B3

State suits / settlement accruals

10 states + LA County suing; 5 settled; $91M accrued 1H26

Direction that mattersNo new filings; no feature-restricting remedies

Q2'26 10-Qtracks A3 B1

Show as table

Signal	Current value	Source	Direction that matters	Tracks
Bookings YoY	+8% (Q2'26); Q3 guide −14% to −18%	Q2'26 letter	Back to growth by 1H27 as comparisons ease	A1 · B2
Bookings per DAU (quarter)	$12.66 vs $12.86 (Q2'25)*	Q2'26 letter; Q2'25 10-Q	Rising, with adults as the driver	A2 · B2
Bookings per monthly payer	≈$58 vs ≈$61 (Q2'25)*	Q2'26 letter	Stops falling by Q2'27	A1 · B2
Age-check penetration	57% global; US >70%; target 90%	Q2'26 letter	Up, with chat back above pre-age-check levels	A3 · B1
Adult share of age-checked DAUs	27% global; ~1/3 US	Q2'26 letter	Rising above 30%	A2 · B4
DevEx % of bookings	23% (Q2'26) vs 22% (Q2'25)	Q2'26 letter	Flat or lower	A4 · B3
FCF minus stock comp	FY25 $224M; 1H26 $333M*	FY25 10-K; Q2'26 letter	Up; diluted shares falling from 752M	A4 · B3
State suits / settlement accruals	10 states + LA County suing; 5 settled; $91M accrued 1H26	Q2'26 10-Q	No new filings; no feature-restricting remedies	A3 · B1

* My calculations from reported bookings, DAUs, payers, operating cash flow, capex and stock comp; Q2'25 payers implied from the reported +15% growth, so per-payer figures are approximate.

6. Bottom line

Why it could workRBLX could work because it owns a platform where creators fund the content and users pay cash up front, it already takes nearly 4% of global gaming spend, and its users and payers kept growing through 2026's hit comparison.
What must go rightBookings must return to growth in 2027 as spend per hour recovers and age-checked adults lift bookings per user above the ~$53-a-year ceiling, while DevEx, AI servers and stock comp grow slower than bookings.
How it breaksThe thesis breaks if youth-safety litigation and the age-check rebuild permanently reduce how much the under-18 core plays and spends, turning safety into a tax rather than a moat.
What changes my mindI would change my mind if bookings per payer keeps falling after the comparison eases in 2027, if new state cases settle with feature restrictions, or if FCF after stock comp stays near zero through 2027.

Section 5 (External challenge) is added only on request.

Sources

Annual reports 2

FY25 10-K — Form 10-K for FY2025, filed 11 Feb 2026 (bookings, FCF, stock comp, DevEx, average DAUs)
FY2021–FY2024 10-Ks — bookings and DevEx history (FY2020 from FY2021 10-K comparatives); FY2024 average DAUs

Quarterly filings 2

Q2'26 10-Q — filed 30 Jul 2026 (legal proceedings, settlements, DevEx rate change, hosting commitments, geographic mix)
Q2'25 10-Q — Q2 2025 DAUs (111.8M)

Shareholder letters (8-K Ex. 99.1) 4

Q2'26 letter — 30 Jul 2026 (Q2 metrics, Q3 guidance, age-check data, buyback)
Q1'26 letter — 30 Apr 2026 (age-check extrapolation note, O18 DevEx 26.6% → 37.8%)
Q4'25 letter — 5 Feb 2026 (FY2026 guidance, 45% age-check penetration)
Q3'25 letter — 30 Oct 2025 (Q3 2025 bookings, 35.8M payers)

Transcripts and governance 2

Q2'26 call — earnings call transcript, 30 Jul 2026
2026 proxy — DEF 14A, 16 Apr 2026 (founder voting power 58.8%)