← back to rankingRBLX · Roblox Corporation
Electronic Gaming & Multimedia · mkt cap $32.2B · calls: Q1 FY2026 vs Q4 FY2025
62.0 conviction · conf-adj 62
conf 2/10 partial
enthusiasm:27.0 · trend:8 · quantifies:0 · impact:0 · under_radar:5 · credibility:12 · business_impact:10 · disruption:-6 · commitment:6 · confirmation:0
Enthusiasm latest 9 / prev 8 (rising)
Roblox’s AI thesis is that proprietary models, creator copilots/agents, 3D generation, NPC testing, discovery, safety, translation, and Roblox Reality can accelerate creation, improve recommendations, and expand gaming into broader entertainment. Credibility improved in Q1 FY2026 because management added concrete creator adoption, inference-scale, proprietary-model, and monetization language, though most financial impact remains indirect rather than tied to revenue or margin dollars.
GROUNDED NEXT-FY IMPACT vs CONSENSUS
Grounded on actual base — revenue $4.9B · net income $-1.1B · net margin -21.8% · diluted EPS -1.54
These are next-fiscal-year annual uplift estimates, not next-quarter numbers.
Aggregate next-FY est. rev uplift: % · next-FY EPS uplift: % · vs analysts: unclear · priced in: medium (model's call-read: high; verdict above is the hard-data one used for ranking) · confidence: 2/10
| Claim | Figure | Arithmetic | Next-FY Rev % | Next-FY EPS % |
|---|
~Half of top 1,000 creators use Roblox Assistant/MCP
productivity · soft | ~50% of top 1k creators (~500) | ~0.5*1,000=~500 top creators using a dev-productivity tool; no disclosed revenue per top creator, creator-revenue concentration, or timeline-to-revenue, so $ impact cannot be sized vs $4.891B revenue. 'Compress dev timelines' unquantified. | | |
400+ models, 1.5M inferences/sec
other · soft | >400 models; >1.5M inf/sec (~47.3T/yr) | 1.5M/sec*31,536,000=47.304T inferences/yr; infra-scale metric on internal (on-prem + own cloud) compute, not sold to third parties. No disclosed inference cost, avoided spend, or revenue rate. | | |
400+ internal AI models (Q4)
other · soft | >400 models | 400+ internal models disclosed, but no revenue, savings, or productivity dollars per model; no calculable uplift. | | |
30,000 years of interaction data/day
other · soft | ~30k yrs/day (~10.95M yrs/yr) | 30,000*365=10.95M years/yr of interaction data; a training-data/moat asset with no disclosed monetization or cost-saving rate. | | |
AI drove double-digit rise in unique experiences surfaced
engagement · soft | double-digit % (>=10%) | % lift is on COUNT of experiences surfaced in 'recommended for you', not bookings/revenue; no disclosed base count or surfaced-to-revenue conversion. Soft engagement signal. | | |
24+ unique experiences/user/month, up double digits
engagement · soft | >24/mo, +double digits (>=~+2.18) | At 10%, 2024 base<=24/1.10=21.82, so increase>=2.18/mo; diversity-of-engagement metric with no take-rate or per-experience revenue disclosed. Soft. | | |
13B hours of 3D/video data for NPC training
other · soft | 13B hours | Data-asset scale for future NPC models; no disclosed revenue uplift, engagement conversion, or monetization rate. | | |
Slight CapEx uptick for in-house GPUs / memory inflation
cost · soft | 'slight uptick' in CapEx | A HEADWIND, not a saving, and unquantified ('slight'). Hits cash flow/later D&A, not a sizable next-FY EPS delta. No $ given. | | |
AI workloads embedded in 2026 margin guidance
cost · soft | 'investments' | Explicitly already inside management's margin guidance (hence in consensus); no dollar cost, margin rate, or opex/depreciation split. A cost, not a saving. | | |
Roblox Reality will be paid (offset real-time inference)
revenue · soft | 'will not be free'; subscribe/pay | Pre-launch product monetized to offset its own inference cost (cost-recovery). No price, subscribers, usage, launch date, or TAM; monetizes RBLX's own product (adopter), not compute sold to others. Cannot size. | | |
Assumptions: Current revenue base $4.891551B; net income -$1.065057B (net margin -21.78%), so EPS uplift % set null under the loss-making-company guardrail and flowing any hypothetical incremental revenue through a negative margin is meaningless. Default tax 21% and default incremental margin=current net margin would apply, but no claim supplies a calculable dollar revenue or cost-saving base. 'Double digit' read as 10% for engagement arithmetic only, not revenue conversion. No phasing assumed (no figure to phase). All claims map to RBLX's own platform = adopter-side.
Top line: Unquantifiable from the disclosures. Every topline AI claim is a count or engagement ratio (>400 models, 1.5M inf/sec ~47.3T/yr, double-digit rise in surfaced experiences, >24 experiences/user/mo, ~500 of top 1k creators, 13B training hours) with no disclosed take-rate, per-unit revenue, or dollar base. Directionally supportive of consensus' strong FY26 growth but no claim attributes a specific point of it to AI.
Bottom line: Net-negative-to-neutral and already in guidance. The only quantified bottomline items are a 'slight uptick' in CapEx for in-house GPUs (amid memory inflation) and unspecified 'AI workloads' — both costs, both explicitly baked into 2026 margin guidance, both unquantified in dollars. RBLX is loss-making (NI -$1.065B), so no EPS-uplift % is computable. Roblox Reality is positioned as cost-recovery for its own inference, not a sized profit driver.
[impact n/m (all claims soft/unanchored); EPS uplift n/m (loss-making base)] Consensus already models FY26 revenue of $7.512B (up from $6.616B in 2025, a +$896M / +18.36% step on the $4.891B base) rising to $8.890B in 2027, while keeping net income negative (-$1.070B 2026, -$879M 2027; EPS ~-$1.50) — i.e. it bakes in continued heavy AI/engagement investment with no near-term profit inflection. Management's AI claims supply zero incremental dollar figures (aggregate rev uplift null, EPS uplift null), and the only quantified bottomline item is stated to be inside the same margin guidance consensus reflects. The AI story is qualitatively priced in.
MODEL CONSENSUS (impact)
partial
Math, verdicts (priced_in high, unclear, adopter) fully agree; only aggregate-pct representation and confidence differed.
Conflicts reconciled
- est_rev_uplift_pct: X=null vs Y=0 -> used null because every claim is unsizeable (cannot compute, not a measured zero)
- supplier_rev_uplift_pct: X=null vs Y=0 -> used null because side is adopter, supplier uplift is N/A
- confidence: X=2 vs Y=3 -> used 2 (more conservative; avg 2.5 rounded down)
| Field | Opus 4.8 | GPT-5.5 |
|---|
| Rev uplift % | – | 0 |
| EPS uplift % | – | – |
| Priced in | high | high |
| vs analysts | unclear | unclear |
| Confidence | 2 | 3 |
| Top line | Unquantifiable from the disclosures. Every topline AI claim is a count or an engagement ratio (>400 models, 1.5M inferences/sec, double-digit rise in surfaced experiences, 24+ experiences/user/mo, 13B training hours) with no disclosed take-rate, per-unit revenue, or dollar base to convert into incremental revenue. Directionally supportive of consensus' strong FY26 revenue growth (+53.6% to $7.512B off the $4.89B base) but no claim lets you attribute a specific point of that growth to AI. | No quantified next-FY revenue uplift can be calculated. The strongest adoption indicators are ~500 of the top 1,000 creators using AI tools, 47.304T annualized inferences at the stated 1.5M/sec run-rate, and at least +2.18 unique experiences/month if 'double digit' is read as 10%, but none discloses a revenue conversion rate. |
| Bottom line | Net-negative-to-neutral and already in guidance. The only quantified bottomline items are a 'slight uptick in CapEx' for in-house GPUs (amid memory-price inflation) and unspecified 'AI workloads' — both costs, both explicitly baked into 2026 margin guidance, and both unquantified in dollars. RBLX is loss-making (NI -$1.065B), so no EPS-uplift % is computable (guardrail). Roblox Reality is positioned to charge users to 'offset' its own inference cost, i.e. cost-recovery, not a sized profit driver. | Bottom-line impact is unquantified and may be a near-term headwind: management cites AI workload investments and a slight CapEx uptick for GPUs/memory, while Roblox Reality is intended to offset inference cost through paid/subscription economics. Because current net income is negative, EPS uplift percentages are not meaningful. |
| Reasoning | Consensus FY26 already models revenue of $7.512B (+53.6% over the $4.89B FY25 base) while keeping the loss essentially flat (NI -$1.070B vs -$1.065B; EPS -$1.50 vs -$1.54) — i.e. it bakes in continued heavy AI/engagement investment with no near-term profit inflection. Management's AI claims provide zero incremental dollar figures: aggregate adopter rev uplift = null, EPS uplift = null. The only quantified bottomline item (CapEx uptick + AI workloads) is stated to be inside the same margin guidance consensus reflects. With nothing in the math pointing measurably above the consensus trajectory, the AI story is qualitatively priced in. | Consensus already models revenue rising from $6.616B in 2025 to $7.512B in 2026, a $896.133M increase or 18.36% of the $4.891B current revenue base, then to $8.890B in 2027. The AI claims provide no hard incremental revenue dollars above that trajectory and no hard savings; consensus also keeps net income negative at -$1.070B in 2026 and -$879.280M in 2027, consistent with AI investment being absorbed rather than visibly accretive. |
Rows highlighted where the two models disagreed.
QUANTIFICATIONS
creator AI tool adoption: Nearly half of the top 1 thousand creators (Q1 FY2026, topline)
“Nearly half of the top 1 thousand creators on Roblox now leverage either Roblox Assistant—our own AI entry point—or MCP (model context protocol) to compress dev timelines.”
AI models and inference volume: over 400 models; over 1.5 million inferences per second (Q1 FY2026, both)
“In general, on the AI side, to wrap up, we have over 400 models running over 1.5 million inferences per second on-prem and on our cloud.”
AI models: over 400 AI models (Q4 FY2025, both)
“I want to highlight that we are internally now running over 400 AI models.”
AI training data: roughly 30,000 years of human interaction data (Every day, both)
“Every day, we capture roughly 30,000 years of human interaction data on Roblox Corporation in a PII and privacy-compliant way.”
AI discovery impact: double-digit increase (Q4 FY2025, topline)
“In discovery, AI has been driving personalization. In Q4, AI drove a double-digit increase in the number of unique experiences that are surfaced in our recommended for you algorithm.”
user experience diversity: over 24 unique experiences per month; up double digits from 2024 (2025, topline)
“In 2025, on average, users engage with over 24 unique experiences per month. This is up double digits from 2024.”
AI/data for NPC training: 13 billion hours (Q4 FY2025, topline)
“training NPCs with the 13 billion hours of not just video data, but intrinsic 3D world data.”
AI infrastructure investment: slight uptick in CapEx spend (2026, bottomline)
“That includes a slight uptick in CapEx spend relative to last year, as we are continuing to land GPUs in our own data centers and also navigating the recent inflation in memory prices.”
AI workloads in margin guidance: investments (2026, bottomline)
“It incorporates investments that we have talked about related to continued growth in users and engagement and also AI workloads.”
Roblox Reality inference economics: will not be free; subscribing or paying (future launch, both)
“When we launch this, it will not be free. This will use cloud compute. We will have some kind of way of subscribing or paying for this, and because of that, we think we will offset the real-time inference side of it.”
PAST (realized)
- Last year, we enhanced the discovery of long-tail content and improved content diversity.
- This week, we launched 4D generation, which allows experiences to include creation of new objects simply by users prompting and creating by AI.
- In discovery, AI has been driving personalization. In Q4, AI drove a double-digit increase in the number of unique experiences that are surfaced in our recommended for you algorithm.
CURRENT (now)
- We continue to see an acceleration of AI toolchain use by our top creators.
- Nearly half of the top 1 thousand creators on Roblox now leverage either Roblox Assistant—our own AI entry point—or MCP (model context protocol) to compress dev timelines.
- In general, on the AI side, to wrap up, we have over 400 models running over 1.5 million inferences per second on-prem and on our cloud.
- This powers, in addition to creation and Roblox Reality, everything from discovery recommendations to communication safety, marketplace recommendations, and 3D generation.
- I want to highlight that we are internally now running over 400 AI models.
FORWARD (guidance)
- We believe AI will fundamentally accelerate gaming, and we are in a unique position to lead in this transformation.
- We are now investing in four in-house proprietary models for 3D generation, NPC behavior, our newly shared video super upsampler initiative, and coding assistance and generation.
- We want to get to the point where devs can have Roblox Studio working overnight on their game and come back in the morning and see those improvements.
- The goal of Roblox Reality is to enable creators to construct interactive environments that are high fidelity and drive this new technical frontier—the introduction of the first easy-to-use multiplayer photorealistic platform that we are uniquely poised to build.
- And the pace of innovation in AI is a tremendous accelerant and has the potential to help us grow beyond gaming as we develop even deeper connections into communication, entertainment, and commerce.
TRACK RECORD — PROMISE vs DELIVERY
82/100 track record delivers 6 calls reviewed
Roblox reliably delivered its specific AI product milestones (text generation, Cube 3D, in-experience and 4D generation) on or near the promised quarter. The only blemishes are safety/age-estimation and multiplayer rollout claims that were directionally achieved but slipped past their dates or lacked full confirmation.
Ship in-experience LLM-powered text generation for AI characters/NPCs in Q1 FY2025 — promised Q4 FY2024
delivered Q1 FY2025 reported a live beta text-generation API using an LLM that creators could use directly in experiences to power conversational NPCs.
Make Cube 3D foundational model available in Studio (Q1 FY2025) — promised Q4 FY2024
delivered Q1 FY2025 reported Cube 3D was live in Studio, trained on 1.5M+ licensed/public datasets, with creators experimenting with generative 3D.
Extend 3D creation/generation into live experiences by Q2 FY2025 — promised Q4 FY2024
delivered Q2 FY2025 reported Cube 3D was live within experiences, with 1M+ 3D models generated in games and Studio.
Enable 4D (functional/interactive) generation over the next few quarters — promised Q1 FY2025
delivered Q4 FY2025 reported 4D generation had launched that week, letting experiences create new interactive objects by prompt.
Use AI facial age estimation to estimate the age of everyone on the platform by end of 2025 — promised Q3 FY2025
partial Rollout began late 2025 and completed globally in January 2026 at ~45-51% global DAU penetration — shipped but slipped past year-end and not universal.
Bring Cube 3D live in multiplayer mode for everyone over the next 1-2 quarters — promised Q3 FY2025
partial Q4 FY2025/Q1 FY2026 advanced 4D and in-experience creation but did not clearly confirm multiplayer-for-everyone within the promised window.
PRICED-IN (REFINED)
MEDIUMEst. revisions falling · Fwd P/E -29.1 · EV/Sales 6.0x
Estimate momentum is falling: recent rating counts show fewer strong-buy/buy ratings than prior months, and price-target averages have moved down from last year to last quarter to last month. Consensus still bakes in strong forward revenue growth and narrowing losses, but the stock remains loss-making, making the negative forward P/E less useful. EV/Sales around 6.0x is rich for a mature company, so valuation already embeds meaningful growth despite weaker revisions. With no usable segment data and a rich valuation offset by falling estimate momentum, the AI upside looks partially but not fully priced in.
COVERAGE — ENTHUSIASM TRAJECTORY + CATALYSTS
8Q4 FY20247Q1 FY20258Q2 FY20259Q3 FY202510Q4 FY20258Q1 FY2026
AI enthusiasm across 6 calls — trend ↗ rising
AI evolved from moderation, discovery, and creator tools into core generation, safety, NPC, world-model, and creator-productivity infrastructure.
RECENT AI CATALYSTS & NEWS
BUSINESS IMPACT - QUALITATIVE MATERIALITY
9/10 qualitative impact transformational medium-term · mixed evidence
Where AI matters: creator tools, discovery, safety, 3D generation
AI is being embedded into Roblox's core platform loop: creator productivity, 3D/4D generation, NPC testing, recommendations, safety, translation, and potential expansion beyond gaming. The evidence is operationally concrete, including top-creator adoption and massive inference scale, but revenue and margin attribution remain mostly unproven.
Caveats: No disclosed AI-driven revenue or EPS uplift; AI infrastructure may raise CapEx and inference costs before monetization catches up; Generated content abundance could worsen discovery, moderation, and quality-control burdens; Competing AI game-creation tools could reduce Roblox Studio's uniqueness
AI DISRUPTION / CANNIBALIZATION RISK two-sided · 4/10
AI could commoditize game/content creation and enable competing AI-native creation platforms, pressuring creator economics and content differentiation. Roblox's social graph, distribution, youth safety stack, payments, and marketplace make the model more durable than standalone content businesses, so the threat is real but not dominant.
OPTIONS / MARKET STRUCTURE
option liquidity: good
proxy inputs — dollar-ADV $485M · beta 1.5 · px $45.02
source: proxy (no options chain on FMP)
FMP /stable/ exposes no options-chain endpoint on this key, so ATM IV, bid-ask spread and open interest are unavailable. Liquidity below is a PROXY from dollar-ADV, beta and price level (a stand-in for option depth), not measured option-market data.
CONFIRMATION — INSIDERS · 13F · LANGUAGE
Mixed — insiders selling, institutions flat, management language 7/10 committed.
INSIDERS selling 63 open-market sell(s) vs 0 buy(s) — net distribution
INSTITUTIONS (13F) flat as of 2026-03-31: 111 new / 225 closed positions; 401 increased / 313 reduced; institutional ownership +1.64pp; -127 net 13F holders
MGMT LANGUAGE 7/10 committed Substantial AI discussion with concrete usage and infrastructure metrics, but strategic impact remains partly framed as vision and belief.
commit “Nearly half of the top 1 thousand creators on Roblox now leverage either Roblox Assistant—our own AI entry point—or MCP”
commit “Creators can now engage in focused conversations with our Assistant about the design, implementation, and test plan of new features.”
commit “we have over 400 models running over 1.5 million inferences per second on-prem and on our cloud.”
VERBATIM AI QUOTES
“We continue to see an acceleration of AI toolchain use by our top creators. Nearly half of the top 1 thousand creators on Roblox now leverage either Roblox Assistant—our own AI entry point—or MCP (model context protocol) to compress dev timelines.”
— David Baszucki, Q1 FY2026
“We want to do for game creation what tools like Codex, Cursor, and Quod Code are doing for coding—radically accelerating speed—and we believe we are uniquely poised to innovate and solve this given the integrated architecture we have from our cloud to our engine to our developer tools to our native AI.”
— David Baszucki, Q1 FY2026
“In April, we called that the month that Roblox Studio went agentic.”
— David Baszucki, Q1 FY2026
“Assistant executes a plan, launches a suite of building agents to test, and delivers new features with minimal creator engagement.”
— David Baszucki, Q1 FY2026
“We believe NPC testing agents can be part of a foundational model that we are building to help creators iterate quickly on their games using NPCs in addition to humans for testing.”
— David Baszucki, Q1 FY2026
“All of this is to enable small teams to produce super high-quality content very quickly.”
— David Baszucki, Q1 FY2026
“We believe AI will fundamentally accelerate gaming, and we are in a unique position to lead in this transformation.”
— David Baszucki, Q1 FY2026
“In general, on the AI side, to wrap up, we have over 400 models running over 1.5 million inferences per second on-prem and on our cloud.”
— David Baszucki, Q1 FY2026
“This powers, in addition to creation and Roblox Reality, everything from discovery recommendations to communication safety, marketplace recommendations, and 3D generation.”
— David Baszucki, Q1 FY2026
“We are now investing in four in-house proprietary models for 3D generation, NPC behavior, our newly shared video super upsampler initiative, and coding assistance and generation.”
— David Baszucki, Q1 FY2026
“We have an enormous amount of high-fidelity AI training data.”
— David Baszucki, Q4 FY2025
“This week, we launched 4D generation, which allows experiences to include creation of new objects simply by users prompting and creating by AI.”
— David Baszucki, Q4 FY2025
“In addition to this AI future, AI is really driving creation, safety, discovery, translation, in addition to these potential areas of user growth on our platform.”
— David Baszucki, Q4 FY2025
“Every day, we capture roughly 30,000 years of human interaction data on Roblox Corporation in a PII and privacy-compliant way.”
— David Baszucki, Q4 FY2025
“We are actively using this data to develop and train AI models that continue to bring our vision to life.”
— David Baszucki, Q4 FY2025
“I want to highlight that we are internally now running over 400 AI models.”
— David Baszucki, Q4 FY2025
“In discovery, AI has been driving personalization. In Q4, AI drove a double-digit increase in the number of unique experiences that are surfaced in our recommended for you algorithm.”
— David Baszucki, Q4 FY2025
“And the pace of innovation in AI is a tremendous accelerant and has the potential to help us grow beyond gaming as we develop even deeper connections into communication, entertainment, and commerce.”
— Naveen Chopra, Q4 FY2025
ANALYST QUESTIONS ON AI
Q (Q1 FY2026, Matthew Andrew Cost): And second, on Roblox Reality, are there any increased cloud expenses or CapEx that you are expecting to support that going forward?
A: When we launch this, it will not be free. This will use cloud compute. We will have some kind of way of subscribing or paying for this, and because of that, we think we will offset the real-time inference side of it. On the training side, we may have additional needs.
Q (Q1 FY2026, Andrew Marroque): Could you give a bit more detail on the shift in discovery algorithms you mentioned—going from shorter-term monetization factors to longer-term health factors? How might that be evidenced from a user perspective, and how does that encourage increased chat density over time?
A: We have made great leaps in discovery using ML to predict many factors for each user-game pair. Now that we have developed this capability, and given the potential for 18 and up, we determined we want to drive user growth of that segment, even if we slightly less weight short-term monetization.
Q (Q4 FY2025, Kenneth Gawrelski): And then the second one, maybe just could you elaborate a little bit more on the you said the increased the better opportunity to target the 18 plus. And maybe within that kind of if you could please if you could please talk a little bit about, you know, the recent developments with Genie and AI gaming. Just how does that inform your view? And help hurt you know, how are you going to use AI to make the platform better for developers?
A: we're optimistic we're going to see an expansion of the definition of really what is gaming, and AI is going to power that.
Q (Q4 FY2025, Matthew Cost): I think the detail on the blog post that you've shared so far on the call about the work you're doing with AI and world models is incredibly helpful. And I want to put a finer point on the concerns that we've seen coming up in the market over the past week about the potential for, you know, disruption to your business from other, you know, advances from with AI coming from other companies out there in the space. And I wonder if you could respond to those concerns and, you know, what you see as the differentiating factor that protects Roblox Corporation.
A: I would flip it and share how we think about it internally, which is an opportunity for disruption in the opposite direction. Which is an opportunity for the expansion of the Roblox Corporation vision beyond gaming into the future mix of what is entertainment and where does gaming end and where do other points begin.