Roblox RBLX

Reference table

1. Data integrity — three problems in this file, and they must come first

This is one of the more compromised files in the batch and the reader needs the caveats before the analysis, not after.

(a) REJECTED — the entire consensus EBITDA series. The estimate block reports negative ebitdaAvg in every forecast year: -$1.771B (FY26E), -$2.032B (FY27E), -$2.396B (FY28E), -$3.028B (FY29E), -$3.389B (FY30E) — a figure that grows more negative every year while the same rows show netIncomeAvg turning positive at +$179.3M in FY30. A company cannot have EBITDA of -$3.39B and net income of +$179M in the same year. The series is internally impossible. No EBITDA-based multiple appears anywhere in this dive. This same failure mode appears in several other files in this batch and looks systematic rather than company-specific.

(b) FLAGGED — the estimate revenue series is on a DIFFERENT BASIS from reported GAAP revenue, and the earnings calendar switches between them mid-series. This is subtle and it invalidates the most obvious comparison an investor would make.

• Consensus FY25 revenue was $6.616B. Actual FY25 GAAP revenue was $4.891B. The estimate exceeds the actual by 35%.
• Consensus FY26 revenue is $7.156B. Actual first-half 2026 GAAP revenue is $2.911B ($1.442B + $1.469B), which annualises to roughly $6.0B. The estimate again sits far above the GAAP run-rate.
• The earnings-calendar revenueActual field switches basis mid-series. For 2025-07-31 it reports $1,080,677,000 and for 2025-10-30 $1,359,646,000 — matching GAAP quarterly revenue to the dollar. But for 2026-02-05 it reports $2,222,275,000 against GAAP revenue of $1,415,021,000; for 2026-04-30 $1,731,000,000 against $1,442,000,000; and for 2026-07-30 $1,557,000,000 against $1,469,000,000. The last three do not match GAAP at all.

The most likely explanation is that the estimate and recent-calendar series are on a bookings basis — the gross cash value of virtual-currency purchases, which for this business model is recognised as revenue over an extended period and therefore substantially exceeds GAAP revenue. That is a standard metric for this company type. But whatever the label, the practical consequence is unambiguous: do not compare consensus revenue to reported revenue, and do not compute EV/sales using the estimate series. Every revenue multiple in this dive uses reported GAAP revenue and says so.

(c) FLAGGED — segment and geographic disclosure are entirely absent. seg_prod and seg_geo are both empty. For a platform business whose economics depend on regional mix, age-cohort mix and the split between platform revenue and advertising, this is a substantial gap. We have no geographic revenue split, no user metrics, no daily-active-user figure, no bookings-per-user, no age-demographic disclosure. For a company where the bear case most plausibly involves regulation or user safety in specific jurisdictions, the absence of any geographic data is exactly the wrong gap to have.

(d) DO NOT MISREAD — the leverage and book-value ratios. The vendor reports debtToEquityRatioTTM of 12.08, priceToBookRatioTTM of 172.9 and currentRatioTTM of 0.82. Read naively these describe a distressed, over-levered company. They describe nothing of the kind. Shareholders' equity is only $394.5M because the balance sheet carries a -$5.061B accumulated deficit and $4.169B of deferred revenue sitting in current liabilities. Recomputing by hand from the balance sheet: cash $1.205B + short-term investments $1.850B − total debt $1.803B = $1.252B of NET CASH. The current ratio of 0.82 is caused by deferred revenue — an obligation to deliver a service, not to pay cash — being classified as a current liability. This company has no leverage problem. We flag this specifically because these are exactly the ratios a screen would flag as distress.

CHECKED AND CLEAN — the most recent quarter. inc_q[0] (Q2 2026, 2026-06-30) passes the standard screen: net income -$183M against 716.767M shares gives -$0.255 versus the reported -$0.26; share counts do not duplicate the prior quarter (716.767M against 711.697M) and continue a steady upward trend; interest expense of $10M is consistent with every prior quarter; the reported EPS matches the earnings-calendar actual of -$0.26 exactly. The quarter is usable.

2. What the operating numbers actually say

Annual record:

Revenue growth has ACCELERATED for three consecutive years — 25.8%, 28.7%, 35.8% — while gross margin expanded every year. Free cash flow has gone from negative to $1.353B in three years. These are not the financials of a business in trouble.

Quarterly record — where the margin improvement shows:

Read the last row against the row four above it. Q2 2025: revenue $1,080.7M, operating loss $322.5M, margin -29.8%. Q2 2026: revenue $1,469.0M (+35.9%), operating loss $229.0M (a $93.5M reduction in absolute dollars), margin -15.6%. The company grew revenue 36% and simultaneously cut its operating loss by 29% in absolute terms. That is a 14.3-percentage-point margin improvement in twelve months, and it is the single strongest operating fact in this dive.

Four consecutive earnings beats. The most recent (2026-07-30) came in at -$0.26 against a -$0.344 consensus — a 24% beat. And the stock is 3% off its 52-week low.

Deferred revenue rose from $3.005B (FY24) to $4.169B (FY25), +38.7% — faster than revenue growth. This is cash already received against future delivery. Customers are prepaying at an accelerating rate, which is difficult to reconcile with any thesis of user abandonment.

3. Balance sheet and cash generation

Balance sheet (FY25, 2025-12-31):

• Cash $1.205B + short-term investments $1.850B = $3.055B
• Total debt $1.803B ($14.7M short-term, $993.1M long-term, the balance in other obligations)
• Net cash $1.252B (my computation; the vendor's netDebt of +$597.4M nets cash only and ignores $1.85B of short-term investments)
• Deferred revenue $4.169B — the largest single liability, and an operating one rather than a financial one
• Accumulated deficit -$5.061B; shareholders' equity $394.5M
• Enterprise value: $26.196B market cap − $1.252B net cash = $24.94B (my computation; the vendor reports $27.04B using the cash-only figure)

Cash generation is the core of the investment case:

• FY25 operating cash flow $1.796B, capex $443.5M, free cash flow $1.353B
• Free-cash-flow margin 27.7% — higher than most profitable software companies
• EV/free cash flow: 18.4x. Free-cash-flow yield on market cap: 5.2%; the vendor's yield on its own enterprise-value basis is 6.3%
• The gap between a -$1.065B net loss and +$1.353B of free cash flow is bridged by $1.129B of stock compensation plus the $1.164B increase in deferred revenue — i.e. non-cash expense plus customer prepayment. Both are real; neither is an accounting trick

The dilution caveat. Diluted shares have grown from 642.8M (Q2 2024) to 716.8M (Q2 2026) — +11.5% in two years, roughly 5.5% annually. There is no buyback (commonStockRepurchased is $0 in every year on file) and net stock issuance is positive. A shareholder's claim on that $1.353B of free cash flow is shrinking by roughly 5% a year. With $1.25B of net cash and a share price at $36.67, a buyback would be highly accretive; management has not initiated one, and that is a legitimate criticism.

4. The drawdown — what we can and cannot say

The facts: $36.67 against a 52-week high of $141.56 — -74.1%. Twelve-month return -73.4% against SPY's +19.9% and QQQ's +23.9%, a 93-point relative deficit. Three-month return -18.7%. The stock is 3.0% above its 52-week low of $35.60. RSI 23.4. MACD -2.42. Both moving averages are falling and price is 47% below the 200-DMA.

A minor data inconsistency: the profile block reports a 52-week range of $33.88-$142.00 while the technical block reports $35.60-$141.56. The conclusion is unaffected.

What this file does NOT contain, and what we therefore cannot assess: any news, regulatory filing, safety or child-protection development, litigation, competitive entry, management guidance, user metric, or analyst commentary. There is no field in this dataset that explains a 74% decline.

What we can rule out from the data available:

• Not a revenue break. Revenue grew 39.3% and 35.9% in the two most recent quarters.
• Not a margin break. Operating margin improved 14.3 points year over year in the most recent quarter.
• Not a cash break. Free cash flow more than doubled in FY25 and the balance sheet holds net cash.
• Not an earnings miss. The last four prints all beat consensus, the most recent by 24%.
• Not a demand break at the customer level. Deferred revenue grew 38.7%, faster than revenue.
• Not, principally, a sell-side capitulation. The consensus target remains $49.47, 34.9% above spot, with 16 buy ratings — though the rating has moved to Hold (18 holds, 3 sells), so there has been partial capitulation.

What remains as candidate explanations, none verifiable here: a regulatory or child-safety development (the most commonly cited structural risk for this business model, and one that would not appear in any financial field); a guidance cut for periods beyond the estimate horizon in this file; a change in the bookings-to-revenue relationship; a broad de-rating of high-multiple growth names; or forced selling.

The intellectually honest position: we are looking at a large, quantifiable dislocation between price and reported operating performance, and we do not know its cause. That is a reason to take a position, because the arithmetic is compelling. It is equally a reason to take a small one, because an unexplained 74% decline usually has an explanation. Every element of the verdict — Stage-In rather than Buy, 1.5-2.5% rather than 3-5%, three tranches rather than one — follows from that single sentence.

5. Knowledge base — bullish, narrow, half-weighted, and silent on the bear case

Nine claims match this company. Five are substantive — and all five come from a SINGLE source on a SINGLE date (no_priors, 2026-02-05):

• conviction 85: "Roblox's 13B hours/month of 3D vector (not raster) interaction data is uniquely powerful training data for NPCs and world models, reproducible from any camera angle."
• conviction 80: "AI superpowers Roblox's 20-year vision of a photorealistic 'holodeck' — scaling multiplayer fidelity from 100 to 10,000 people with realistic physics/acoustics."
• conviction 70: "Roblox Studio runs native code assistant plus environmental-generation engine, integrating industry AI workflows; agents will autonomously test/iterate experiences against the cloud."
• conviction 70 (principle): "'Take the long view + get stuff done' — pair 20-year vision with weekly rapid iteration; break big products into shippable weekly increments toward a clear target."
• conviction 65 (fact): "Top Roblox creators now earn tens of millions/year; long-tail revenue growing faster than #1 creator, top-20 experiences increasingly competitive with rapid live-ops updates."

These are half-weighted, and the reason is important. The framing is unmistakably first-person company language — "Roblox's 20-year vision", a company operating principle stated as a maxim, internal product-roadmap detail about what agents will do. Under the house rule on management framing, claims sourced from company voices carry half weight. We flag this rather than counting five high-conviction claims at face value, and it is the single largest reason the conviction rating is Low-Moderate rather than High despite an average conviction of 74 across the substantive claims.

Four further claims are tangential and add breadth without content:

• real_vision, 2022-08-15 (bullish, 60): "Platform-era winners quietly build for decades before breaking out; Nvidia, Epic and Roblox went unnoticed for years before mattering."
• invest_like_the_best, 2025-02-01 (neutral, 70): "Virtual worlds succeed on simplicity and low learning curve (Minecraft/Roblox); replicating enterprise meetings in-world is a useless use of the tech."
• raoul_pal, 2025-06-05 (neutral, 55): "Blockchain gaming hasn't fully lived up to its promise, but AAA on-chain games (Roblox-style) are inevitable long-term."
• invest_like_the_best, 2025-07-29 (neutral, 85): "Excluding Uber and Roblox, almost no power-law company (>$5B) had its seed round led by a seed-stage VC firm." — a claim about venture capital structure, not about Roblox as an investment.

The most important observation about this entire lane: not one claim addresses the bear case. There is nothing in the knowledge base about regulation, child safety, user-growth deceleration, competitive pressure, or valuation. The panel's coverage of this name is a product and technology lane, dated February 2026, before the bulk of the decline. It tells us the asset is interesting. It tells us nothing about why the stock fell 74%. Presenting it as conviction support for buying the dip would be a misuse of it, and we decline to do that.

6. Valuation — priced in or room?

At $36.67 (market cap $26.196B, net cash $1.252B, EV $24.94B):

Consensus (with the basis caveat from section 1(b) applying to every revenue figure):

Consensus does not model GAAP profitability until FY30. That is four and a half years away and it is the central reason this stock carries no earnings-based support.

6a. What today's price assumes (the inversion)

At $36.67 — approximately 18.4x trailing free cash flow and 4.39x trailing GAAP revenue — the price embeds the following falsifiable claims:

• Free cash flow does not grow materially from the FY25 $1.353B base. An 18.4x multiple on a business growing revenue 36% is what the market pays for a cash flow it expects to stall. If free cash flow simply grows in line with revenue, the forward multiple is roughly 13x — which is a value multiple for a growth asset.
• GAAP losses persist through FY29 (consensus: -$1.45, -$1.15, -$0.93, -$0.42 per share), i.e. the 14.3-point annual margin improvement just delivered does not continue at anything like that rate. This is the assumption most directly contradicted by the last reported quarter.
• Revenue growth decelerates from 36% to roughly 15% by FY28 (consensus, on the mixed basis: +14.7% FY27, +18.0% FY28) — a halving of the current rate within two years.
• Dilution continues at roughly 5% annually with no buyback, despite $1.25B of net cash and a share price 74% off its high.
• Whatever caused the 74% decline is permanent and material. This is the assumption we cannot test, and it is doing the most work in the price.

The single most fragile assumption: that the operating-margin improvement stops. Q2 2026 delivered a 14.3-point year-over-year margin gain and a $93.5M absolute reduction in operating loss on 36% revenue growth. Consensus models losses persisting for four more years. One of those two things is wrong, and the reported quarter is the one with evidence behind it.

6b. The return bridge (why the multiple moves)

Expected return ≈ free-cash-flow growth + multiple drift + shareholder yield. We use free cash flow rather than EPS because there are no earnings.

• Free-cash-flow growth: FY25 delivered $1.353B against $641.3M in FY24 — a 111% increase. Even assuming that pace halves and then halves again, free cash flow above $1.6B within the horizon is a modest assumption given 36% revenue growth and a narrowing operating loss.
• Multiple drift: our base case assumes modest EXPANSION, from ~18.4x to ~22x free cash flow, and we flag this as the fragile leg of the bridge. The justification: 18.4x for a 36% grower with net cash and improving unit economics is a distressed multiple, and even partial normalisation of sentiment moves it. But a thesis that requires multiple expansion is inherently weaker than one that does not, and we say so explicitly rather than burying it. If the multiple simply holds at 18.4x and free cash flow grows 20%, fair value is approximately $38 — barely above spot. The base case needs both legs.
• Shareholder yield: NEGATIVE, approximately -5%. No dividend, no buyback, and roughly 5% annual dilution. This is a real and under-discussed drag — the only name in this batch where the shareholder yield subtracts from returns rather than adding to them.

The bridge reads: solid free-cash-flow growth, plus assumed multiple normalisation, less ~5% annual dilution. Net approximately +18% to our base case. The dilution term is why the base case is $43 rather than $46.

Scenario bands (stated arithmetic, no DCF):

• Bear ~$22 — 12x a declining $1.2B of free cash flow, plus $1.25B of net cash, across 714.4M shares. This is the scenario in which the invisible cause of the drawdown is real and material: growth decelerates hard, the cash flow that has been compounding stalls, and the market applies a structurally-impaired multiple. -40% from spot.
• Base ~$43 — 22x the FY25 free cash flow of $1.353B, plus net cash. Growth moderates but continues, the margin improvement persists, sentiment partially normalises, and dilution continues. +18% from spot.
• Bull ~$64 — 28x a forward $1.6B of free cash flow, plus net cash. The operating-margin trajectory continues, GAAP breakeven arrives well ahead of the FY30 consensus, the data-asset thesis begins converting to revenue, and the multiple normalises toward a growth-platform level. +76% from spot.

Base sits 18% above spot and 13% below the $49.47 street consensus. We are more cautious than the sell-side, principally because of the dilution drag and because we discount for the information we do not have.

6c. Variant perception (where we differ, what would surprise)

• Our variant perception is on the operating-margin trajectory, and it is specific. Consensus models GAAP losses through FY29. The most recent quarter improved operating margin by 14.3 percentage points year over year and reduced the operating loss by $93.5M in absolute dollars while revenue grew 36%. Extrapolating even a third of that rate produces GAAP breakeven around FY28, two years ahead of consensus. That is where our base case gets its multiple expansion, and it is testable every ninety days.
• We differ from the sell-side by being MORE cautious on the price, despite being more constructive on the margins. Our $43 base is 13% below the $49.47 consensus, because we explicitly charge the ~5% annual dilution and because we apply a discount for underwriting an unexplained drawdown. We think the street is under-weighting both.
• We hold NO variant perception on the cause of the decline, and we state that plainly. We do not know why the stock fell 74%. Any dive that offered a confident explanation from this dataset would be inventing one.
• Positive surprise that would force a repricing: Q3 2026 on 2026-10-29 delivering a further absolute reduction in operating loss (below $229M) on continued 30%+ growth — the third consecutive quarter of the pattern — or, more powerfully, the announcement of a buyback, which with $1.25B of net cash at $36.67 would be highly accretive and would signal that management regards the price as dislocated. An insider purchase would be the single most informative event available.
• Negative surprise: the operating loss widening again; deferred revenue growth falling below 20%; free cash flow declining year over year; the 52-week low of $35.60 breaking on volume; or the emergence — through any channel — of a regulatory or safety development that would retrospectively explain the drawdown.

7. Moat and competitive position

The moat has three layers, and the knowledge base describes all three.

The creator economy. Roblox does not make the content; users do, and they are paid for it. The knowledge base records that top creators earn tens of millions of dollars annually, and — the more important detail — that long-tail creator revenue is growing faster than the top creator's. A creator ecosystem where the tail grows faster than the head is a healthy one: it means opportunity is widening rather than concentrating, which is what keeps new creators arriving. This is a two-sided network that a competitor cannot buy.

The data asset. The strongest claim in the lane: 13 billion hours per month of three-dimensional vector interaction data — scene geometry rather than rendered video, reproducible from any camera angle. For training world models and non-player-character behaviour that is a materially more useful corpus than video, because it contains the underlying structure rather than a projection of it. No competitor has an equivalent, and the asset compounds with usage.

Switching costs. Users' social graphs, purchased items, creations and reputations live on the platform. For the core cohort these are not transferable.

Where the moat is weakest. The business is concentrated in a demographic that ages out, so cohort replacement is a permanent requirement. The 78.1% gross margin is struck after creator payouts, so the economics depend on a revenue-share that competitors could bid against. And — the honest gap — with no user metrics, no geographic split and no age-cohort disclosure in this file, we cannot verify that the moat is holding. We are inferring platform health from revenue growth and deferred revenue, which are good proxies but not direct evidence.

The vendor's assigned peer set (CoreWeave, Electronic Arts, Fortinet, Corning, Infosys, Motorola Solutions, Strategy, NetEase, TE Connectivity, Take-Two) is a market-capitalisation cohort rather than a competitive set. The only genuinely comparable names in it are Electronic Arts ($52.98B) and Take-Two ($45.84B) — both traditional publishers with fundamentally different economics, and both carrying market capitalisations roughly twice Roblox's on materially slower growth.

8. Insider activity

The record shows three sellers and no buyers, all in early July 2026:

Two observations, and the second is the one that matters.

First, the sales are small — roughly 8,300 shares by Baszucki (a director, retaining 8,969,734 shares in one holding plus 1,285,518 in another) and roughly 4,500 by Reinstra (retaining 422,248). These are trivial fractions of very large positions and carry little signal. The Rawlings transaction is an equity award, not a sale.

Second, and materially: every one of those sales was executed between $55.77 and $58.03 — that is 52% to 58% ABOVE today's price of $36.67. The stock has fallen roughly 35% in the four weeks since. These insiders did not sell into a top; they sold into what has turned out to be the beginning of a further collapse. That tells us the decline was not anticipated by the people best placed to anticipate it — which is mild evidence against the hypothesis that a known company-specific problem drove the fall, and mild evidence for the dislocation reading.

And there are no purchases. With the stock 74% off its high and $1.25B of net cash on the balance sheet, the absence of a single insider buy — and of any buyback — is the most conspicuous gap in the entire file.

9. Verdict, kill-criteria and flip conditions

Stage-In (deep-value, high-uncertainty). At $36.67 Roblox trades at roughly 18.4x trailing free cash flow and 4.39x trailing GAAP revenue, with $1.25B of net cash, against a business that grew revenue 35.9% in the quarter just reported, improved operating margin by 14.3 percentage points year over year, generated $1.353B of free cash flow at a 27.7% margin (more than double the prior year), grew deferred revenue 38.7%, and beat consensus in each of its last four prints. Our base fair value of $43 is 18% above spot; the street's is $49.47.

Three reasons this is staged rather than full-size, each falsifiable:

1. This dataset contains no explanation for a 74.1% decline. Not one financial line deteriorated. An unexplained collapse of that magnitude usually has a cause, and we cannot see it. Sizing is the only honest response to an information gap of this size.

2. The knowledge-base lane, though bullish, is silent on the bear case — five substantive claims, all from one source on one date in February 2026, all in company-vision framing and therefore half-weighted, none addressing regulation, safety, competition or valuation.

3. Shareholder yield is negative — roughly 5% annual dilution with no buyback, despite $1.25B of net cash and a 74% drawdown. That is a real drag and a real criticism of capital allocation.

Staged entry — build a 1.5-2.5% position in three tranches:

1. Tranche 1 — now (~1/3 of target), at ~$37. RSI 23.4, 3% off the 52-week low, ~18x free cash flow, net cash, four consecutive beats. You are buying the arithmetic, deliberately small.

2. Tranche 2 — the reaction to the 2026-10-29 print. Add on a third consecutive quarter of absolute operating-loss reduction with revenue growth above 30%. If the loss widens or growth falls below 25%, do not add.

3. Tranche 3 — either a reclaim of the 50-DMA (~$49) on volume, confirming a trend change, or a capitulation below $30 where the bear case is close to fully priced. Whichever comes first.

Pre-registered KILL criteria (any one converts Stage-In → Avoid, and stops further tranches):

• The operating loss widening in absolute terms for two consecutive quarters — the margin-improvement trend, which is the entire basis of our variant perception, breaking.
• Revenue growth below 25% on 2026-10-29.
• Deferred revenue declining, or growing below 20% — the earliest available signal that customer prepayment behaviour is deteriorating, and one that leads revenue.
• Free cash flow below $1.0B on a trailing basis — removing the only valuation anchor this stock has.
• The 52-week low of $35.60 breaking decisively, taking the stock into territory with no reference point in this dataset.
• Any emerging regulatory, safety or legal development that retrospectively explains the drawdown — at which point the thesis was never about valuation.

Pre-registered FLIP TO HIGH-CONVICTION (upsize toward 3-4%):

• A share buyback announced. With $1.25B of net cash at a 74% drawdown this would be both highly accretive and a strong management signal, and it would eliminate the negative shareholder yield that is currently costing 5% a year.
• Any insider PURCHASE. The single most informative event available given the information gap.
• A third consecutive quarter of absolute operating-loss reduction on 30%+ growth, putting GAAP breakeven visibly ahead of the FY30 consensus.
• A second, independent tracked source naming the company — which would broaden a lane currently resting entirely on one half-weighted voice.

Where RBLX fits in the Synthos Framework Portfolio. A high-uncertainty / dislocation sleeve position at 1.5-2.5%, explicitly below a normal weight. It is not an AI-infrastructure holding and should not be sized as one; the AI angle here is a data asset that may compound, not a revenue line that does. Logged as a tracked Synthos call (Stage-In) as of 2026-08-04 at $36.67.

Single biggest risk: that the market knows something this file does not. Revenue accelerated, margins improved 14 points, cash flow doubled, and the stock fell 74%. Either that is one of the larger mispricings available, or there is an input we cannot see. The entire position-sizing discipline in this dive exists because we cannot distinguish between those two possibilities from the data provided.

Most fragile assumption in the price: that the operating-margin improvement stops. Consensus requires GAAP losses through FY29. The most recently reported quarter cut the absolute operating loss by $93.5M while growing revenue 36%. Those two statements cannot both survive another four quarters, and the resolution — in either direction — is the single largest determinant of what this stock is worth.

Provenance & disclosures

• Traceability: 9 KB claims match this company (kb_claim_count: 9, breadth 4). Five are substantive and ALL come from a single source on a single date (no_priors, 2026-02-05, conviction 65-85). These five are HALF-WEIGHTED under the house rule on management framing — the language is first-person company voice ("Roblox's 20-year vision", an internal operating principle, forward product-roadmap detail), which is the fingerprint of a guest or management source rather than an independent analyst. Four further claims (real_vision 2022-08-15, invest_like_the_best 2025-02-01 and 2025-07-29, raoul_pal 2025-06-05) are tangential and carry minimal weight. No claim in the knowledge base addresses the bear case, the regulatory environment, or the valuation — the lane is a product-and-technology lane predating most of the decline, and it is presented as such rather than as conviction support for buying.
• Data as-of: fundamentals 2026-06-30 (Q2 2026, reported 2026-07-30; fiscal year ends December) · estimates 2026-08-04 · prices 2026-08-04 ($36.67, 50-DMA $48.79, 200-DMA $69.40, RSI 23.4) · KB claims 2026-08-04. Sole market-data source: the Synthos vendor file for this ticker.
• Data REJECTED — the entire consensus EBITDA series. Every forecast year shows negative ebitdaAvg, growing to -$3.389B in FY30, while the same FY30 row shows netIncomeAvg of +$179.3M. Negative EBITDA of $3.4B alongside positive net income is arithmetically impossible. No EBITDA-based multiple appears anywhere in this dive. This failure mode recurs across several files in this batch and appears systematic.
• Data FLAGGED — mixed revenue bases. The consensus revenue series is not GAAP: consensus FY25 revenue of $6.616B against actual GAAP revenue of $4.891B (+35%), and consensus FY26 of $7.156B against a first-half GAAP run-rate of roughly $6.0B. The earnings-calendar revenueActual field switches basis mid-series — the 2025-07-31 and 2025-10-30 entries match GAAP to the dollar, while the 2026-02-05, 2026-04-30 and 2026-07-30 entries do not ($2,222M / $1,731M / $1,557M against GAAP $1,415M / $1,442M / $1,469M). The most likely explanation is a bookings basis. Every revenue multiple in this dive uses reported GAAP revenue only, and the Q3 revenue estimate of $1.664B is explicitly labelled as not comparable to reported revenue.
• Data gaps explicitly noted: seg_prod and seg_geo are both EMPTY — no product, geographic, user, age-cohort or bookings-per-user disclosure whatsoever. For a platform business whose most plausible structural risk is jurisdictional regulation, the absence of any geographic split is a significant gap. And critically: this file contains no news, regulatory, safety, litigation, competitive or guidance information — there is nothing here that explains a 74% drawdown, and the dive says so rather than inventing a narrative.
• Vendor ratios that must NOT be read naively: debtToEquityRatioTTM 12.08, priceToBookRatioTTM 172.9, currentRatioTTM 0.82. All three are artefacts of a $394.5M equity base compressed by a -$5.061B accumulated deficit and $4.169B of deferred revenue classified as a current liability. Recomputed by hand: net CASH of $1.252B (cash $1.205B + short-term investments $1.850B − total debt $1.803B). The vendor's netDebt of +$597.4M nets cash only. Enterprise value used throughout: $24.94B (vendor: $27.04B).
• Data VALIDATED: the most recent quarter (inc_q[0], Q2 2026) passes the standard corruption screen — EPS reconciles against net income and share count (-$183M ÷ 716.767M = -$0.255 versus -$0.26 reported), share counts do not duplicate the prior quarter and continue a steady trend, interest expense is consistent across all quarters, and the reported EPS matches the earnings-calendar actual exactly.
• Fair-value caveat: the $22 / $43 / $64 anchors are scenario multiples applied to free cash flow (there are no earnings), plus net cash, divided by the current share count — not a discounted cash-flow model. The base case requires BOTH free-cash-flow growth AND modest multiple expansion (18.4x → ~22x); if the multiple merely holds and free cash flow grows 20%, fair value is approximately $38, barely above spot. That dependence is disclosed rather than hidden, and it is why the verdict is Stage-In rather than Buy. The base case also charges approximately 5% annual dilution, which no buyback currently offsets.
• Not investment advice. Independent research, educational and informational only, never personalised.
• Version: 2026-08-04-full.