Rocket Companies RKT

Reference table

And we still will not underwrite it, because the balance sheet in this file is unusable. bal_a for FY2025 reports total debt of exactly $0, long-term debt $0, short-term debt $0, and net cash of $2,696M — for a mortgage originator holding $60,685M of assets that reported $13,976M of total debt one year earlier. The same file's km_ttm carries an enterprise value of $67,623M against a $38,633M market capitalisation, implying roughly $29,000M of net DEBT. That is a $32 billion internal contradiction inside one record. Alongside it: current liabilities of $285M against $60.7B of assets; a basic share count that jumps from 148M to 2,828M mid-series; earnings-calendar revenue that disagrees with the income statement in three consecutive quarters by 7-13%; and a dividend field showing $0.80 while the trailing dividend yield reads zero. A lender whose leverage cannot be determined cannot be valued. Add a 2.176 beta — the highest in this batch — and twelve-month performance of −7.4% against +19.9% for the index, and the honest verdict is Watch.

1. Data integrity — this section governs the entire dive

This is the most badly corrupted file in this batch and the corruption is concentrated in exactly the place that matters most for a lender. Stated in full:

REJECTED — the entire FY2025 balance sheet debt and liability presentation

bal_a for the period ended 2025-12-31 reports:

A mortgage originator funds loans held for sale with warehouse facilities. Reporting zero total debt against $60,685M of assets and $4,736M of current assets is not a rounding issue — it is a broken record. The prior year, on a balance sheet less than half the size, carried $13,976M. We reject every debt, leverage and liability field in bal_a[0].

The $32 billion contradiction

The same file's km_ttm reports enterpriseValueTTM: 67,622,514,037 against a market capitalisation of $38,632,514,037 — a difference of $28,990M, i.e. the enterprise-value calculation is using roughly $29.0B of net debt. ratios_ttm corroborates: debtToMarketCapTTM: 0.6115 implies debt of ~$23.6B, and debtToEquityRatioTTM: 1.364 and debtToAssetsRatioTTM: 0.533 both describe a levered entity.

So one part of the file says net cash of $2.7B and another says net debt of $29.0B. The spread is $31.7B — 82% of the market capitalisation.

Consequences, stated plainly:

• No enterprise-value multiple appears in this dive. EV/revenue, EV/EBITDA and net-debt-to-EBITDA are all unavailable.
• No leverage or coverage assessment is offered. netDebtToEBITDATTM of 16.76x and interestCoverageRatioTTM of 1.246x are both computed from contradictory inputs and are discarded.
• The valuation in §7 is built on earnings multiples only, which are the only metrics that do not require a debt figure.

REJECTED — the dividend fields

profile.lastDividend reports $0.80, while ratios_ttm reports dividendYieldTTM: 0 and dividendPerShareTTM: 0 — and simultaneously dividendPayoutRatioTTM: 0.9748, a 97% payout ratio on a zero dividend. All three cannot be true. The most coherent reading is that $0.80 is a stale historical figure (Rocket has paid special dividends in the past) and that no dividend has been paid in the trailing period. We treat RKT as paying no current dividend and reject the payout-ratio field.

FLAGGED — the Up-C share count makes every per-share figure ambiguous

inc_q shows this progression:

This is an Up-C structure: a small publicly traded Class A float alongside a much larger controlling interest held in units of an underlying partnership, exchangeable into Class A shares. Basic EPS is computed on the small float; diluted on the full economic count. The presentation changed between Q2'25 and Q3'25, which is why the ratio collapses from 12x to 1x.

Consequence: ratios_ttm.priceToEarningsRatioTTM of 210.8x cannot be reconciled with our own computation. Our figures: trailing net income of $239.4M (−$1.785M − $123.854M + $68M + $297M) divided by 2,828.5M shares gives $0.0846 of trailing GAAP EPS, and $13.68 ÷ $0.0846 = 161.6x. The vendor's 210.8x uses a different share base. We use the 2,828.5M economic count throughout and label it. The quote's market capitalisation of $38,632.5M at $13.68 implies 2,823.5M shares, which confirms that the vendor's market capitalisation already uses the full economic count — an important consistency check that we did pass.

FLAGGED — earn_cal revenue disagrees with inc_q in three consecutive quarters

Three consecutive quarters, in both directions. The most likely explanation is that earn_cal carries an adjusted or "total revenue net" measure while inc_q carries the GAAP line. We use inc_q for all revenue figures and note that the 2026-08-06 consensus revenue of $2,811.8M may not be like-for-like with the income statement.

FLAGGED but NOT corrupt — negative operating cash flow

cf_a reports operating cash flow of −$3,927M (FY25) and −$2,629M (FY24), with free cash flow of −$4,018M and −$3,434M. This is not distress and must not be read as such. For a mortgage originator, loans originated and held for sale are classified within operating activities, so a period of rising origination volume produces large negative operating cash flow by construction. FY22, a year of shrinking volumes, produced +$10,823M of operating cash flow for exactly the mirror-image reason. Negative operating cash flow here is a volume signal, not a solvency signal, and ratios_ttm's price-to-free-cash-flow of −27.7x should be ignored rather than interpreted.

GAP — seg_geo is empty

No geographic revenue disclosure at all. Immaterial for a US-and-Canada lender, but noted.

CLEAN — the income statement

FY25 quarterly revenue sums to $6,839.4M against the stated annual $6,880M (a 0.6% difference, likely a restatement). Quarterly net income for FY25 sums to −$68.0M, matching inc_a exactly. inc_q[0] (2026-03-31) is internally consistent: revenue $2,738M, gross profit $2,446M, operating income $749M, net income $297M, EPS $0.11 on 2,828.5M basic shares ($297M ÷ 2,828.5M = $0.105). The income statement is usable; the balance sheet is not.

2. What happened in 2025 — the transformation, from the numbers only

The file does not name any acquisition, and this dive names none, because no counterparty appears anywhere in the data. What the data does show, unambiguously:

The signature is complete and consistent: goodwill up $9,384M, total assets up $36,175M, share count up 792M (+38.5%), and $2,323M of net cash spent on acquisitions. Rocket bought at least one very large business during 2025, paying predominantly in equity.

Two consequences that matter for the verdict.

First, $10,611M of goodwill is 17.5% of total assets and 27.5% of the market capitalisation. None of it has been tested through a housing downturn. In FY22 Rocket's revenue fell from $13,176M to $6,005M in a single year; a repeat with $10.6B of goodwill on the books is an impairment risk of a size that would dwarf any operating result.

Second, the revenue growth is purchased, and consensus knows it. FY26E revenue of $11,472M is +66.7% over FY25 — the arithmetic of owning the acquired business for a full year. Then it stops: FY27E $12,724M (+10.9%) and FY28E $13,341M (+4.8%). The street models one year of acquisition arithmetic followed by a mature business. Anyone reading "+67% growth" as an organic rate is misreading it.

3. The operating trajectory — the genuine bull case

Adjusted earnings per share, actual versus estimate, five consecutive quarters:

Five for five, with beats averaging +32%, and the absolute level rising monotonically from $0.04 to $0.15 — a near-quadrupling in twelve months. Trailing adjusted EPS is $0.37 ($0.04 + $0.07 + $0.11 + $0.15), putting the stock at 37.0x trailing adjusted earnings and, on consensus, 20.4x FY26E and 14.0x FY27E.

The GAAP picture is far weaker and the divergence should be understood. Quarterly GAAP net income: −$10.4M (Q1'25), −$1.8M (Q2'25), −$123.9M (Q3'25), +$68M (Q4'25), +$297M (Q1'26). Trailing GAAP net income is $239.4M, giving trailing GAAP EPS of $0.085 — a 161.6x multiple. The gap between $0.37 adjusted and $0.085 GAAP is roughly $0.29 per share, or $820M annually, and the file does not disclose its composition. Given $343M of stock-based compensation and a large acquisition, the most plausible components are acquisition-related amortisation, integration costs and share-based pay. We flag that the adjusted series — the one the street values on — excludes roughly three-quarters of the gap between the two measures, and that this is an unusually wide adjustment.

Segments (FY25): Direct-to-Customer $4,791M (87.8%), Partner Network $668M (12.2%). The two sum to $5,459M against $6,880M of total revenue, leaving $1,421M unallocated — a further reconciliation gap, flagged. The direct-to-consumer concentration has risen from 87.2% (FY23) and is the structural feature of the franchise: Rocket owns the customer relationship rather than renting it from brokers.

4. Cyclicality — the amplitude that governs everything

Revenue fell 70% from peak to trough in two years ($13,176M to $4,006M) and operating income swung from +$6,185M to −$403M. That is the amplitude of the mortgage cycle, and it is the single most important context for any forward estimate. Note also that even in the boom year FY21, net income attributable was only $308.2M against $6,185M of operating income — the Up-C structure allocates the great majority of earnings to the non-controlling interest, which is why GAAP EPS of $2.36 accompanied $6.2B of operating profit.

The bull case is a cycle bet. Consensus FY27 revenue of $12,724M would be roughly the FY21 boom level — but on a much larger, acquisition-enlarged platform. The bear case is that the cycle does not turn, in which case a business generating $600M of operating income against $10,611M of goodwill and an undetermined debt load is a very different security.

5. Financials — what can be said with confidence

Given §1, this section is deliberately restricted to income-statement and cash-flow facts we trust.

• Revenue: FY25 $6,880M (+27.4%); trailing $8,476.1M. Quarterly run-rate now $2,738M.
• Profitability: FY25 operating income $600M (8.7% margin), net income −$68M. Trailing: operating margin 15.6%, net margin 2.8%, net income $239.4M.
• Gross margin: 90.2% trailing — but this is a financial-services presentation where "cost of revenue" is largely direct origination cost, and it carries no information. Not used.
• Operating expense: SG&A $2,527M (36.7% of FY25 revenue), trailing 26.7%. Stock compensation $343M (5.0% of FY25 revenue, 5.1% trailing) — moderate.
• Cash flow: operating cash flow −$3,927M (FY25), which as explained in §1 reflects origination volume rather than distress. Capital expenditure $91M. No meaningful free-cash-flow measure exists for this business model and none is used.
• Returns: return on equity 1.5%, return on assets 0.4%, return on invested capital 1.5%. These are genuinely low and are computed from fields we trust (income statement and equity). A 1.5% return on equity is the strongest single argument against the bull case — even granting the adjusted earnings, the capital base is enormous relative to the profit it produces.
• Book value: $8.213 per share (ratios_ttm), against a $13.68 price — price-to-book 1.67x. Tangible book value $3.716 per share, so price-to-tangible-book is 3.68x, and the gap is the $10,611M of goodwill.

The return-on-equity figure deserves emphasis. At 1.5%, Rocket currently earns less on its equity than a Treasury bill. Consensus FY27 net income of $2,406.0M against $22,898M of equity would be roughly 10.5% — a normal figure for a lender. The entire investment case is the journey from 1.5% to 10.5%, and it depends on the cycle.

6. Knowledge base — one hit, and it is not about the stock

A search of 51,928 knowledge-base entries for "RKT", "Rocket Companies" and "Rocket Mortgage" returns one match:

a16z (2026-02-09, conviction 68, bullish, horizon: thesis) — entities: [empty]; category: enterprise AI adoption: "Over the next 5 years there will be a reckoning: companies that push through change management and adopt the best AI tools (e.g. Chime -60% support cost, Rocket Mortgage $40M savings) gain huge productivity advantages over laggards."

How to read this honestly, and it requires care. This is not a view on Rocket Companies as a security. The entity field is empty. The claim is a general argument about enterprise technology adoption, and Rocket Mortgage is cited as one of two illustrative examples alongside an unrelated private company. The $40M figure is offered as evidence for a thesis about change management, not as a valuation input — and $40M against $6,880M of FY25 revenue is 0.6%, immaterial to any earnings estimate.

kb_claim_count: 1, but the substantive count is zero. No tracked voice has expressed a view on this security, its valuation, its balance sheet or its prospects. This dive is fundamentals- and quant-driven — an accepted and precedented outcome — and it is being conducted under the severe data limitations documented in §1. That combination, a corrupted file and no independent expert view to triangulate against, is the specific reason the verdict is Watch rather than a directional call.

7. Valuation — earnings multiples only, because nothing else is available

At $13.68 (market cap $38.63B, net debt undeterminable — see §1):

Note what is missing and why it matters. For a lender, the enterprise-value and leverage metrics are ordinarily the most informative available. Here they are unavailable because of the $32B contradiction in §1. Valuing a mortgage originator on price-to-earnings alone is like valuing a bank on revenue — it can be done, but the most important dimension of the risk is invisible.

7a. What today's price assumes (the inversion)

Reverse-engineering $13.68 into falsifiable claims. Consensus-derived arithmetic, labelled as such.

• Adjusted EPS reaches $0.977 by FY27, a 2.6x increase from the $0.37 trailing. Consensus (10 analysts). At 14.0x FY27E the price is roughly consistent with delivery. Falsifiable quarterly, and the beat streak (§3) supports it — five consecutive beats, absolute level up 4x in a year.
• Revenue reaches $11,472M in FY26 — the arithmetic of a full year of acquired business. Consensus (8 analysts). This one is close to mechanical and is the least fragile assumption in the price; the acquisitions closed, the revenue exists.
• Return on equity travels from 1.5% to roughly 10.5%. FY27 consensus net income of $2,406.0M against $22,898M of equity. This is the most fragile assumption in the price, and it is the same assumption stated a different way: it requires both the acquired platform to earn its cost of capital and the mortgage cycle to cooperate. Neither has been demonstrated.
• The $10,611M of goodwill is not impaired. It is 27.5% of market capitalisation. FY23 saw revenue fall to $4,006M; a repeat with this goodwill on the books would produce a write-down larger than any plausible year of earnings. The price embeds zero probability of impairment.
• The share count stabilises near 2,847M. It rose 38.5% in twelve months. Any further equity issuance dilutes the $0.977 directly.
• Leverage is manageable. The price must be assuming something about this, and we cannot verify what, because the file reports both zero debt and $29B of net debt. This is not an assumption we can test, and its untestability is the core of the verdict.

7b. The return bridge (why the multiple moves)

Expected return over an eighteen-to-twenty-four-month horizon: EPS growth (+164% cumulative to FY27 on consensus) + multiple compression (large) + shareholder yield (zero — no dividend, no buyback recorded).

Our base case assumes very substantial multiple COMPRESSION, and this needs saying clearly because it is unusual. The market pays 37.0x trailing adjusted earnings today. Our base case applies 15x to FY27E EPS of $0.977 — a de-rating of well over half. The justification is not pessimism about the earnings; it is that 37x is a multiple appropriate to a growth company, and Rocket is a cyclical lender whose consensus revenue growth falls to 4.8% by FY28. Cyclical lenders in the middle of a cycle trade at low-to-mid-teens earnings multiples, not high-thirties.

So the base-case return of roughly +8% is: a 164% increase in earnings almost entirely offset by a de-rating from 37x to 15x. That is the whole bridge, and it explains why a stock the street targets 42.5% higher can carry a base case only 8% above spot — the disagreement is not about the earnings, it is about what multiple a mid-cycle mortgage lender deserves.

The bull case at $19.50 requires 20x FY27E — the street's implicit multiple. That is defensible if one believes the cycle is turning and the platform is structurally larger. It is the fragile leg, and it depends on a rate and volume environment this dive does not forecast.

7c. Variant perception (where we differ, what would surprise)

• We differ from the street on the multiple, not the earnings. The consensus $19.50 target implies ~20x FY27E adjusted EPS. We think a lender with a 1.5% current return on equity, a 2.176 beta, $10.6B of untested goodwill, consensus revenue growth of 4.8% by FY28 and an unreadable balance sheet deserves ~15x, not 20x. That single difference is the entire $4.75 gap between our base case and theirs. Falsifiable over time, not at any single print.
• We think the adjusted-versus-GAAP gap is being under-scrutinised. Trailing adjusted EPS is $0.37; trailing GAAP EPS is $0.085. The $0.29 per share difference is roughly $820M annually on a company whose FY25 operating income was $600M. An adjustment larger than the operating profit is unusual and the file does not disclose its composition. Negative surprise, watchable: any quarter where the adjustment widens further, or an impairment that reveals what was being added back.
• We may be wrong on timing, and the beat streak is the reason. Five consecutive beats averaging +32%, with the absolute level rising from $0.04 to $0.15, is a genuine momentum signal in the operating results. If the 2026-08-06 print delivers a sixth beat above ~$0.20, the FY27 estimate starts looking conservative rather than heroic, and our 15x becomes the binding constraint rather than the estimate. Positive surprise, watchable 2026-08-06: adjusted EPS above ~$0.20.
• The genuinely under-priced positive is the option value of a rate cycle. FY21 produced $6,185M of operating income on a smaller platform. If origination volumes normalise toward that environment on a post-acquisition base, the FY28 consensus EPS of $1.113 is far too low. We assign it no value because the file gives no basis to forecast rates, but it is why the bull case is a real bull case rather than a courtesy. Not watchable in the file; watchable in the rate environment.
• Where we have NO edge and say so plainly: on the balance sheet. We cannot determine the leverage of a leveraged lender from this file. That is not a variant perception; it is an absence of information, and the correct response to an absence of information is not to take a position.

Synthos fair values

• Bear ~$8.00 — a housing-freeze scenario in which FY27 adjusted EPS lands near $0.50 rather than $0.977, valued at 16x. Or, equivalently, the current $0.37 trailing adjusted EPS valued at ~21x with no recovery. Would also likely involve a goodwill review. −41.5%.
• Base ~$14.75 — 15x FY27E adjusted EPS ($0.977) = $14.66; cross-checked at 13x FY28E ($1.113) = $14.47. Takes the midpoint at $14.75. Assumes the acquisitions deliver, the beat streak continues, and the market applies a normal mid-cycle lender multiple. +7.8%.
• Bull ~$19.50 — 20x FY27E adjusted EPS ($0.977) = $19.54, which lands precisely on the street's $19.50 consensus target. Assumes a cycle turn, full integration, and the market paying a growth multiple. +42.5%.

Base sits ~8% above spot with a −41.5% bear and a +42.5% bull. That is a very wide, roughly symmetric distribution around a thin central expectation — the definition of a security to watch rather than own, particularly on a 2.176 beta.

8. Technicals — a broken chart with one loud session

• Price $13.68, up 6.1% on the day on 16.2M shares — a notably heavy session against the file's typical volume.
• −41.6% from the $23.44 fifty-two-week high; fifty-two-week low $12.35, so the stock sits just 10.8% above its low. max_dd_from_peak is −67.1%.
• Below a falling 50-day average ($13.99) by 2.2% and 17.1% below a falling 200-day average ($16.51). Both above the price, both declining.
• RSI 44.6, MACD −0.27. Neutral-to-weak; no oversold signal despite proximity to the low.
• Relative performance is the worst in this batch: −7.4% over twelve months against +19.9% for the S&P (a 27-point deficit), −23.7% over six months against +9.5%, −6.6% over three months against +5.1%. Rocket has underperformed on every horizon.
• Beta 2.176 — the highest in this batch, more than twice the market. A 1% index move implies a 2.2% move here.
• Read: the 50-day at $13.99 is 2.2% above and is the first level to reclaim; the 200-day at $16.51 is 20.7% above and would confirm a trend change. Today's 6.1% session, two days before earnings, is a positioning move rather than a technical repair. There is no base and no support until $12.35.

9. Insider activity

The file records only routine annual director equity grants, all dated 2026-06-10 at $0 (awards, not purchases): Suzanne F. Shank 16,312 shares (holding 119,558), Tagar Olson 11,255 (holding 289,595), Jonathan D. Mariner 16,312 (holding 102,300), and a further partial record.

Read: no directional signal in either direction. There are no open-market purchases and no sales anywhere in the file. Notably, there is no insider buying into a 67% drawdown — which for a controlled company with a large founder interest is not necessarily meaningful, but is worth stating as an absence.

Governance note. The Up-C structure documented in §1 means the public Class A float is a small minority of the economic interest, with the majority held in exchangeable units. Public shareholders do not control this company, and the structural allocation of earnings between Class A holders and the non-controlling interest is what caused FY21's $6,185M of operating income to produce only $2.36 of Class A EPS. Any per-share forecast for this security is a forecast about an allocation as much as about a business.

10. Franchise and competitive position

Per the profile, Rocket spans Rocket Mortgage (origination), Amrock (title insurance, valuation, settlement), Rocket Homes (search and agent referral), Rocket Auto, Rocket Loans (personal lending), Core Digital Media (digital advertising), Rocket Solar and a personal-finance application. 23,500 employees.

The genuine competitive asset is the direct-to-consumer relationship. Direct-to-Customer is 87.8% of segment revenue, against a Partner Network at 12.2%. Owning the customer rather than renting them from a broker network means lower acquisition cost at scale, a brand that survives cycle troughs, and — critically — a captive channel for the adjacent title, valuation and settlement businesses that carry better margins than origination itself.

The bounds. Mortgage origination is a commodity product sold on rate and speed; switching costs are essentially zero at the transaction level; and the volume of the addressable market is set by interest rates rather than by anything Rocket does. The franchise determines share; the cycle determines profit. FY21 to FY23 is the demonstration: a 70% revenue decline with no change in competitive position.

Moat: real on customer acquisition, non-existent on pricing, and entirely subordinate to the rate cycle.

11. Capital allocation

• The dominant act of the period was a large equity-funded acquisition — $2,323M of net cash plus roughly 792M shares (38.5% dilution), producing $9,384M of goodwill. Whether this was good allocation is unknowable for at least a full cycle, and it is the single largest open question about the management team.
• No buybacks in FY23, FY24 or FY25 ($0 recorded). FY22 saw $177.7M.
• No current dividend (see §1 — the $0.80 field is stale).
• Capital expenditure $91M (FY25) against $805M (FY24) — an 89% reduction, and the FY24 figure is itself anomalous for an asset-light lender. Flagged as possibly containing capitalised software or acquired assets; not used in any per-share computation.
• Stock compensation $343M (FY25), 5.0% of revenue, up from $145.5M — consistent with an enlarged, acquisition-integrated workforce.

The forward allocation question: with a 1.5% return on equity and $22,898M of book equity, the highest-return use of capital is almost certainly buying back stock at 1.67x book — and none is being done. That absence is informative, and the most likely explanation is that the balance sheet cannot support it, which brings us back to §1.

12. Verdict, kill-criteria and flip conditions

Watch.

There is a real bull case and it should be stated fairly: adjusted earnings per share have beaten consensus in five consecutive quarters and risen from $0.04 to $0.15, revenue has grown 149% in four quarters on an acquisition-enlarged platform, consensus models $0.977 of FY27 EPS putting the stock at 14.0x, and the street's $19.50 target is 42.5% above spot — the widest positive gap in this batch. If the mortgage cycle turns, a larger direct-to-consumer platform captures more of it than the old one would have, and FY28's $1.113 consensus is conservative.

We are not taking that bet, for one governing reason and several supporting ones.

The governing reason: we cannot determine the leverage of a leveraged lender. bal_a reports total debt of $0 and net cash of $2,696M against $60,685M of assets, one year after reporting $13,976M of debt on a balance sheet less than half the size. The same file's enterprise value implies roughly $29,000M of net debt. That is a $32 billion contradiction, it is 82% of the market capitalisation, and it renders every leverage, coverage and enterprise-value metric unavailable. For a mortgage originator — a business whose entire model is funding loans with borrowed money — that is not a peripheral gap. It is the risk assessment itself.

Supporting reasons: a 2.176 beta, the highest in this batch; an Up-C structure that makes every per-share figure ambiguous and allocates most economics away from the public float; $10,611M of untested goodwill at 27.5% of market capitalisation; a 1.5% return on equity; a trailing adjusted-to-GAAP adjustment of roughly $820M annually with no disclosed composition; three consecutive quarters where the vendor's own revenue series disagrees with itself by 7-13%; zero substantive knowledge-base coverage; and twelve-month performance of −7.4% against +19.9% with the price below both a falling 50-day and a falling 200-day average.

Position: none. This is a Watch, and the watching has specific conditions.

Pre-registered conditions to move from Watch to a starter position:

1. A legible balance sheet. A subsequent filing or data refresh showing a coherent total-debt figure, consistent with the enterprise-value calculation, is the necessary condition. Without it there is no position at any price.

2. A reclaim of the 200-day average ($16.51) on rising volume — confirmation that the twelve-month underperformance has ended.

3. A sixth consecutive adjusted-EPS beat on 2026-08-06 (above ~$0.20 against the $0.1642 consensus), extending a streak that is now the strongest operating evidence in the file.

4. Only then, and only at 1-1.5% maximum, given a 2.176 beta.

Pre-registered KILL / do-not-buy criteria:

• Any goodwill impairment. $10,611M is 27.5% of market capitalisation and none of it has seen a downturn.
• Further equity issuance beyond the current ~2,847M diluted count — the share count already rose 38.5% in twelve months, and every new share dilutes the $0.977 directly.
• The adjusted-EPS beat streak breaking — it is the single strongest operating fact in the file and its loss removes the case for the FY27 estimate.
• Revenue growth in FY27 falling below the modelled +10.9% — which would confirm that FY26's +66.7% was purely acquisition arithmetic with nothing behind it.
• Return on equity failing to improve from 1.5% over the next four quarters — the entire investment thesis is the journey to ~10.5%, and no progress would falsify it.
• A rising-rate environment that suppresses origination volumes — FY21 to FY23 saw a 70% revenue decline for exactly this reason.

Pre-registered FLIP TO BUY:

• A clean balance sheet showing net debt below ~$20B with adequate coverage, combined with the 200-day reclaim. This is the primary gate and it is a data gate, not a business gate.
• Two consecutive quarters with return on equity above ~5% — halfway to the consensus destination, and the first hard evidence the enlarged platform earns its capital.
• A buyback authorisation. At 1.67x book with a 1.5% return on equity, repurchase is the obvious highest-return use of capital; management not doing it is currently the most eloquent statement available about the balance sheet's true capacity, and reversing it would be equally eloquent.
• A price below ~$10.50 with the operating trajectory intact — roughly 10.7x FY27E, at which point the multiple compensates for the opacity.

Where RKT fits in the Synthos Framework Portfolio. The financials / housing-cycle sleeve, as a high-beta cyclical option — currently 0%, maximum 1-1.5% if the gates above clear. It is not redundant with anything else in this batch: with a 2.176 beta and a return driven by mortgage rates and origination volumes, it is the only pure interest-rate-cycle exposure here and it correlates with none of the others. Its portfolio function would be convexity, and convexity should be sized small. Logged as a tracked Synthos call (Watch, no position; primary gate is a legible balance sheet) as of 2026-08-04 at $13.68.

Single biggest risk: we cannot determine the leverage of a leveraged lender. A file that reports a mortgage originator with zero total debt, $285M of current liabilities against $60.7B of assets, and an enterprise value implying $29B of net debt is not a file that supports a risk assessment. Everything else in this dive — the beat streak, the acquisitions, the cycle option — is downstream of a number we do not have.

Most fragile assumption in the price: that return on equity travels from 1.5% today to roughly 10.5% by FY27. That is what the consensus $2,406M of FY27 net income against $22,898M of equity requires, it is the entire distance between a $13.68 price and a $19.50 target, and it depends simultaneously on integration of a very large acquisition and on a mortgage cycle that has been at a multi-decade low. Neither has been demonstrated, and one of them is outside management's control.

Provenance and disclosures

• Traceability: 1 knowledge-base hit for "RKT", "Rocket Companies" and "Rocket Mortgage" across 51,928 entries, and it is not a view on this security. a16z (2026-02-09, conviction 68) argues a general thesis about enterprise AI adoption and cites "Rocket Mortgage $40M savings" as one of two illustrative examples; the entity list on the claim is empty, and $40M is 0.6% of FY25 revenue. kb_claim_count: 1, substantive count: 0. No tracked voice has expressed a view on this company, its valuation or its balance sheet. This dive is fundamentals- and quant-driven, conducted under the severe data limitations below.
• Data as-of: fundamentals 2026-03-31 (Q1'26) · estimates, prices, technicals and knowledge-base claims 2026-08-04. All market data drawn exclusively from scripts/deepdive/vti_data/RKT_data.json.
• PRIMARY DATA FAILURE (§1) — the FY2025 balance sheet is rejected in full on the debt and liability lines. bal_a[0] reports totalDebt: 0, longTermDebt: 0, shortTermDebt: 0 and netDebt: −2,696,000,000 (net cash) for a mortgage originator holding $60,685M of assets, against $13,975.8M of total debt reported for FY2024 on a balance sheet less than half the size. It further reports totalCurrentLiabilities: 285,000,000 against those $60,685M of assets. Meanwhile the same file's km_ttm reports enterpriseValueTTM: 67,622,514,037 against a $38,632,514,037 market capitalisation — implying roughly $29,000M of NET DEBT. The contradiction is $31.7B, or 82% of market capitalisation. Consequence: no enterprise-value multiple, no leverage ratio and no coverage ratio appears anywhere in this dive. netDebtToEBITDATTM (16.76x) and interestCoverageRatioTTM (1.246x) are computed from contradictory inputs and are discarded. The valuation in §7 uses earnings multiples only.
• Further vendor data rejected: the dividend fields — profile.lastDividend of $0.80 against ratios_ttm.dividendYieldTTM of 0, dividendPerShareTTM of 0, and simultaneously dividendPayoutRatioTTM of 0.9748 (a 97% payout on a zero dividend). All three cannot hold; RKT is treated as paying no current dividend and the payout field is discarded. Also rejected: ratios_ttm.priceToEarningsRatioTTM of 210.8x, which cannot be reconciled with our own computation of 161.6x on the 2,828.5M economic share count (trailing net income $239.4M ÷ 2,828.5M = $0.0846; $13.68 ÷ $0.0846 = 161.6x).
• Vendor data flagged: earn_cal.revenueActual disagrees with inc_q revenue in three consecutive quarters — Q3'25 $1,562.0M versus $1,790.0M (−12.7%), Q4'25 $2,692.0M versus $2,497.0M (+7.8%), Q1'26 $2,941.0M versus $2,738.0M (+7.4%). inc_q was used throughout, and the 2026-08-06 consensus revenue of $2,811.8M may not be like-for-like with the income statement. The Up-C share count changes presentation mid-series (basic 148M-171M in early 2025 against diluted 2,055M, converging to ~2,828M from Q3'25), making every per-share ratio in the file ambiguous; we use 2,828.5M throughout and label it, noting that the vendor's own market capitalisation already uses the full economic count. seg_prod FY25 sums to $5,459M against $6,880M of revenue — $1,421M unallocated. seg_geo is empty.
• Vendor data checked and found SOUND despite appearances: cf_a operating cash flow of −$3,927M (FY25) is not distress — for a mortgage originator, loans held for sale flow through operating activities, so negative operating cash flow is a volume signal (FY22, a shrinking year, produced +$10,823M for the mirror-image reason). ratios_ttm price-to-free-cash-flow of −27.7x should be ignored rather than interpreted. The income statement itself is clean: FY25 quarterly net income sums to −$68.0M, matching inc_a exactly.
• No counterparty is named for the 2025 acquisitions anywhere in this dive, because no acquisition counterparty appears anywhere in the file. The transformation is described strictly from the observable data (goodwill $1,228M → $10,611M, total assets $24,510M → $60,685M, diluted shares 2,055M → 2,847M, acquisitions net −$2,323M).
• Fair-value method: scenario multiples applied to consensus FY27E adjusted EPS ($0.977, 10 analysts) and FY28E ($1.113, 5 analysts). Bear a housing-freeze EPS of ~$0.50 at 16x, base 15x FY27E (cross-checked at 13x FY28E), bull 20x FY27E (which lands on the street's $19.50). Stated arithmetic only — no discounted-cash-flow model was built, and no enterprise-value method was available.
• Timing caveat: written two days before the Q2'26 print (2026-08-06) on a security with a 2.176 beta. The recommendation is explicitly no position.
• Not investment advice. Independent research, educational and informational only, never personalised.
• Version: 2026-08-04-full.