Rocket Companies, Inc. (NYSE: RKT) — The Bear Case Was Built on an Accounting Mistake, and the Rate Bet Just Paid Without the Rates

Independent equity research · Date: 2026-08-22 · Price: $13.93 (close 2026-08-21) · Market cap: ~$39.4B (2,831.6M Class A + Class L shares) · EV: ~$47.1B ex-non-recourse funding / ~$63.8B headline Prior coverage: my previous analysis of Rocket, published 2026-06-20 at a price of $14.42. This article is an update to it.

⚡ Claude’s Take

This block is the author’s own subjective opinion. It is general information and not investment advice. The analytical body of this article (Sections 1–15 below) takes no position and contains no price target — it discusses valuation only as embedded expectations.

Verdict: upgrade to BUY-on-weakness from the prior HOLD, in a ~$12–14 accumulation zone (roughly 3.1–3.6x tangible book and ~20–23x a ~$0.60 run-rate adjusted EPS), scaling out above ~$19–20. Medium conviction, up from low-to-medium. The call changed. Two of the three quantitative pillars of the prior bear case were wrong, and the company just printed the prior memo’s base case inside the prior memo’s bear rate scenario.

The prior report called Rocket “a levered, negative-tangible-book call on falling mortgage rates.” The strategic architecture in that sentence was right. The accounting was not. Tangible book equity is not −$8.9B; it is +$10.9B, about $3.86 per share. The prior analysis counted the $19.4B mortgage-servicing-rights asset as an acquired intangible and subtracted it. MSRs are a separately-captioned financial asset carried at fair value — and Rocket proved their liquidity this very quarter by selling $53B of unpaid balance for $795M of cash at essentially 100% of carrying value. JPMorgan’s syndicate wrote a minimum-tangible-net-worth covenant into the new $2.5B revolver in July; you do not do that against negative tangible equity. Two related corrections follow: the MSR is carried at 1.56% of owned UPB, not the 0.92% the prior memo computed by dividing into the subserviced book — so the mark is fair and liquid rather than conservative, which kills the hidden-upside leg of the old sum-of-the-parts; and EV/adjusted revenue is ~4.2x on the combined entity’s run-rate, not the “6–8x, richest ever” produced by dividing a post-Cooper enterprise value by a partly pre-Cooper revenue base.

What actually happened since June is better than the setup deserved. The prior memo’s bull-falsification test said rates staying high while earnings stalled near ~$0.20 would break the thesis. Rates did stay high — the 30-year hit 6.8%, a one-year peak, and management conceded “the expected housing recovery in 2026 has not materialized” — and Rocket earned ~$0.61 annualized, above its own ~$0.45 base case. Adjusted EBITDA margin rose 150bp sequentially on a 2.2% revenue decline, which is what synergy capture looks like in a P&L; the Cooper program was raised to $500M and is half-realized. The flywheel stopped being a slide: 57% of refinance closed volume now comes from existing servicing clients, up from 54%, over half of Rocket Loans volume goes to servicing clients, Redfin doubled mortgage leads, and Rocket became the #1 home-equity lender. Meanwhile the competitive landscape broke Rocket’s way in the most violent manner imaginable — UWM lost $452M on a botched hedge from the collapsed Two Harbors deal, suspended its dividend, took a $2.05B Oaktree rescue, and is under a securities investigation. Rocket out-originated it $49.1B to $39.7B. The prior memo’s “UWM out-originates Rocket in dollars” is dead.

So why only medium conviction, and why a zone below the current price? Because none of this makes the stock cheap. At $13.93 you pay 3.6x tangible book and 23x run-rate adjusted earnings for a franchise guiding revenue down ~5.8% sequentially, whose servicing book is shrinking on UPB (customers pay off ~$48.6B a quarter), whose broker-share gains are being bought with a ~100bp pricing incentive worth roughly $170M a year that may not survive its own withdrawal, and whose dominant factor loading is InterestRate at −1.9 — the largest coefficient in every model, meaning this remains, empirically, a levered short position in the rate factor. The framing is no longer “abandoned falling knife.” It is a quality-operator-in-a-wrecked-industry at a full-but-defensible price, with the balance-sheet fear removed and the rate option still free. Three top-tier long-onlys — T. Rowe, Capital Group, Fidelity — filed 13Gs within nine days; this is not an abandoned name. Flip more bullish on evidence that Rocket Pro share holds after the Compass incentive normalizes — that converts a purchased share gain into a franchise gain. Flip bearish on the $320B of >6% note-rate loans failing to recapture when a cutting cycle finally arrives, which would mean the whole $10.6B of goodwill is unsupported. Tag: “They fixed the company; we had mis-read the balance sheet.”

🔄 Changes Since 2026-06-20

Six things moved the thesis in nine weeks. In order of importance:

1. The negative-tangible-book claim was an error and is corrected. Goodwill $10,611M + intangibles $1,995M = $12,606M against $23,546M of equity leaves +$10,940M of tangible equity (~$3.86/share), confirmed directly in XBRL. The prior figure of −$8.9B added the MSR asset into intangibles. This removes the “no equity cushion” risk that ran through the previous article’s opening Take, Executive Summary, Financial Quality, Capital Allocation, risk matrix and valuation.
2. Q2-2026 beat the prior memo’s base case in the prior memo’s bear environment. Adjusted EPS $0.16 (H1 $0.30, ~$0.61 annualized) versus a ~$0.45 base and ~$0.20 bear, with 30-year mortgage rates at a one-year high of 6.8%.
3. UWM — the #1 wholesale originator — was impaired. A $451.9M Q2 loss from an over-hedged position left over from the failed Two Harbors acquisition, dividend suspended, $2.05B Oaktree/Ishbia recapitalization, securities investigation, stock −71% over twelve months. Rocket now out-originates it.
4. The flywheel produced its first hard number. 57% of refinance closed volume from existing servicing clients (54% in Q1); record purchase share 6.2% and refi share 14.3%; servicing fully migrated onto one platform.
5. A deliberate capital-light MSR rotation. $53B of UPB sold for $795M at ~carrying value, subservicing and recapture retained on ~80%, concentrating the book in the $320B of >6% note-rate loans that carry the actual refi option. Total portfolio fell $2.1T → $2.0T by design.
6. Balance-sheet housekeeping. $1.5B of notes issued (7x oversubscribed, upsized) to retire the 2.875% 2026s; weighted-average coupon up 5.03% → 5.43% but net corporate leverage down to 0.9x, −20% since YE25; a new undrawn $2.5B revolver to 2029; $11.2B total liquidity.

Two things did not change: the revenue trajectory is still deteriorating (Q3 guided $2.5–2.7B, −5.8% sequentially at the midpoint), and the stock is still a rate instrument (InterestRate loading −1.9).

📈 Stock Price Action — Five-Year Event Map

Factual price history — no recommendation, no price target.

Rocket has round-tripped an entire rate cycle and is now grinding in the lower half of a wide post-deal range. From an August-2020 IPO at $18, it spiked to a meme/refi-boom peak of $41.60 unadjusted (~$34.95 split-and-dividend-adjusted) on 2 March 2021, collapsed to $5.94 in October 2022 as mortgage rates went from ~3% to ~7%, ground through three years of “higher-for-longer,” re-rated to a 52-week high of $23.44 on 15 January 2026 as the Redfin and Mr. Cooper integrations took hold, and has since fallen to $13.93 — below its IPO price, ~60.1% off the adjusted all-time high, −29.9% year-to-date, and in the lower third of a 52-week $12.35–$23.44 range, trading below its 21-day ($14.11), 50-day ($14.15) and 200-day ($15.31) EMAs. The last three months have been approximately flat (+4.1% annualized), not falling — the decline is a six-month story that has stopped.

Cycle narrative. Events 1–3 are one rate story told twice: the COVID refi mania compounded by a retail squeeze, then the fastest hiking cycle in forty years erasing it, an 86% peak-to-trough drawdown. Event 4 is three years of purgatory in which every dovish hint produced a rally that died. Events 5–6 are the strategic pivot — the market first punished the all-stock Redfin and Mr. Cooper dilution, then rewarded the #1-originator-plus-#1-servicer scale story into a January 2026 high of $23.44. Event 7 is this year’s unwind, and the factor model attributes it cleanly: the InterestRate factor rose +7.01% over 126 days (z = +1.32) against RKT’s −1.9 loading, which on the model’s own arithmetic accounts for essentially the whole six-month decline. Event 8 is the most informative leg for this update. On 6 August Rocket fell 4.6% ($13.86 → $13.22) — the session after UWM’s blow-up and before the market could digest Rocket’s own release, which was issued after the close; on 7 August, the first full session to price the print, it rose 3.8% to $13.72. Sector contagion down, own-results recovery up. Each price move is a fact; each attributed driver is interpretation, cross-referenced to earnings releases, 8-K filings, the rate path and the factor-return series.

1. Executive Summary

Rocket Companies is the Detroit parent of Rocket Mortgage, the largest US retail mortgage originator, and — since the October 2025 Mr. Cooper close — the largest US mortgage servicer, plus the owner of Redfin. This is an update to coverage initiated 2026-06-20. The prior report’s strategic reading was sound; three of its quantitative conclusions were not, and the second quarter delivered a result the prior framework did not anticipate.

The correction first, because it is load-bearing. The prior memo’s central balance-sheet claim — negative tangible book equity of ~−$8.9B, “the only large mortgage franchise with negative tangible equity” — was produced by treating the $19.4B mortgage-servicing-rights asset as an acquired intangible. It is not one. Against $23,546M of equity sit $10,611M of goodwill and $1,995M of intangibles, leaving +$10,940M of tangible equity, roughly $3.86 per share (confirmed independently in XBRL). Two corollary corrections follow. The MSR is carried at 1.558% of owned UPB ($18,905M on $1,213,607M), not the 0.92% the prior memo obtained by dividing into a $2.1T total that includes $803B of subserviced balances on which Rocket owns no asset — so the mark is at, not below, the bulk market, and the Q2 sale of $52,926M of UPB for $795M cleared at 1.502%, validating the mark as fair and liquid while eliminating the prior sum-of-the-parts’ embedded upside. And EV/adjusted revenue is ~4.22x on run-rate (~4.87x TTM), not “6–8x and near richest ever,” a figure the prior memo obtained by dividing a post-Cooper enterprise value into a partly pre-Cooper revenue base.

Then the quarter. Q2-2026 produced total revenue of $2,784M, GAAP net income of $229M, adjusted net income of $441M, adjusted EBITDA of $766M and adjusted diluted EPS of $0.16 — with adjusted EBITDA margin expanding 150bp sequentially (26.2% → 27.7%) on a 2.2% sequential revenue decline, the signature of cost synergies rather than volume leverage. The Cooper synergy program was raised from $400M to $500M and is half-realized. This happened while 30-year mortgage rates reached 6.8%, the highest in more than a year, and management conceded that “the expected housing recovery in 2026 has not materialized.” The prior memo’s bull-falsification test required that a high-rate environment hold normalized adjusted EPS near ~$0.20; the actual run-rate is ~$0.61, above the prior memo’s own ~$0.45 base case. The test was administered under adverse conditions and passed.

The flywheel acquired its first hard number. 57% of refinance closed volume came from existing servicing clients, up from 54% in Q1. Purchase share hit a record 6.2% (from 5.5% in Q4-25) and refi share 14.3% (from 12.2%); over half of Rocket Loans’ near-doubled volume went to servicing clients; Redfin doubled mortgage leads year-over-year; and Rocket became the #1 US home-equity lender. Servicing migrated fully onto one platform. Simultaneously the company executed a deliberate capital-light rotation, selling $53B of low-coupon MSR UPB — loans that would never refinance under any plausible rate path — for $795M of cash while retaining subservicing and recapture on ~80%, concentrating the remaining book in the $320B of loans with note rates above 6% that constitute the actual refinance option. Total portfolio fell $2.1T → $2.0T by design, not by attrition.

And the industry broke Rocket’s way. On 5 August UWM Holdings reported a $451.9M quarterly loss caused by an over-hedged position left standing after its Two Harbors acquisition collapsed, suspended its dividend, announced a $2.05B Oaktree/Ishbia recapitalization, and drew a securities investigation; its shares fell 34.78% in a session and are −71% over twelve months. Rocket originated $49.1B against UWM’s $39.7B — a reversal of dollar-volume leadership. Rocket’s aggressive 0.69% Rocket Pro gain-on-sale margin, explicitly a Compass-partnership pricing incentive, is being spent precisely while the channel’s incumbent price-setter cannot respond.

What has not improved is the top line and the price. Q3 adjusted revenue is guided to $2.5–2.7B, a ~5.8% sequential decline at the midpoint. The servicing book shrinks by ~$48.6B of customer payoffs per quarter. The Compass incentive costs roughly $170M annualized and has not been tested against its own withdrawal. And at $13.93 the equity trades at 3.61x tangible book, 23.0x run-rate adjusted EPS and 15.7x EV/adjusted EBITDA — a full price. The dominant factor loading remains InterestRate at −1.9, the largest coefficient in every nested model: this is still, empirically, a levered long-duration instrument. The thesis is no longer about balance-sheet fragility. It is about whether a demonstrably superior operator, now trading at a fair-to-full multiple with a free rate option and an impaired principal competitor, is worth owning before the cycle turns.

2. Business Overview

Rocket is a vertically integrated residential-real-estate platform that, as of the second quarter of 2026, reports itself differently than it did three months ago — a change worth noting up front because it reduces disclosure.

The new segment structure. Beginning in Q2-2026 the company reports one segment, “Mortgage,” encompassing origination, servicing, title, closing and appraisal, with real estate (Redfin) and personal finance (Rocket Money, Rocket Loans) collapsed into “All Other.” This replaces the prior Direct-to-Consumer and Partner Network segments. The stated rationale is that origination and servicing “are connected by our recapture engine,” which is analytically fair — the whole thesis is that these are one business, not two. But the practical effect is that the DTC-versus-Partner contribution-margin comparison the prior memo relied on no longer exists in the financial statements. Channel gain-on-sale margins are still disclosed in the earnings release highlights; segment profitability by channel is not. Treat this as a transparency downgrade even where the logic is defensible. In Q2 the Mortgage segment produced $2,274M of revenue, $2,251M of adjusted revenue and $1,174M of contribution margin (52.2% of adjusted revenue); All Other produced $510M of revenue and $181M of contribution margin (35.5%).

(1) Mortgage origination (gain-on-sale). Rocket originates residential loans and sells them, overwhelmingly into the agency channel, booking the spread. Q2-2026: $49.1B of total closed volume and $47.0B of net rate locks, at a blended gain-on-sale margin of 2.48%. The channel detail matters more than the blend:

Direct-to-Consumer at 4.13% is the economic engine and is holding up. Rocket Pro at 0.69% has collapsed from the ~1.08% the Partner Network earned in FY2025 — deliberately, as a Compass pricing incentive (see Capital Allocation). Correspondent at 0.19% is essentially a volume-and-servicing-acquisition channel, not a margin business. Ex-correspondent margin of 3.11% is down from 3.22% in Q1.

(2) Servicing. The centre of gravity. Total servicing portfolio UPB of $2,016,711M across 9.1 million loans at 30 June 2026, split between $1,213,607M of owned MSR and $803,104M of subserviced-and-other. Servicing fee income was $1,066M in Q2 (versus $401M in Q2-25 pre-Cooper). The MSR asset is carried at $18,905M, or 1.558% of owned UPB, marked to fair value quarterly. Servicing is recurring, counter-cyclical and the fuel for recapture — and the distinction between owned and subserviced UPB, missed in the prior memo, is essential to valuing it.

(3) Real estate and adjacent services. Redfin contributes the home-search portal and brokerage; Amrock/Rocket Close provides title, valuation and settlement; Rocket Homes is the referral network. In June, Redfin doubled mortgage leads year-over-year and hit record attach rates, aided by an expanded “Preferred Pricing” offer under which eligible Rocket servicing clients can save up to $20,000 buying and selling with a Redfin agent and financing through Rocket.

(4) Consumer fintech. Rocket Money (subscription personal finance) and Rocket Loans (personal lending). Rocket Loans nearly doubled volume year-over-year in H1-2026 and hit a record month in June, with more than half of those loans made to Rocket servicing clients — a small but clean demonstration that the customer graph monetizes across products.

(5) The AI layer, which is now producing measurable operating results rather than slogans. AI Voice handled more than one million inbound servicing calls within three months of launch; over 50% would otherwise have required a human; task resolution was ~25% faster than traditional IVR with client satisfaction at 4.5/5. Loan officers using the new AI pipeline-prioritization tools are handling ~40% more clients than a year prior. Broker partners adopting the Jupiter loan-origination system and Navigate AI are growing applications and closings at five times the pace of those who do not.

Verdict on the model. The architecture the prior memo described as “coherent but brand-new and unproven” has, in one quarter, produced four independent quantified traces of the cross-sell working: 57% of refis from the servicing book, >50% of personal loans to servicing clients, doubled Redfin mortgage leads, and #1 in home equity. The pieces genuinely connect. What remains unproven is not the routing but the volume — whether this machine, pointed at a falling-rate cycle, converts at the scale the valuation implies. The single-segment reporting change makes verifying that harder, not easier.

3. Industry Dynamics

The prior memo’s structural read — two industries of opposite quality joined at the loan — stands. What changed is the state of the cycle and, dramatically, the health of the principal competitor.

Origination remains structurally bad, and 2026 proved it again. The industry has no entry barriers, a commodity product, near-zero borrower switching costs and severe fragmentation. Management’s own framing on the Q2 call — the 30-year fixed “reached 6.8% in recent weeks, the highest level in more than a year,” and “the expected housing recovery in 2026 has not materialized” — describes an industry in its third consecutive year of famine. Rocket guided Q3 revenue down ~5.8% sequentially into it. In Marathon’s capital-cycle terms this is the correct phase to be acquiring capacity, which is exactly what Rocket is doing with the Compass incentive.

Servicing remains structurally better — and Rocket demonstrated its liquidity. Servicing is concentrated, capital-heavy, recurring, scale-rewarding and a natural rate hedge. The hedge worked visibly this half: the 10-Q attributes a +$349M change in MSR valuation model inputs in H1-2026 (versus −$286M in H1-2025) to “an increase in interest rates during the current period.” When origination got harder, the servicing asset got more valuable — precisely the counter-cyclicality the 2025 deals were bought for. The Q2 bulk sale at 1.502% of UPB, essentially at carrying value, also establishes that this asset is not a mark-to-model abstraction: it converts to cash on demand at book.

The competitive landscape changed violently. On 5 August 2026 UWM Holdings — the wholesale channel’s dominant price-setter and Rocket’s principal rival — reported Q2 revenue of $888.0M, a net loss of $451.9M, adjusted EBITDA of $185.9M and origination volume of $39.7B. It suspended its quarterly dividend and announced a $2.05B equity investment by Oaktree Capital Management and SFS Group Capital LLC (an Ishbia-family vehicle). UWM’s CEO attributed the loss to being “over-hedged” on a position left standing when its Two Harbors acquisition fell through. Shares fell 34.78% in one session (to $1.20) and closed at $1.40 on 21 August, −71.0% over twelve months. Kaplan Fox & Kilsheimer announced a securities-law investigation.

The implications run three ways and should be weighed honestly:

• For Rocket, positive. Volume leadership in dollars flipped — $49.1B versus $39.7B in the same quarter. The prior memo’s statement that “UWM out-originates Rocket in dollars” is no longer true. More importantly, a wholesale channel whose price-setter is capital-constrained, dilutively recapitalized and legally distracted is a channel where a well-capitalized entrant can buy durable share. That is the strategic logic of Rocket Pro’s 0.69% margin.
• For the sector, a warning. UWM’s loss was a hedging accident on an MSR-adjacent book, not an operating loss. Rocket runs an $18.9B MSR book with its own hedge program and reports “change in fair value of MSRs and related liabilities due to valuation assumptions (net of hedges)” as a single netted line. The tail risk UWM demonstrated is present, in kind if not in degree, at Rocket.
• For the market’s read of Rocket, mixed. RKT fell 4.6% on 6 August in sympathy before its own results were digested. Sector contagion is a live cost of being classified with UWM — and the factor model confirms UWMC is now RKT’s single closest factor neighbour at 0.915 similarity.

Peer scoreboard (price, 21 August 2026). RKT −3.4% since the prior report and −20.8% over twelve months; UWMC −36.9% and −71.0%; PFSI −8.5% and −26.4%; Zillow +13.2% and −52.8%. Rocket outperformed every mortgage comparable over both windows.

The structural tailwind: bank retreat. Post-crisis capital rules continue to push banks out of origination and servicing toward non-banks. Unchanged, and now compounded by UWM’s impairment: the non-bank field itself is consolidating toward the strongest balance sheet.

The structural swing factor: GSE reform. Fannie and Freddie remain in conservatorship with the administration signalling interest in release. Higher guarantee fees would raise mortgage rates and dent volumes; Rocket’s scale confers relative pricing advantage. Timing remains uncertain and the market treats it as a live overhang. Nothing in the Q2 disclosures advances this either way.

Verdict: still a structurally mixed-to-bad industry — but Rocket’s relative position within it improved materially. Origination is as commodity and cyclical as ever, and 2026 is a genuinely bad vintage. Servicing is better and Rocket proved both its counter-cyclicality and its liquidity this quarter. The decisive change is that the industry’s second-largest player just impaired itself in a way that is not quickly reversible, at the exact moment Rocket has $11.2B of liquidity and 0.9x net corporate leverage with which to take share. A bad industry in which your main competitor breaks is a materially better place to be than a bad industry with two healthy giants.

4. Competitive Position

Does Rocket have a durable competitive advantage? The prior memo answered “narrow and emerging, not wide.” Nine weeks later the evidence is stronger and the answer should be upgraded to narrow but now partially demonstrated — with the demonstration confined to the routing mechanism, not yet the volume it produces.

The recapture flywheel produced its first hard number. The mechanism was always specific: owning the servicing relationship gives Rocket a refinance shot at its own customer at near-zero acquisition cost. The prior memo could only cite management’s ~83% recapture claim and called validation “the single most important unproven datapoint.” Q2 supplied a measurable, sequentially improving statistic from the CFO: “existing servicing clients accounted for 57% of refinance closed volume, up from 54% in Q1.” A majority of Rocket’s refinance business is now sourced from its own book, and the share is rising quarter over quarter. Management adds that “recapture rates on the Mr. Cooper portfolio reached another record.”

Be precise about what this does and does not prove. 57% is a mix statistic, not a retention rate. It tells you where refinance volume originated, not what fraction of departing customers were kept. The retention question is answered less flatteringly by the balance sheet (below). But as evidence that the routing works — that owning servicing genuinely feeds origination at scale — it is the strongest datapoint the company has produced, and it did so in a quarter with almost no refinance market to work with.

The corroborating cross-sell traces. Three further independent signals point the same way: over half of Rocket Loans’ near-doubled H1 volume went to Rocket servicing clients; Redfin doubled mortgage leads year-over-year in June with record attach rates; and Rocket became the #1 US home-equity lender, the first independent mortgage company to lead that category, with >250,000 homeowners and >$24B of equity accessed since a mid-2022 launch. Four different products, one customer graph. This is what demand-side captivity looks like when it is working, and it is no longer a slide.

Scale and cost advantage (Greenwald: supply-side). The AI operating results are now quantified and material: AI Voice absorbed >1 million inbound servicing calls in three months, with >50% otherwise requiring a human and ~25% faster resolution; loan officers using AI prioritization handle ~40% more clients than a year ago. The 150bp of sequential adjusted-EBITDA-margin expansion on declining revenue is the financial trace. The completion of “one of the largest servicing migrations in industry history” — the entire portfolio now on a single platform — is the enabling infrastructure and a genuine execution achievement given how many servicing transfers go badly.

The counter-evidence, which is real. Three things cut against the moat claim:

1. The book is shrinking. Customers voluntarily paid off $48,597M of UPB in Q2 alone (~9.6% of the $2.0T book annualized) against $43,264M of MSR added from originations. Rocket is roughly treading water on retention at the portfolio level even as the mix statistic improves. If recapture were the overwhelming force management describes, the book would compound; it does not.
2. The Rocket Pro share is purchased. A 0.69% gain-on-sale margin against ~1.08% historically is share bought with roughly 39bp of foregone margin — about $43M per quarter, ~$170M annualized on Q2 volumes. Purchased share is not moat evidence until it survives the withdrawal of the purchase.
3. Returns still fail the franchise test. Greenwald’s marker for a genuine barrier to entry is sustained 15–25% ROIC with stable market share. Rocket’s returns remain boom/bust, its share is rising (which is good but is the opposite of the stability signature), and normalized returns on the ~$23.5B equity base are mid-single-digit. On the strict test, this is not yet a franchise.

Versus the competitive set. The comparison improved sharply on one axis and not at all on another. Against UWM, Rocket now leads in dollar volume ($49.1B vs $39.7B), is vastly better capitalized, and faces an opponent that has suspended its dividend, taken rescue capital and drawn a securities probe — this is the best relative position Rocket has occupied versus its wholesale rival in the company’s public history. Against PennyMac, Rocket is far larger in servicing and origination but PFSI runs the more conservative balance sheet. Against Zillow, on the purchase-search funnel, Rocket is still the challenger — Redfin’s doubled leads are growth off a smaller base, and Zillow’s shares (+13.2% since June against Rocket’s −3.4%) suggest the market still gives Zillow the funnel.

Verdict: a narrow moat, now partially demonstrated rather than merely asserted — resting on a genuine and rising recapture mix, an AI-driven cost advantage with measurable output, and an impaired principal competitor — but still unvalidated on the metric that matters most (portfolio retention) and still resting on a commodity core. The prior memo’s core scepticism was correct in kind. The evidence has moved in the company’s favour in degree, and it moved in the hardest quarter available to move in. It has not moved far enough to call this a franchise.

5. Growth History and Forward Opportunities

History. Rocket’s organic growth history is the rate cycle: revenue ~$15.5B (2020) and ~$12.7B (2021) at the peak, collapsing to ~$3.6B (2023), recovering to ~$4.9B (2024) and ~$6.3B GAAP / $6.9B adjusted (2025, M&A-aided). H1-2026 adjusted revenue of $5,583M annualizes to ~$11.2B — but that is the combined entity’s first clean half, not an organic growth rate. The acquired step-change is now fully in the base.

The current trajectory is downward, and this is the update’s principal negative. Sequential adjusted revenue: Q4-25 $2,440M → Q1-26 $2,822M → Q2-26 $2,761M (−2.2%) → Q3-26 guided $2,500–2,700M (−5.8% at the midpoint). Two consecutive sequential declines, guided into a third. This is what a 6.8% mortgage rate and a failed housing recovery do to an origination platform, and it is happening despite record market share. The share gains are real; the market they are taken from is shrinking faster.

The organic signals worth weighting are genuinely good, and they are share signals. Purchase market share rose to a record 6.2% from 5.5% in Q4-25; refinance share to a record 14.3% from 12.2%. Both in “one of the toughest spring housing markets in years,” per the CEO. Rocket became #1 in home equity. Rocket Loans nearly doubled. Redfin doubled mortgage leads. In a business where the cycle is exogenous, taking two points of purchase share in three quarters is the only growth metric management controls, and it is being delivered.

Forward opportunities, reordered by how the evidence has moved:

1. The rate-cut refi wave — now with a sharpened instrument. Still the largest lever, still exogenous. What changed is the quality of the book pointed at it. Rocket deliberately sold $53B of low-WAC MSRs — loans whose note rates are so far below market that they will never refinance under any plausible path — and retained the $320B of loans with note rates above 6%. That $320B is the live refinance inventory. A 100–150bp decline in mortgage rates puts a large fraction of it in the money against an origination machine that already sources 57% of refi volume from its own book.
2. Synergies — raised, not trimmed. The Cooper cost program went from $400M to $500M, with “an additional $100 million in annualized savings identified for the first half of 2027,” and is half-realized at the end of Q2. The 150bp of sequential margin expansion on falling revenue is the independent corroboration.
3. Wholesale share against an impaired UWM. The Compass “Power Play” (up to 100bp of stacked pricing credit) has produced >$2B of net rate lock volume since inception in February 2026, plus same-business-day conditional approvals and a 12-business-day clear-to-close commitment. Brokers adopting Jupiter and Navigate AI grow 5x faster than those who do not — a retention mechanism that could outlast the price incentive.
4. Cross-sell attach. Redfin’s doubled leads, the up-to-$20,000 Preferred Pricing offer, Rocket Money, Rocket Loans and title.
5. Continued bank retreat and now non-bank consolidation.

Verdict: MIXED-quality growth, with the mix improving. The revenue line is going the wrong way and will continue to for at least one more quarter — that is the honest headline and it is why the stock is where it is. But the composition of what growth exists has improved materially: record share in both purchase and refi, four working cross-sell channels, a servicing book deliberately concentrated into its most refinanceable cohort, and a synergy program that was raised. The prior memo asked whether the flywheel and synergies would validate the ~19x dilution and said credit could only accrue in 2026–27. Two quarters of 2026 are in, and they have accrued credit — on margin and share, not yet on revenue.

6. Financial Quality

Earnings quality is low in form and improving in substance. GAAP remains dominated by MSR marks: Q2 GAAP net income of $229M against adjusted net income of $441M, with the bridge running through $90M of share-based compensation, $99M of acquisition-related expenses, $112M of acquired-intangible amortization, a $28M litigation accrual and a −$23M MSR valuation adjustment. GAAP diluted EPS of $0.08 versus adjusted $0.16. Trailing GAAP P/E is ~115x on ~$0.12 of TTM GAAP EPS and remains uninformative. Use adjusted EPS and adjusted EBITDA. The one improvement worth noting is that GAAP is no longer negative — H1-2026 GAAP net income of $526M compares with an H1-2025 loss of $178M and a FY2025 loss of $234M — so the gap between GAAP and adjusted is narrowing as deal costs roll off.

The operating result, and why the margin matters more than the revenue.

Adjusted EBITDA rose 3.8% sequentially while adjusted revenue fell 2.2% — 150bp of margin expansion. On a fixed-cost origination platform, margin expansion into declining revenue does not happen by accident; it is cost coming out. That is the synergy program showing up where it can be verified, and it is the single most encouraging line in the quarter. Total expenses of $2,503M against $2,784M of revenue leave a 10.1% GAAP pretax margin; the Mortgage segment’s contribution margin was $1,174M on $2,251M of adjusted revenue (52.2%).

Cyclicality — the annuity is working as designed. The prior memo’s key structural claim was that the servicing book would dampen the violent cyclicality of gain-on-sale. H1-2026 is the first real test and it passed: servicing fee income of $2,149M for the half (versus $802M in H1-2025), and a +$349M swing in MSR valuation model inputs driven explicitly by “an increase in interest rates during the current period” — against −$286M in H1-2025 when rates fell. Rates rose, origination suffered, and the servicing asset appreciated. Interest income of $583M in Q2 (versus $237M) is largely custodial-deposit income on the servicing portfolio — a second, less-discussed rate hedge: higher short rates earn more on escrow float.

The tangible-book correction

The prior memo stated that goodwill plus intangibles of ~$32.1B against ~$23.2B of equity left tangible equity of ~−$8.9B, and built a substantial part of its risk framing on it. That is wrong. From the 30 June 2026 balance sheet:

Independently confirmed in XBRL: us-gaap:Goodwill = $10,611M, us-gaap:IntangibleAssetsNetExcludingGoodwill = $1,995M, us-gaap:StockholdersEquityIncludingPortionAttributableToNoncontrollingInterest = $23,546M at 30 June 2026. The arithmetic of the prior error is transparent: $32,100M − $10,611M = $21,489M, which is $19,442M of MSRs at fair value plus $2,109M of true intangibles. The MSR asset was counted as an acquired intangible and subtracted.

MSRs are a separately-captioned financial asset carried at fair value, not an acquired intangible; every comparable mortgage franchise (PFSI, COOP, UWMC) includes them in tangible book. Three independent confirmations that the corrected figure is the right one: (a) the balance sheet itself captions them separately; (b) Rocket sold $52,926M of MSR UPB for $795M of cash in Q2 at ~100% of carrying value, which is the definition of a realizable asset; and © the new $2.5B revolver signed 16 July 2026, with JPMorgan as administrative agent, requires the Company “to maintain minimum liquidity and tangible net worth” — a covenant no syndicate writes against negative tangible equity.

This changes the risk profile, not the valuation verdict. Tangible equity is +$10.9B and grew $877M in the half. But at $13.93 the stock trades at 3.61x tangible book — which is a full multiple for a mortgage franchise, not a cheap one. The correction removes a solvency fear; it does not create value.

The MSR capitalization-rate correction

The prior memo computed the MSR at “~0.92% of UPB” and concluded, against a 1.0–1.5% bulk market, that the asset was conservatively marked and worth “$19–26B.” That calculation divided the MSR asset by the total servicing portfolio, which includes $803,104M of subserviced balances on which Rocket owns no MSR at all.

At 1.558% of owned UPB the MSR is marked at or slightly above the bulk market, not below it. And the Q2 sale — $795M on $52,926M of UPB = 1.502% — is a market-clearing print at essentially carrying value. The mark is validated as fair and liquid, which is genuinely valuable evidence; but there is no embedded gain, and the prior memo’s $19–26B upper bound should come down to approximately the carried $18.9B.

Balance sheet and leverage. Total assets $60,977M. Cash $3,103M (from $2,696M). Loans held for sale $15,548M. Advance receivables $1,542M (down from $2,040M — a working-capital release). Against these: secured financing $16,639M (down from $17,936M; largely non-recourse warehouse plumbing backed by the loans held for sale) and unsecured financing, net $10,772M. Senior notes principal is $10,805M versus $10,455M at YE25, but the weighted-average coupon rose from 5.03% to 5.43% after the June refinancing swapped 2.875% 2026 paper for 6.125%/6.500% money. Total liquidity is $11.2B ($3.1B cash + $2.3B undrawn credit lines + $5.8B undrawn MSR/advance lines). Management reports net corporate leverage of 0.9x, down ~20% since YE2025. The 2027 convertible ($503M at 0.500%, conversion price ~$118) is deeply out of the money and will be settled in cash at par.

Note the nuance the prior memo missed: absolute corporate debt is flat-to-slightly-up and more expensive, while leverage measured against earnings fell sharply because EBITDA grew. Both are true. The refinancing cost roughly $43M a year of incremental interest; it also removed the 2026 maturity wall entirely.

Returns. GAAP ROE annualizes to ~4.5% on H1-2026; return on tangible equity to ~15.8% on the annualized adjusted net income of $1,726M against $10,940M of tangible equity — a respectable figure, though it is flattered by measuring an adjusted numerator against a tangible denominator and by an unusually favourable MSR mark. Through-cycle, returns remain boom/bust.

Verdict: MIXED financial quality, upgraded from the prior NEGATIVE-to-MIXED. The upgrade rests on three things: the balance sheet is not fragile (tangible equity +$10.9B, leverage 0.9x, $11.2B liquidity, an untested-but-real minimum-TNW covenant); the MSR asset is provably liquid at book; and the cost structure is demonstrably improving (150bp of sequential margin on falling revenue). What keeps it from GOOD: GAAP earnings are still MSR-mark noise, the top line is declining, adjusted metrics strip $211M of Q2 items including real cash costs like $90M of share-based compensation, and through-cycle returns on the origination core remain poor.

7. Capital Allocation

Capital allocation remains the crux, and the quarter delivered the first genuinely skilful capital-allocation decision of the post-merger era alongside one that is deliberately expensive.

The MSR rotation — the good decision. Rocket sold $53B of UPB for $795M of cash, retaining subservicing and recapture services on nearly 80% of what it sold. The CFO was explicit about the selection criterion: “We focused on the low WAC MSRs, and the good news is something to the tune of 80% of those sales went to our partners… We will still be the subservicer of those loans, and we will still collect the recapture economic on those loans too.”

This is better than it looks at first glance. Low-WAC (low note-rate) MSRs are the loans that will never refinance under any plausible rate path — pure long-duration fee annuity with zero embedded refi option. They are also, precisely because they never prepay, the most highly valued per unit of UPB. Rocket sold the dead duration at a full price, kept the servicing relationship and the recapture right via subservicing, converted the proceeds to cash, and concentrated the remaining owned book in the $320B of loans with note rates above 6% that constitute the actual refinance call option. The headline “$2.1T → $2.0T, 9.5M → 9.1M loans” reads as deterioration to anyone anchoring on UPB. It is the opposite: the portfolio got smaller, lighter, and more optioned.

The Compass incentive — the expensive decision. Rocket Pro’s gain-on-sale margin fell to 0.69% from ~1.08% historically. The company attributes this to “investments in the Compass partnership,” and the CFO did not hedge: “We are offering a pricing incentive. Which is the right thing to do when you enter a big partnership, and we need to get these Compass agents excited.” The “Power Play” initiative launched in February 2026 offers brokers up to 100 basis points of stacked pricing credit, and has generated >$2B of net rate lock volume since inception.

Quantify it: ~39bp of foregone margin on $11.1B of quarterly Rocket Pro closings is roughly $43M per quarter, ~$170M annualized — real, discretionary, and material against $1,726M of annualized adjusted net income. The defence is timing: this incentive is being spent precisely while UWM, the channel’s price-setter, is capital-constrained and legally distracted. Buying share from a wounded incumbent at the bottom of a capital cycle is textbook Marathon behaviour. The test is durability — purchased share is not franchise share until it survives the withdrawal of the purchase. The Jupiter/Navigate AI adoption data (partners using them grow 5x faster) hints at a stickiness mechanism beyond price, but it is a hint.

The M&A scorecard, two quarters on. The 2025 Redfin (~$1.75B) and Mr. Cooper (~$14.2B closed) all-stock deals cost ~19x share dilution (~150M → 2,831.6M shares). Two quarters of combined operation now exist to judge them by:

The first four have gone the company’s way. The fifth is the one that matters and it will take a rate cycle to settle.

Returns of capital — still absent. No dividend, no buyback, and neither was discussed on the Q2 call. The dormant $1B authorization (~$312M used) remains dormant. With the stock below its IPO price, tangible book at $3.86 and net corporate leverage at 0.9x, the absence of any repurchase is a legitimate criticism — management is choosing balance-sheet capacity and growth investment over per-share accretion, which is defensible in a cyclical trough but means holders underwrite growth rather than collect it.

Governance — modestly improved. The board expanded from nine to ten on 17 August 2026 with the appointment of Sarah Watterson as an independent Class III director: Goldman Sachs, then Managing Director at Fortress Investment Group evaluating and managing businesses across financial services including mortgage origination and servicing, currently President of 3 Star Sports & Entertainment and previously President of Brightline West. Compensation is standard ($75,000 cash retainer, $215,000 initial RSU grant). This is the first director with direct MSR/mortgage capital-markets underwriting experience — competence-matching to a balance sheet that now carries an $18.9B MSR book, not window dressing. Dan Gilbert / Rock Holdings retain effective control; related-party flows (Note 7) are unchanged in character and immaterial in size — $17M of Q2 lease expense to Bedrock affiliates, $13M of related-party G&A, $4M of marketing, $6M of notes receivable. The Tax Receivable Agreement remains in place.

Insider behaviour — neutral, and worth stating plainly. The Form 4 sweep for 1 June to 22 August 2026 returns six filings, all directors, and zero open-market purchases (code P) and zero open-market sales (code S). Five are routine post-annual-meeting RSU grants (code A: Rampell 16,312; Mariner 16,312; Shank 16,312; Olson 11,255; Watterson 14,566 on appointment); one is a director option exercise (Rizik, codes M/J, 12,261 shares at $12.65, aff10b5One = 0). No conviction buying despite a sub-IPO price; equally, no distribution. Score this as genuinely neutral, not as a signal in either direction.

Institutional sponsorship — a new and notable datapoint. Three top-tier long-only managers filed Schedule 13Gs within nine days in August 2026: Capital World Investors 69,428,716 shares (12 Aug), T. Rowe Price Associates 55,268,794 shares (14 Aug), and FMR LLC / Abigail Johnson 52,657,948 shares (6 Aug) — together ~177.4M shares, roughly 9.3% of the 1,906.4M Class A shares outstanding. This is meaningful active-manager sponsorship in a stock trading below its IPO price and argues directly against the prior memo’s “abandoned name” characterization.

Verdict: upgraded to MIXED-POSITIVE from the prior MIXED. The MSR rotation is a genuinely skilful, value-additive decision that the prior framework would not have predicted. Synergies were raised and are half-banked. Leverage fell. The board added relevant expertise. Against that: the Compass incentive is a real ~$170M annual cost of unproven durability, there is still no return of capital, incentive metrics still do not gate on per-share value, and the ~19x dilution remains unvalidated on the only test that counts. The jury is still out — but two quarters of evidence have come in, and most of it favours management.

8. Changes and Headwinds — Last Two Years

The Krishna-era pivot (2023→) has moved from strategy to measurable output. The AI/automation programme now shows in verifiable operating statistics — >1 million AI-handled servicing calls in three months, ~40% more clients per loan officer, brokers on Jupiter/Navigate AI growing 5x faster — and in the P&L, as 150bp of sequential margin expansion on falling revenue.

The two megadeals are integrated faster than planned. Redfin (July 2025) and Mr. Cooper (October 2025). Q2-2026 saw the completion of “one of the largest servicing migrations in industry history,” putting all servicing clients on a single platform. The Cooper cost synergy target was raised to $500M and is half-realized.

A new segment structure reduces disclosure (Q2-2026). Reporting collapsed from Direct-to-Consumer / Partner Network to a single Mortgage segment plus All Other. Channel gain-on-sale margins survive in the release; channel profitability does not. Prior periods were recast. Defensible on the logic that origination and servicing are one recapture-linked business — but it removes a line of analysis, and $9,982M of DTC goodwill was reassigned to the Mortgage segment, requiring a quantitative impairment test immediately before and after (no impairment found; no headroom disclosed).

The rate back-up got worse, not better (2026 headwind). The 30-year fixed reached 6.8%, the highest in more than a year, and management stated plainly that “the expected housing recovery in 2026 has not materialized.” Q3 adjusted revenue is guided to $2.5–2.7B, down ~5.8% sequentially at the midpoint. This is the dominant near-term headwind and it is entirely exogenous.

Balance-sheet refinancing at a higher cost (June–July 2026). $1.5B of senior notes issued 16 June — $900M at 6.125% due 2031 and $600M at 6.500% due 2034, upsized from $1.2B and more than seven times oversubscribed — used to redeem the $1,150M of 2.875% notes due October 2026 and repay other debt. The 5.25% 2028s ($62M) were redeemed after quarter-end. A new $2.5B unsecured revolver (JPMorgan agent, maturing July 2029) replaced the April-2025 facility with no prepayment penalty. Net effect: the 2026 maturity wall is gone, liquidity is $11.2B, net corporate leverage is 0.9x — at the cost of a weighted-average coupon rising from 5.03% to 5.43%. The seven-times-oversubscribed reception is itself a credit datapoint.

UWM’s impairment (August 2026) — the largest competitive change since the Cooper close. Covered in Industry Dynamics above. A $451.9M quarterly loss from an over-hedged position surviving the collapsed Two Harbors deal, a suspended dividend, a $2.05B Oaktree/Ishbia recapitalization, a securities investigation, and a 34.78% single-session decline. Rocket out-originated UWM $49.1B to $39.7B. It also cost Rocket a 4.6% sympathy decline on 6 August.

Litigation. A $28M litigation accrual was recorded in Q2 “related to a specific legal matter” and is added back in adjusted net income. Total ASC 450 legal reserves rose from $74M at YE25 to $103M at 30 June 2026. The 10-Q’s standard language states no matter is currently expected to be material. Modest but rising.

Sell-side, post-print. RBC raised its target to $16 from $15 (12 Aug, Hold); Wells Fargo cut to $15 from $17 (7 Aug); Stephens cut to $20 from $22.50 (7 Aug); Benchmark initiated at Buy with $21. Average ~$17.78. Two cuts and one raise immediately after a profitability beat — consensus credits the cost execution and marks down the revenue trajectory.

GSE privatization remains a live, unresolved overhang. Nothing in the quarter advanced it.

Verdict: the changes CONTINUE to strengthen the franchise, and the shareholder thesis is now less muddy than it was. The prior memo concluded “a better company and a more complicated investment.” Two of the three complications it identified — negative tangible book and a near-richest-ever multiple — were artifacts. The third, dilution, remains. What is new is that the operating evidence has begun to arrive on schedule while the principal competitor removed itself from the field. The company is better than it was in June and the market’s read of it has, if anything, become less accurate.

9. Risk Analysis (Risk Matrix)

Reading the matrix. The risk profile improved in kind, not merely in degree. One risk — negative tangible equity — was removed because it was never real, and financing risk was downgraded on hard evidence (leverage, liquidity, maturity profile, oversubscription). What replaced them are risks of a healthier sort: that the revenue line keeps falling, that purchased share evaporates, and that recapture converts on mix but not on volume. The concentration in a single exogenous variable — the rate path — is unchanged and, if anything, more precisely measured now: a −1.90 loading to the InterestRate factor is the largest coefficient in every nested model. UWM’s blow-up is the cautionary note the sector needed: the tail risk in this asset class is a hedging accident on the servicing book, not a credit event on the loans.

10. Valuation Discussion (Embedded Expectations)

No price target; no recommendation. This section frames what the market is underwriting.

The capital structure, rebuilt from primary sources. The 10-Q cover page reports 1,906,422,676 Class A + 925,219,745 Class L = 2,831,642,421 shares at 31 July 2026, and the EPS note confirms both classes “have equal rights to participate in earnings and dividends on a per share basis” — so both count.

Retire the own-history percentile screen entirely for this name. The AZI valuation index reads P/E 115.5 (70.9th percentile), P/B 1.696 (10.2nd percentile), P/S 3.522 (97.3rd percentile) and composite 40.2 (59.5th). Read naively, the P/S leg says Rocket has never been more expensive. It is an artifact, and the mechanism is specific: until 30 June 2025 Rocket was an Up-C. Only ~171M Class A shares existed against ~2,000M of total economic interests — the Q2-2025 EPS reconciliation shows 171,438,105 weighted-average Class A plus 1,828,562,126 pro forma Class D. A per-share history computed on Class-A-only share counts against whole-company revenue produced artificially high sales-per-share and therefore artificially low P/S, which is exactly why the prior memo observed a historical P/S range of “~0.14–0.51.” Today’s 3.52x is measured across all 2,831.6M shares. These are not the same series and cannot be ranked against one another. The identical break corrupts the P/B history, and the P/E leg is separately meaningless because GAAP earnings are MSR-mark static. All three legs are broken; discard the screen rather than caveat it.

The corrected cross-period multiples. The prior memo, having correctly identified the share-base break, then reached the same “expensive” conclusion via EV/sales — but computed it by dividing a post-Cooper enterprise value into a revenue base that still contained pre-Cooper standalone quarters. On like-for-like bases:

~4.2x EV/adjusted revenue is full but not extreme — meaningfully different from the “6–8x, near richest ever” that anchored the prior memo’s conclusion. Note the honest caveat on the run-rate: Q3 is guided down ~5.8%, so annualizing H1 flatters the base. Using Q2 actual plus the Q3 guide midpoint gives ~$10.7B, and ~4.4x. Either way the multiple is in the low-to-mid 4s, not 6–8x.

Sum-of-the-parts, revised. The servicing asset is worth approximately its carried $18,905M — and that is now an observed value rather than an estimated range, because the Q2 sale of $52,926M of UPB cleared at 1.502% against a 1.558% carrying rate. Against a $39,445M market capitalization the servicing asset covers roughly 48% of the equity. The remaining ~$20.5B capitalizes the origination franchise (~$196B of annualized closed volume at a 2.48% blended margin), Redfin, Rocket Money and Rocket Loans, the ~$500M synergy programme, and the recapture option on the $320B of >6% note-rate loans. The prior memo’s “roughly half for a defensible annuity, half for a cyclical option” structure survives; only the upper bound on the annuity comes down.

Scenario analysis on normalized adjusted EPS. The prior memo’s scenario table has been overtaken on one axis: its bear case assumed ~$0.20 of EPS in a high-rate world, and the actual high-rate world produced ~$0.61. Recalibrated:

What must be true at $13.93. The current price embeds something close to the bear-to-base path — roughly 20–35x on plausible normalized earnings, 3.6x tangible book, 15.7x EV/EBITDA. That is a materially different statement from the prior memo’s conclusion that the price embedded “the base-to-bull path.” The reason is not that the stock fell (it is down only 3.4%); it is that the earnings base moved up — Q2 delivered ~$0.61 annualized where the prior framework expected ~$0.20–0.45 — and the balance-sheet discount that was being applied for negative tangible equity was applied to a fact that was not true. Where the market may be offsides: it is pricing the revenue decline (correctly) and the rate environment (correctly) but appears to give little weight to (a) the demonstrated liquidity and fair marking of the MSR book, (b) the removal of the sector’s second-largest competitor as an effective price-setter, and © an origination machine that now sources a majority of its refinance volume internally and holds $320B of in-the-money-on-a-cut inventory. Where the bear is right: 3.6x tangible book and 23x run-rate earnings is not a distressed price, the top line is still falling, and ~$170M a year of the share gain is rented.

Consensus. Hold, with an average target of ~$17.78 — RBC $16 (raised, 12 Aug), Wells Fargo $15 (cut, 7 Aug), Stephens $20 (cut, 7 Aug), Benchmark $21 (initiated Buy). Two cuts and one raise following a profitability beat.

Verdict (embedded expectations): fairly-to-fully priced on today’s earnings, with the rate option approximately free and the balance-sheet discount removed. The prior memo’s conclusion — “a fairly-priced option on falling rates with an annuity floor” — was directionally right but rested on two wrong numbers and one wrong multiple. Corrected, the same sentence is more favourable to the equity: the floor is harder than described (tangible equity is positive and the MSR proved saleable at book), the multiple is lower than described, and the earnings base is roughly 35% higher than the base case that supported it.

11. Variant Perception

Consensus. Hold, ~$17.78 average target. The sell-side view is that Rocket is a strategically improved, well-executing #1 platform whose earnings are hostage to a mortgage market that keeps getting worse — hence two target cuts and one raise on a quarter that beat on profit and missed on trajectory. The crowd is watching the revenue line.

Strongest bull case. Rocket executed a counter-cyclical acquisition of the two missing pieces of an integrated real-estate flywheel at a rate-cycle trough, and two quarters in, the pieces are demonstrably connected: 57% of refinance volume now comes from its own servicing book (up from 54%), over half of Rocket Loans volume goes to servicing clients, Redfin doubled mortgage leads, and it is now #1 in home equity. It did this while taking record purchase (6.2%) and refinance (14.3%) share in “one of the toughest spring housing markets in years.” Cost synergies were raised to $500M and are half-banked, producing 150bp of sequential margin expansion on declining revenue. It rotated its servicing book intelligently — selling dead low-coupon duration at full value for $795M of cash while retaining recapture, concentrating the book in $320B of >6% note-rate loans that are the actual refi option. Its balance sheet is stronger than described anywhere: +$10.9B of tangible equity, 0.9x net corporate leverage, $11.2B of liquidity, a seven-times-oversubscribed bond deal. And its principal competitor just lost $452M on a hedging accident, suspended its dividend, took rescue capital and drew a securities investigation. You are buying the last healthy giant in a broken industry at 3.6x tangible book and ~20x a plausible normalized number, with a free option on the rate cut that eventually comes.

Strongest bear case. Origination is a commodity with no barriers, and none of the above changes that. Revenue has fallen sequentially twice and is guided to fall a third time; share gains are being taken in a market shrinking faster than the share is growing. A meaningful part of that share is rented — Rocket Pro’s 0.69% margin represents ~$170M a year of purchased volume that may reverse the moment the Compass incentive normalizes. The servicing book is shrinking: customers voluntarily paid off $48.6B in Q2 alone, more than the $43.3B of MSR added by origination — so on the portfolio metric, recapture is not even holding serve. The company sold MSRs partly to raise cash. Goodwill of $10.6B sits on the balance sheet against a $39.4B market cap with no disclosed headroom, and the segment restructuring forced two quantitative impairment tests in a single quarter. Adjusted earnings strip $211M of Q2 items including $90M of genuinely cash-costly share-based compensation. At 3.61x tangible book and 23x run-rate adjusted EPS the stock is not cheap on any measure, there is no dividend or buyback, insiders bought nothing, and the dominant factor loading (InterestRate −1.90) says the whole position is a leveraged bet on a rate cut that has been “coming” for three years.

The 3–5 assumptions that matter most:

1. The rate path. Still dominant, still exogenous — and now precisely measured at a −1.90 factor loading. Everything else is second-order.
2. Does the $320B of >6% note-rate inventory recapture at scale when rates fall? The mix evidence (57%) says the routing works; the payoff evidence ($48.6B/quarter) says the volume is not yet there. This is the single unresolved question.
3. Does Rocket Pro share survive the withdrawal of the Compass incentive? Rented share versus franchise share, worth ~$170M/yr.
4. Does UWM’s impairment translate into durable share, or does Oaktree’s $2.05B simply restore the status quo? A recapitalized UWM in 2027 is a different competitor from a wounded one in 2026.
5. Is the $10.6B of goodwill supportable? Untested publicly; the trigger path is a sustained de-rating plus a lowered forecast.

Falsification. The bull is falsified if adjusted revenue continues declining ~5%+ per quarter through 2027 while the synergy benefit exhausts, holding normalized adjusted EPS near ~$0.40 — at which point 35x is an untenable multiple for a shrinking commodity lender — or if Rocket Pro volumes fall back toward pre-Compass levels once pricing normalizes. The bear is falsified if a cutting cycle arrives and the >6% cohort demonstrably refinances internally at the already-demonstrated 57%+ mix while portfolio UPB stabilizes and grows, taking adjusted EPS toward ~$1.20.

The factor-positioning read. The empirical tape has changed character since the prior report and the change is informative. RKT’s InterestRate loading is −1.896 in the Base+Sector+Industry model (R²=0.517) — the largest coefficient in all four nested models, and, since every factor is vol-scaled to 10%, the rigorous statement that this stock is a levered long-duration instrument. The InterestRate factor rose +7.01% over 126 days (z = +1.32), which on the model’s own arithmetic explains essentially all of the −24.5% six-month relative decline; positive loadings to Market (+1.18) and SmallSize (+1.19) partially offset. But two things shifted. First, the factor-neighbour set: UWMC is now the single closest neighbour at 0.915 similarity, ahead of the preferred and high-yield ETFs (PFF, PFFD, PGX, HYG) that topped the prior memo’s list, and homebuilders (LGIH, KBH, MTH, CCS, ITB, XHB) are prominent alongside a +0.900 Home Construction industry loading. The market has moved from trading Rocket as pure duration carry toward trading it as a housing-and-mortgage operating company with heavy rate duration — a modest upgrade in classification. Second, the decline has stopped: the three-month return is +4.1% annualized with a −19.4% max drawdown, versus −33.6% annualized over six months. Five-year annualized return is −2.74% with a −62.8% max drawdown and a −0.087 Sharpe; idiosyncratic volatility is 40.0%.

Where consensus is most likely offsides: the market is correctly pricing a deteriorating revenue line and a punishing rate environment, and it correctly identified this as a rate instrument. What it appears to under-weight is that the relative competitive position improved discontinuously in August — RKT −3.4% against UWMC −36.9% over the same window is not noise — and that the balance sheet is materially stronger than the widely-circulated characterization. Three top-tier long-onlys filing 13Gs within nine days suggests at least some of the market has noticed. This is no longer a falling knife; it is a full-priced quality operator holding a free option in a wrecked industry, and the risk is boredom and continued revenue attrition rather than impairment.

12. Fact vs. Interpretation Table

13. Open Questions

1. Portfolio retention versus mix. The 57% refi-mix statistic and the $48.6B of quarterly voluntary payoffs point in opposite directions. What is the actual retention rate on the combined book — the fraction of departing customers recaptured — and is it rising? Management cites records on the Cooper portfolio but does not publish a combined figure.
2. Compass durability. What happens to Rocket Pro volume when the up-to-100bp pricing credit normalizes? Is the Jupiter/Navigate AI adoption (5x growth for adopters) sufficient to retain the brokers acquired with price?
3. Goodwill headroom. $10.6B carried, two quantitative tests in Q2 on segment reassignment, no headroom percentage disclosed. How much cushion exists, and at what forecast does it close?
4. Gross MSR hedge exposure. The “net of hedges” presentation obscures the gross position. Given that UWM lost $451.9M on precisely this exposure, what is Rocket’s gross notional and its sensitivity to a rapid rate move in either direction?
5. Normalized earnings power. Is the ~$0.61 annualized H1 run-rate sustainable given the Q3 guide-down, or does it require the unrealized half of the $500M synergy programme to offset continued revenue attrition?
6. The lost segment disclosure. With DTC and Partner Network folded into one Mortgage segment, how do we independently monitor whether the Rocket Pro margin sacrifice is destroying segment profitability?
7. TRA residual. The Tax Receivable Agreement remains in place post-amendment. What is the remaining liability and its effect on free cash to common?
8. UWM’s trajectory. Does the $2.05B Oaktree/Ishbia recapitalization restore UWM as an aggressive price-setter in 2027, and on what timeline?
9. GSE reform. Unchanged and unresolved: what does privatization and higher g-fees do to Rocket’s volumes specifically?

14. What Must Be True

Bull case — what must be true:

• Adjusted revenue stabilizes within two to three quarters rather than continuing to decline ~5% sequentially, with the unrealized half of the $500M synergy programme bridging the gap.
• The $320B of >6% note-rate loans recaptures internally at or near the already-demonstrated 57% mix when a cutting cycle arrives, and portfolio UPB stabilizes rather than shrinking.
• The Rocket Pro share survives normalization of the Compass pricing incentive, converting rented volume into franchise volume.
• UWM does not return as an effective wholesale price-setter quickly enough to compete away the share Rocket is buying now.
• Normalized adjusted EPS marches toward ~$1.20, re-rating the stock from ~12x on that number.

Falsification test (bull): If, over the next four quarters, adjusted revenue continues to decline ~5% or more sequentially while the synergy benefit exhausts — holding normalized adjusted EPS near ~$0.40 — or if Rocket Pro closed volume falls back below ~$8B/quarter once the Compass incentive normalizes, the bull thesis is broken. At ~$0.40 of normalized EPS, 35x is untenable for a shrinking commodity lender no matter how good the operator.

Bear case — what must be true:

• Rates stay at or above ~6.75% through 2027, keeping origination volumes grinding lower and the refi option out of the money.
• The recapture flywheel converts on mix but not on volume — the 57% statistic keeps improving while the book keeps shrinking ~5% a year, proving the mechanism routes business it was going to get anyway.
• The Compass share reverses, revealing ~$170M/yr of purchased volume rather than a durable channel gain.
• The multiple de-rates from 3.6x tangible book and 23x toward mortgage-peer norms as the market concludes the 2025 dilution was never validated.

Falsification test (bear): If a genuine cutting cycle arrives and Rocket’s owned MSR UPB stops shrinking and grows for two consecutive quarters while the refi-from-servicing-clients mix holds at 57%+ — demonstrating that the $320B cohort recaptures at scale — and adjusted EPS heads toward ~$1.20, the bear thesis is broken and the franchise re-rates as a genuine flywheel rather than a well-marketed roll-up. Note the specific, checkable metric: owned MSR UPB, from the 10-Q rollforward. That is the number this thesis turns on, and it is published quarterly.

15. Source Appendix

See Appendix B — Source Appendix below for the full list of primary and secondary sources with URLs and access dates. Primary sources for this update include: Rocket Companies’ Form 10-Q for the quarter ended 30 June 2026 (filed 2026-08-07); the Q2-2026 earnings release (8-K of 2026-08-06, Exhibit 99.1); 8-K filings of 2026-06-09, 2026-06-10, 2026-06-16 (senior notes), 2026-07-16 (2026 Credit Agreement) and 2026-08-17 (board appointment); Forms 3/4 mirrored as raw XML; Schedule 13G filings by Capital World Investors, T. Rowe Price Associates and FMR LLC; SEC XBRL company-concept data; the Q2-2026 earnings call transcript; UWM Holdings’ Q2-2026 release and related trade-press coverage; AZI price and valuation-index data; and FactorsToday factor, leaderboard and regime data. Management-stated operating metrics (recapture, synergy targets, net corporate leverage) are labeled as such and treated as hypotheses pending validation. All fundamentals in this article were sourced from SEC filings and XBRL data directly.

Sections 1–15 of this article carry no investment recommendation and no price target. The only position expressed anywhere in this piece is the clearly-labeled “Claude’s Take” block at the top, which is the author’s own subjective view and is general information, not investment advice. The author holds no position in any security mentioned.

APPENDIX A — Standard Diligence Questionnaire

Rocket Companies, Inc. (NYSE: RKT) — as of 2026-08-22

A standing due-diligence checklist applied to the company. Fact / Interpretation / Assumption labels applied where material. This is an update to the version published 2026-06-20; answers that materially changed are marked [CHANGED].

General

What thoughtful questions have other investors asked about this company? The institutional debate has shifted meaningfully in nine weeks. It previously clustered on balance-sheet fear; it now clusters on trajectory. The live questions are: (1) Is the recapture flywheel real? — with a new, checkable answer: 57% of refinance closed volume now comes from existing servicing clients, up from 54% in Q1 (Fact, mgmt-stated), against a portfolio that is nonetheless shrinking (Fact). (2) Why is revenue falling while share is rising? — the market Rocket is taking share in is contracting faster than the share is growing (Interpretation). (3) How much of the Rocket Pro share is rented? — the 0.69% gain-on-sale margin implies roughly $170M/yr of purchased volume. (4) Did the 2025 all-stock deals create per-share value? given ~19x dilution — four of five sub-tests now favourable, the per-share test unresolved. (5) What is the gross MSR hedge exposure? — newly urgent after UWM lost $451.9M on exactly that risk. (6) Is the $10.6B of goodwill supportable? — no headroom disclosed. [CHANGED] The question “how do you value negative tangible book?” is retired: the premise was false. Tangible equity is +$10.9B, roughly $3.86/share.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Low, and demonstrably so. The 30-year fixed rate reached 6.8% in Q2-2026, the highest in more than a year, and management stated that “the expected housing recovery in 2026 has not materialized.” Adjusted revenue has declined sequentially twice and is guided to decline a third time (Q3: $2.5–2.7B versus $2,761M actual). [CHANGED] The important nuance is that Rocket earned ~$0.61 of annualized adjusted EPS in this trough — roughly triple what the prior report’s bear scenario assumed a high-rate environment would permit. (Fact + Interpretation.)

Driven by external environment or internal actions? Overwhelmingly external for the level; increasingly internal for the margin. The rate path drives volume, gain-on-sale margin, MSR marks and refi optionality simultaneously — quantified by a FactorsToday InterestRate factor loading of −1.90, the largest coefficient in every nested model. But the 150bp of sequential adjusted-EBITDA-margin expansion on a 2.2% revenue decline is internal: cost synergies, not cycle. (Fact/Interpretation.)

How stable are revenues? Historically very unstable (~$15.5B in 2020 → ~$3.6B in 2023, −76%). The 2025 acquisitions added a recurring servicing annuity — servicing fee income of $1,066M in Q2 alone versus $401M pre-deal — and H1-2026 gave the first real proof it dampens cyclicality: the MSR valuation-model input change was +$349M in H1-2026 (versus −$286M in H1-2025), explicitly “driven by an increase in interest rates.” Rates rose, origination suffered, the servicing asset appreciated. That is the hedge working. (Fact.)

Outlook for products/services; how big is the market? US origination is a multi-trillion-dollar but deeply cyclical annual market currently in its third famine year; servicing is a steady multi-trillion-dollar fee pool consolidating toward non-banks as banks retreat — and, [CHANGED], consolidating further within non-banks after UWM’s impairment. Domestic (US plus small Canada). (Fact/Interpretation.)

Business Quality & Competitive Moat

Is the industry getting more or less competitive? [CHANGED] — materially less, at least temporarily. Origination remains structurally commodity, but the wholesale channel’s dominant price-setter is impaired: UWM reported a $451.9M Q2 loss from an over-hedged position left after its collapsed Two Harbors acquisition, suspended its dividend, took a $2.05B Oaktree/Ishbia recapitalization, and drew a securities investigation. Rocket out-originated it $49.1B to $39.7B. Servicing continues to consolidate. Whether this is durable depends on how quickly the recapitalization restores UWM. (Fact + Interpretation.)

How profitable is the business (ROIC, ROE)? GAAP ROE annualizes to ~4.5% on H1-2026; return on tangible equity to ~15.8% (annualized adjusted net income $1,726M against tangible equity $10,940M) — though that is flattered by an adjusted numerator over a tangible denominator and a favourable MSR mark. Through-cycle returns remain boom/bust and fail Greenwald’s sustained-15–25%-ROIC franchise test. (Fact/Interpretation.)

How profitable is the industry; barriers to entry? Origination: low industry profitability, ~zero entry barriers, fragmented — and currently loss-making for at least one major participant. Servicing: better — capital, scale and operational barriers, recurring fees, and Rocket has just demonstrated that MSR scale is genuinely liquid by selling $53B of UPB at carrying value. (Fact/Interpretation.)

Can the business be easily understood? [CHANGED] — harder than before. The strategy is clear, but the accounting traps are demonstrably dangerous: the prior report misclassified the MSR asset as an intangible and mis-derived the MSR capitalization rate by dividing into subserviced balances. Beginning Q2-2026 the company also collapsed from two reportable segments to one (Mortgage plus All Other), removing channel-level profitability disclosure. Adjusted metrics strip $211M of Q2 items. Approach the financial statements with care. (Interpretation.)

Undermined by foreign low-cost labor? No — US-regulated mortgage origination and servicing is domestic and compliance-bound. The relevant labor substitution is AI, and Rocket is the one doing it: AI Voice handled >1 million inbound servicing calls within three months of launch, >50% of which would otherwise have required a person. (Fact.)

Do brands matter? Yes, more than in most commodity products, because a mortgage is an infrequent, high-anxiety, trust-laden purchase. Rocket has the strongest US brand awareness. But brand has still not conferred pricing power: Rocket Pro’s margin fell to 0.69% because Rocket chose to buy share on price. (Interpretation.)

Nature of competition / switching costs? Competition is on rate, speed and certainty (hence Rocket’s 12-business-day clear-to-close commitment and same-business-day conditional approvals). Borrower switching costs at origination remain ~zero. The only genuine stickiness is the servicing relationship plus the recapture funnel — which is the entire moat thesis and now has its first hard measurement at 57% of refi volume. (Fact/Interpretation.)

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? The brand, the customer graph and the recapture data asset are not capitalized. Conversely $12.6B of goodwill and intangibles are capitalized from the 2025 deals. (Interpretation.)

Off-balance-sheet liabilities? Servicing advance obligations (advance receivables of $1,542M are on-balance-sheet and fell from $2,040M — a working-capital release), representation-and-warranty buyback exposure on sold loans, the residual Tax Receivable Agreement liability, and $5,768M of loans subject to repurchase right from Ginnie Mae (grossed up on both sides of the balance sheet). Legal reserves under ASC 450 rose from $74M at YE25 to $103M at 30 June 2026, including a $28M Q2 accrual on “a specific legal matter.” (Fact/Open Question.)

How conservative is the accounting? Mixed, with one important positive. MSRs are carried at fair value and the Q2 bulk sale at 1.502% of UPB against a 1.558% carrying rate is a market-clearing validation that the mark is honest — that is meaningful evidence of conservatism in the largest single asset. Against that: heavy reliance on adjusted metrics that strip $90M of quarterly share-based compensation (a real cost), $99M of acquisition expenses and $112M of intangible amortization; and $10.6B of goodwill with no disclosed headroom despite two quantitative impairment tests in Q2. (Interpretation.)

How CapEx-hungry is the business? Not physically capital-hungry — property and equipment is $277M against $60,977M of total assets. It is financially capital-hungry: MSRs must be financed, servicing advances funded, and originations warehoused ($16,639M of secured financing against $15,548M of loans held for sale). The Q2 MSR sale is best read as a deliberate reduction in that financial capital intensity. (Fact/Interpretation.)

Capital Allocation & Management

How much FCF does the business generate, and how is it used? Rocket’s cash generation is best proxied by adjusted EBITDA of $766M in Q2 ($1,504M in H1) plus episodic MSR monetization ($795M in Q2). Uses in 2026: retiring the 2.875% 2026 notes, funding the Compass pricing incentive, reducing net corporate leverage to 0.9x (−20% since YE25), and building liquidity to $11.2B. No dividend and no buyback, and neither was discussed on the Q2 call. (Fact.)

Significant acquisitions recently? None new. Redfin (July 2025) and Mr. Cooper (October 2025) are being integrated ahead of plan — full servicing migration onto a single platform completed in Q2, cost synergies raised from $400M to $500M and half-realized. The notable transaction this quarter was a disposal: $53B of low-coupon MSR UPB sold for $795M with subservicing and recapture retained on ~80%. (Fact.)

Buying back shares? No. The ~$1B authorization (~$312M used) remains dormant. With the stock below its IPO price, at 3.6x tangible book, and net leverage at 0.9x, the absence of repurchase is a legitimate criticism. (Fact/Interpretation.)

Issuing large amounts of new shares to insiders? No new equity issuance of consequence — shares outstanding are 2,831,642,421 (1,906,422,676 Class A plus 925,219,745 Class L) at 31 July 2026, essentially unchanged since the deals closed. Share-based compensation runs ~$90M/quarter. Director grants in the window totalled ~74,757 RSUs. (Fact.)

Compensation policy of directors/management? CEO incentive metrics remain Adjusted Revenue and Adjusted EBITDA (50/50 RSU/PSU) with no ROIC or relative-TSR gate — management is paid on scale and adjusted profit, the metrics the dilutive deals inflate, without a per-share hurdle. Unchanged and still a legitimate governance criticism. New director compensation is standard ($75,000 cash retainer plus a $215,000 initial RSU grant). (Fact/Interpretation.)

Motivations of management? Founder Dan Gilbert / Rock Holdings retain effective control. Related-party flows are unchanged in character and immaterial in scale: $17M of Q2 lease expense to Bedrock affiliates, $13M of related-party G&A, $4M of marketing, $6M of notes receivable. [CHANGED] The board expanded from nine to ten on 17 August 2026 with the appointment of Sarah Watterson — Goldman Sachs, then Managing Director at Fortress Investment Group covering financial services including mortgage origination and servicing — the first director with direct MSR capital-markets underwriting experience. Read as competence-matching to a balance sheet now carrying an $18.9B MSR book. (Fact/Interpretation.)

Valuation & Market Data

Is the stock an ADR, MLP, or K-1 issuer? No. Rocket Companies, Inc. is a Delaware C-corporation filing standard 10-K/10-Q, issuing Form 1099 not K-1. The historical Up-C structure was collapsed on 30 June 2025 (Class D shares surrendered and retired); two participating classes remain, Class A (NYSE: RKT) and Class L, with equal rights to earnings and dividends. (Fact.)

Dividend policy? No regular dividend. One special dividend of $1.01/share was paid in February 2022. No dividend is currently contemplated. Trailing dividend yield 0%. (Fact.)

How profitable is the business? Q2-2026: 10.1% GAAP pretax margin ($281M on $2,784M); 27.7% adjusted EBITDA margin; Mortgage segment contribution margin of 52.2% ($1,174M on $2,251M of adjusted revenue). (Fact.)

Is net income diverging from cash from operations? Yes, structurally and unavoidably — this is a mortgage originator whose operating cash flow swings violently with loans-held-for-sale balances ($15,548M) and whose GAAP earnings are dominated by non-cash MSR fair-value marks (−$616M in Q2). GAAP net income of $229M versus adjusted net income of $441M is the more informative divergence, bridged by $90M of share-based compensation, $99M of acquisition expenses, $112M of intangible amortization and a $28M litigation accrual. Neither GAAP net income nor reported operating cash flow is a useful proxy for economic earnings here; use adjusted EBITDA and adjusted net income, and cross-check against the MSR rollforward. (Fact/Interpretation.)

Valuation snapshot at $13.93 (2026-08-21). Market capitalization $39,445M; corporate net debt $7,702M; EV excluding non-recourse funding $47,147M. EV/adjusted revenue 4.22x on run-rate (4.87x TTM); EV/adjusted EBITDA 15.7x; price/adjusted EPS 23.0x; price/book 1.68x; price/tangible book 3.61x. [CHANGED] The AZI own-history percentile screen (P/S at the 97.3rd percentile) should be discarded in full for this name: prior to 30 June 2025 only ~171M Class A shares existed against ~2,000M total economic interests, so the historical per-share series is not comparable to today’s 2,831.6M-share basis. All three legs are broken. (Fact/Interpretation.)

Risks & Downside

What factors would cause the stock to decline? In order: (1) mortgage rates staying at or above ~6.75% through 2027, keeping the refi option out of the money — the dominant factor, quantified at a −1.90 InterestRate loading; (2) continued ~5% sequential adjusted-revenue declines once the synergy benefit is exhausted; (3) Rocket Pro volume reversing when the Compass pricing incentive normalizes, revealing ~$170M/yr of rented share; (4) a goodwill impairment against the $10.6B carrying value; (5) an MSR hedging accident of the kind that cost UWM $451.9M; (6) GSE privatization raising guarantee fees and depressing volumes. (Interpretation.)

Risk of a catastrophic loss? Moderate in the sense of drawdown, low in the sense of impairment. The five-year maximum drawdown is −62.8% with 54.4% annualized volatility, beta 1.26 and 40.0% idiosyncratic volatility — a 40–50% drawdown on a rate or sector shock is entirely plausible and has happened repeatedly. [CHANGED] But the balance-sheet path to catastrophe described in the prior report does not exist: tangible equity is +$10.9B, net corporate leverage is 0.9x, liquidity is $11.2B, the 2026 maturity wall has been retired, and the June notes issue was seven times oversubscribed. (Fact/Interpretation.)

Chance of a total loss? Low, and better supported than previously stated. Behind the equity sit $10,940M of tangible equity and an $18,905M MSR asset whose liquidity was demonstrated this quarter by a $53B-UPB sale clearing at ~100% of carrying value. The new $2.5B revolver carries a minimum-tangible-net-worth covenant that a JPMorgan-led syndicate was willing to underwrite. (Fact/Interpretation.)

Recent News & Events

Has the business environment changed recently? Yes, in two opposite directions. Adversely: the 30-year fixed rate reached 6.8%, the highest in more than a year; the anticipated 2026 housing recovery did not arrive; Q3 adjusted revenue is guided down ~5.8% sequentially. Favourably: UWM Holdings — the largest wholesale originator — reported a $451.9M Q2 loss caused by an over-hedged position surviving its collapsed Two Harbors acquisition, suspended its dividend, announced a $2.05B equity investment by Oaktree Capital Management and SFS Group Capital (an Ishbia-family vehicle), fell 34.78% in a single session, and became the subject of a securities-law investigation. Rocket out-originated it $49.1B to $39.7B. (Fact.)

Significant acquisitions? None. The material transaction was a disposal — $53B of low-coupon MSR UPB sold for $795M of cash, subservicing and recapture retained on ~80%, concentrating the remaining book in the $320B of loans with note rates above 6%. (Fact.)

Change in accounting policies? [CHANGED] Yes, two. (1) Beginning Q2-2026 the company reports one segment, Mortgage (origination, servicing, title, closing, appraisal), with real estate and personal finance in All Other; prior periods were recast, and $9,982M of Direct-to-Consumer goodwill was reassigned to Mortgage, requiring quantitative impairment tests immediately before and after (no impairment found). (2) Beginning Q1-2026 certain interest-related activity was reclassified — “Interest income, net” retitled “Interest income,” funding-facility interest expense moved out of contra-revenue, and custodial deposit income moved from Other income to Interest income. No effect on net income, financial position or cash flows, but prior-period comparatives were restated. Both changes reduce comparability with pre-2026 analysis. (Fact.)

Recent changes — new markets, facilities, management? New $2.5B unsecured revolving credit facility (16 July 2026, JPMorgan agent, maturing 2029), replacing the April-2025 facility. $1.5B of senior notes issued 16 June ($900M at 6.125% due 2031, $600M at 6.500% due 2034), upsized from $1.2B and more than seven times oversubscribed, retiring the $1,150M of 2.875% notes due October 2026. Sarah Watterson appointed to the board 17 August 2026 (nine to ten directors). Completion of “one of the largest servicing migrations in industry history,” unifying all servicing clients on a single platform. Rocket Mortgage became the #1 US home-equity lender. The Compass “Power Play” broker initiative expanded, with same-business-day conditional approvals and a 12-business-day clear-to-close commitment. Three top-tier long-only managers — Capital World Investors, T. Rowe Price and FMR — filed Schedule 13Gs within nine days in August, together holding ~9.3% of the Class A shares. (Fact.)

APPENDIX B — Source Appendix

Rocket Companies, Inc. (NYSE: RKT) — research dated 2026-08-22

All URLs accessed 2026-08-22 unless otherwise noted. This is an update to coverage dated 2026-06-20; sources below are those relied on for the update. Primary sources (SEC filings and XBRL) are authoritative; third-party aggregated data is labeled as such and was reconciled to filings wherever it drove a conclusion.

1. Primary — SEC filings

Insider filings (Forms 3/4), 2026-06-01 to 2026-08-22 — read as raw XML from EDGAR (transaction codes and the aff10b5One flag parsed from the source XML rather than EDGAR’s rendered HTML):

Net: zero open-market purchases (code P) and zero open-market sales (code S) by any officer or director in the window.

Schedule 13G institutional ownership filings (August 2026), SEC EDGAR CIK 0001805284:

SEC XBRL company-concept data (https://data.sec.gov/api/xbrl/companyconcept/CIK0001805284/...), used to independently confirm the tangible-book correction: us-gaap:Goodwill, us-gaap:IntangibleAssetsNetExcludingGoodwill, us-gaap:FiniteLivedIntangibleAssetsNet, us-gaap:StockholdersEquityIncludingPortionAttributableToNoncontrollingInterest.

2. Primary — company earnings releases and call transcripts

3. Primary — competitor filings and releases

4. Trade press and financial media (secondary — used for event confirmation, not for figures that drive conclusions)

5. Quantitative data services (third-party aggregated — reconciled to filings)

AZI (azitrading.com)

• Daily price history CSV for RKT, and for UWMC / PFSI / ZG used in the peer-cohort control test: https://azitrading.com/controls/download-data.php?t={TICKER}. Supports: close of $13.93 (2026-08-21); 52-week range $12.35 (2026-06-08) – $23.44 (2026-01-15); adjusted all-time high $34.95 and unadjusted $41.60 (2021-03-02); −60.1% from peak; −29.9% year-to-date; 21/50/200-day EMAs of $14.11 / $14.15 / $15.31; the 2026-08-06 decline of −4.6% and the 2026-08-07 recovery of +3.8%; and the cross-sectional peer returns.
• Valuation index (own-history percentile ranks, 2026-08-21): P/E 115.50 (70.9th percentile), P/B 1.696 (10.2nd percentile), P/S 3.522 (97.3rd percentile), composite 40.24 (59.5th percentile), n_components 3. Reported in the memo as an artifact and explicitly discarded — the pre-2025-06-30 Up-C share base makes the own-history series non-comparable.

FactorsToday (factorstoday.com/api) — third-party statistical estimates; loadings and realized returns are reportable facts, forward statements are interpretation.

• /stock-loadings/RKT (2026-08-21): Base+Sector+Industry model, R² 0.517 — InterestRate −1.896, SmallSize +1.193, Market +1.179, Industry: Home Construction +0.900, OilPrice −0.379, USDollar −0.376, Industry: Banks +0.337, Sector: Real Estate +0.310. Base model — InterestRate −1.885, SmallSize +1.298, Market +0.935, USDollar −0.718, Value +0.516.
• /leaderboard/RKT (all figures annualized): y5 return −2.74%, volatility 54.4%, max drawdown −62.8%, Sharpe −0.087; y3 return +11.19%, Sharpe +0.162; y1 return −20.76%, Sharpe −0.387; m6 return −33.56%, Sharpe −0.576; m3 return +4.12%, Sharpe +0.034.
• /stock-info/RKT: beta 1.2554, alpha −0.14687, rs_ytd −28.05%, rs_6m −24.50%, rs_12m −21.57%, rs_peak −60.15%, market cap $39.83B, 90-day average volume 26.97M shares. Note: the descriptive text in this endpoint is stale — it still describes two segments, Direct to Consumer and Partner Network, superseded by the Q2-2026 single-segment structure. No quantitative field is affected.
• /stock-specific-vol/RKT: idiosyncratic volatility 40.0% annualized, model R² 0.508.
• /related-stocks/RKT: UWMC 0.915, UWMC-WS 0.911, PFF 0.853, PFFD 0.845, PGX 0.839, FND 0.835, LGIH 0.828, LEGH 0.827, PSK 0.827, KBH 0.824, HYG 0.817, JPM-PK 0.814, MTH 0.813, CCS 0.812.
• /factor-returns/historic: InterestRate +7.01% over 126 days (z = +1.32), +1.56% over 63d, +0.80% over 21d; Market +11.69% over 126d; SmallSize +5.04%; Value +0.31%.

Data note. All fundamentals in this article were taken directly from SEC filings and XBRL, which are the authoritative sources; no conclusion here depends on third-party aggregated fundamental data.

6. Analytical frameworks

• Bruce Greenwald & Judd Kahn, Competition Demystified — barriers-to-entry taxonomy (supply/cost, demand/captivity, economies of scale plus captivity), market-share-stability and sustained-ROIC franchise tests, applied in the Competitive Position section.
• Edward Chancellor (ed.), Capital Returns (Marathon Asset Management) — supply-side capital-cycle analysis and the asset-growth anomaly, applied in the Industry Dynamics, Growth and Capital Allocation sections.