Robinhood (HOOD) — The Tax Rate Arrived. The Price Still Assumes a Decade.

Robinhood's earnings per share fell 16% in a quarter when revenue rose 12%, and four-fifths of the fall is the tax rate reaching a normal 23.8%, as our July note said it would. Before investment marks and at a constant tax rate, earnings rose about 8%. The quarter barely moves our July method: $81.77 against $84.53. Valued on the returns its tangible equity earns, as we value brokers, a share is worth $35.91. The $94.56 price needs net income to compound at about 20% a year from the bull case's 2028 through 2036.

TON618 Capital Research · As of 27 October 2026 · Download PDF · Download model

In July we called Robinhood's price a growth commitment. Revenue had kept rising through a halving of crypto revenue, and we argued that the reported multiple flattered the company because its tax rate was far below normal. The third quarter tested both claims at once. Crypto revenue fell 55% from a year earlier and total revenue still rose 12%, the third quarter in a row of that pattern. The tax rate rose to 23.8% from 12.3%, and diluted earnings per share fell to $0.51 from $0.61. Event contracts, the line on contested legal ground, reached $173 million, 12% of revenue. In the same weeks, Robinhood lost its own Ninth Circuit appeal and filed a conditional petition asking the Supreme Court to hear it. This note asks what a share is worth at $94.56 after the print, and what the price needs to be true.

Terms used throughout, defined once. Normalized earnings: pre-tax income less other income (investment marks), taxed at 24%, less the non-controlling interest where per-share figures are given. Event contracts: exchange-listed yes/no contracts on outcomes (sports, elections, economics) that pay $1 if the outcome occurs; Robinhood offers them as a futures commission merchant. Take rate: annualized revenue divided by average Total Platform Assets. Tangible common equity (TCE): equity attributable to Robinhood less goodwill and intangible assets, $9.26 billion at 30 September. ROTCE: net income to common over average TCE. Excess-return model: equity value = TCE today plus the present value of each future year's earnings above the cost of equity on the TCE at the start of that year. Cost of equity (CoE): the return shareholders require, here argued from the capital asset pricing model (CAPM). SBC: stock-based compensation. Net deposits: customer money moved in less money moved out.

Revenue rose 12% to $1,424 million. Transaction revenue rose 19% to $867 million: options $399 million (+31%), event contracts $173 million (from $27 million), equities $121 million (+41%), crypto $120 million (−55%). Net interest fell 9% to $415 million and other revenue rose 61% to $142 million on Trump Accounts service fees and Gold. Operating expenses rose 16% to $739 million, led by credit-card provisions ($71 million, from $26 million, as the card book grew) and operations ($79 million, from $33 million). Operating income still rose 8%, to $685 million from $635 million.

The fall in earnings per share happens below the operating line. Exhibit 1 walks net income from the third quarter of 2025 to the third quarter of 2026.

Three items explain the drop. The tax rate rose to 23.8% from 12.3%, which on this year's pre-tax income costs $72 million. The 10-Q gives the reason: last year's rate was low "primarily due to excess tax benefits from SBC"; this quarter's is above the federal rate "primarily driven by state income taxes." For nine months the rate is 20.1%, held down by a valuation-allowance release on California deferred tax assets earlier in the year. Other income swung to −$60 million from −$1 million, mostly a $67 million mark-to-market loss on Robinhood's retained stake in Robinhood Ventures Fund I (RVI), which it deconsolidated in June and now carries at fair value ($361 million at 30 September). A new non-controlling interest took $8 million. The 10-Q does not name the entity. On our reading it is Rothera, the prediction-markets exchange Robinhood owns jointly with Susquehanna (SIG) and consolidates because it appoints the chief executive and a board majority. RVI and RVII were deconsolidated on 25 June and 14 August, which leaves Rothera as the only consolidated entity with outside owners the filing names at 30 September.

Set those aside and the business grew. Pre-tax income less other income, taxed at 24% in both years, was $521 million against $483 million: +8%. After the $8 million outside interest, net income to Robinhood on that basis rose 6%, and 7% per share on 911 million diluted shares against 918 million.

The July claim, re-tested. July said the reported price-to-earnings multiple understated the true one because the tax rate was abnormally low: 53 times reported, about 62 times at a 24% rate. At the $94.56 close the trailing multiples are 43.4 reported and 49.3 normalized, a gap of 5.9 turns against 8.6 in July and 6.8 at the September scorecard. The gap is closing the way the claim said it would: by reported earnings falling to the normalized level, not by normalized earnings rising to the reported one. The claim has played out.

The 8-K/A. On 28 October Robinhood refiled the release, "solely to revise the presentation of income tax effects of non-GAAP adjustments"; the revision "does not affect any GAAP financial results." Three rows differ: the tax effect of the non-GAAP adjustments, whose sign is reversed in every column, and the two rows computed from it. The filing calls this a revision of presentation; on the figures, the original carried the tax effect with the wrong sign. Adjusted Net Income moves from $550 million to $542 million for the quarter, $454 million to $500 million for the June quarter, and $565 million to $561 million a year earlier. Adjusted EPS for the quarter stays at $0.60; the June quarter's moves to $0.55 from $0.50 and last year's to $0.61 from $0.62. Nothing else changed but the headline sentence and two typographic labels.

Which earnings the "beat" measured. The Robinhood earnings feed shows $0.60 against a $0.56 estimate, without stating the estimate's basis. That $0.60 is Adjusted EPS, which the company introduced this quarter "to provide a view of profitability before certain non-operating items." It adds back investment marks ($67 million), CFO-transition stock compensation ($2 million) and amortization of acquired intangibles ($9 million), less their tax effect ($4 million): $468 million becomes $542 million. Diluted EPS under GAAP was $0.51. We use neither: the model uses normalized earnings.

Costs. The 2026 outlook for adjusted operating expense and stock compensation is now $2.675 to $2.75 billion, with the top end lowered by $25 million. It excludes credit-card provisions, the fastest-growing cost, and "costs from equity modifications of restructuring and executive awards in connection with our CFO transition." The release puts that transition at $22 million of stock compensation for nine months, $2 million of it in the third quarter; the Item 5.02 filings on it describe no award modification. A $23 million restructuring charge and its $7 million of stock compensation fell in the June quarter.

July named three engines that would decide the value: event contracts, net interest and crypto. Each moved this quarter, net interest and crypto in directions July did not expect.

Net interest fell 9%, mostly because of securities lending. Securities lending swung from +$87 million to −$19 million: a $106 million fall in a line that fell $41 million in total. Lower rates on cash took about $60 million more; balance growth more than made that up. Excluding securities lending, net interest rose 18%, to $434 million from $369 million. Margin interest rose to $244 million from $153 million on a record $22.1 billion margin book, and the credit card added $50 million (from $17 million). What fell with rates was income on cash and deposits, $102 million from $142 million as its yield fell to 2.09% from 3.96% (the net segregated-cash line fell to $55 million from $103 million), and the cash sweep, $45 million from $64 million. Average interest-earning assets grew to $73.3 billion from $60.8 billion.

The rate path also turned. July's base case assumed 75 basis points of cuts. In September the Fed raised its target by 25 basis points instead: effective fed funds stepped up on 17 September and is 3.88%, from 3.62% when the July note published. The September projections put the median at 4.1% at the end of 2026 and 2027 and 3.9% in 2028. The 10-Q puts the exposure at $366 million of pre-tax revenue a year per 100 basis points, from $294 million in July. The CFO gave the same order of size on the call: about $100 million a year for each 25 basis points, on about $40 billion of rate-sensitive assets.

Crypto: stabilizing, not yet a floor. We test the line against the Bitcoin price path rather than any forecast of it. The table shows crypto revenue for each quarter against the average Coinbase Bitcoin price that quarter.

Revenue rose 20% on the quarter while the average Bitcoin price barely moved. That is the first sequential rise after three falls. Two facts temper it. The 10-Q says the year-on-year fall was "partially offset by revenues earned from Robinhood Chain," the company's new blockchain, whose revenue the 10-Q does not quantify. And activity in the app is still shrinking: users trading crypto fell 37% from a year earlier and app notional volume fell 48% to $21 billion. The CFO attributed the lower take rate to the stablecoin product, Robinhood Earn, which the company monetizes less. One quarter, with a new revenue source inside it, is a stabilization, not a floor. The model assumes no Bitcoin return in any case.

Event contracts: volume times take rate, then the law. Robinhood traded 16.4 billion event contracts in the quarter, more than seven times a year earlier, and earned $173 million on them: about 1.05 cents a contract. October volumes started about 10% above the third quarter; we do not count that, because pricing changed when Rothera began operating. At the third quarter times four, the line is worth about $692 million a year, 12% of revenue. About 70% of activity is sports (the CFO said 30% of the quarter's activity was non-sports, up from 15% in the June quarter); we use that as the revenue share.

The legal record moved, mostly one way. On 28 August the Ninth Circuit ruled against Kalshi in Nevada (KalshiEX v. Assad, No. 25-7516) and, the same day, decided Robinhood's own Nevada appeal (No. 25-7831); the 10-Q says it "affirmed the district court's denial of RHD's motion." In September the Sixth Circuit joined, affirming the denial of Kalshi's injunction in Ohio and reversing its grant in Tennessee. Of the appellate rulings the 10-Q describes, only the Third Circuit's (New Jersey, Pennsylvania, Delaware) went for preemption. Two petitions now sit at the Supreme Court. New Jersey's against Kalshi (Flaherty v. KalshiEX, No. 26-299, docketed 8 September; response due 9 November after one extension) carries eleven amicus briefs, among them the National Football League and the American Gaming Association. Robinhood's is what the 10-Q calls a conditional petition, asking the Court to take its case if it grants New Jersey's or another (Robinhood Derivatives v. Dreitzer, No. 26-338, filed 10 September; response due 13 November). No call for the Solicitor General's views has issued.

Robinhood has already stepped back state by state. New sports contracts stopped in Nevada on 1 December 2025. In Michigan, Robinhood agreed to stop offering and to close sports contracts while appeals run. In Ohio, after a 2 October cease-and-desist letter, it agreed on 26 October to stop new sports contracts by 2 November. Washington's non-enforcement agreement lasts while Kalshi's petition for rehearing by the full Ninth Circuit is pending. Two further costs sit outside the courts. Robinhood has not collected or remitted the federal excise tax on wagers (Internal Revenue Code section 4401), because it does not believe the contracts are subject to it. And Illinois, North Carolina and Kentucky have enacted state taxes on prediction-market operators.

The company does not report event-contract revenue by state, so we carry the exposure as a grid, using population as the proxy for customers. Exhibit 2 shows what each outcome costs the run-rate.

The base case assumes the split stands through 2028: Ohio, Michigan, the eight remaining Ninth Circuit states and the two Sixth Circuit states are lost, 20% of event revenue, about 2.5% of total revenue. The bear case assumes the Supreme Court sides with the states in June 2027 and sports contracts end everywhere, 69% of event revenue. The bull case assumes preemption wins and Nevada returns.

A ruling binds all three scenarios, so we run it through all three, each keeping its own growth and the weights held at 25 / 50 / 25.

The loss is larger because the bull case carries the most event revenue and values each dollar of it most, and because the base case has 49 points further to lose against 21 to regain. A ruling would also remove the argument for equal tails in §3, but we move the weights only on the conditions under "What would change our view." July's sensitivity ($15 to $25) flexed the whole line between zero and $1.6 billion under multiples, so the two are not like for like.

Take rate and deposits. Annualized revenue was 1.49% of average platform assets in the quarter, against 1.67% a year earlier: an 11% fall in a year in which the newest line grew sixfold. Excluding crypto, the take rate rose, from 1.32% to 1.36%; the headline fall is the crypto line, not the rest of the business. Platform assets rose 19% to $395.9 billion, with $15.7 billion of net deposits and $11.5 billion of market gains in the quarter. October deposits topped $13 billion by the call, $6 billion of it the Dell family's gift to Trump Accounts. The remaining $7 billion, annualized, is about 21% of platform assets, in line with the trailing twelve months. We treat the gift as a one-off. The rest still includes ordinary Trump Account contributions and accounts from acquisitions, so it is not purely organic.

July, reproduced. The July engine, run unchanged, gives $86.94; the published figure was $87. That engine builds six revenue lines to 2028, costs them at 12% of transaction revenue plus fixed expense, taxes them at 24%, and capitalizes the result at 21, 28 or 34 times, plus net corporate cash. July carried $5,165 million of net cash and no funded debt. In fact Robinhood had closed $2.2 billion of 0% convertible notes due October 2029 on 25 June, eighteen days before the note published. Counting them as debt and changing nothing else gives $84.53. The error ran in the company's favor. Every comparison below starts from the corrected figure.

The quarter's data on July's method. Exhibit 3 re-runs July's engine with one input at a time moved to today's value. Shares and the 30 September balance sheet add a dollar. Options and net interest add +$5.03 and +$7.01, because volumes are at records and rates went up instead of down. Crypto takes away $10.27: its 2028 base falls to $600 million from $1,050 million, since July's recovery has not started and we assume no Bitcoin return. Event contracts take away $5.17 once the state grid is applied. The rest nets to about zero. On July's method the quarter moves fair value to $81.77, under three dollars below where it started.

The cost of equity is a judgment, argued here, with the reference reads beside it. For the explicit years we use 12.9%: the 10-year Treasury at 4.99% plus a beta of 1.75 times a 4.5% equity premium. Robinhood's own beta is high in every window we can measure: 2.17 on its last 60 monthly returns against the S&P 500 (October 2021 to September 2026, standard error 0.49) and 3.25 on the last 36, a window that starts after the post-listing fall. So the high reading is the business, not the listing: about three-fifths of the quarter's revenue was transactions, in options, equities, crypto and event contracts that rise and fall with retail traders' appetite for risk. We shrink the own estimate by its precision toward Interactive Brokers' beta (1.33), Robinhood's closest analog: its clients are active traders, and commissions were 36% of its net revenue in the quarter to June ($673 million of $1,896 million), with most of the rest interest on those clients' balances. We do not shrink toward the median of Schwab's and Interactive Brokers' betas, which includes Schwab (0.77), a broker built on advice and banking for long-term investors. The own estimate keeps 51% of the weight.

For the terminal value we use 10.5%: Interactive Brokers' beta adjusted toward 1.0 for maturity (1.22), reached in a straight line across each scenario's convergence stage. The mature business the model describes in 2036 is still a trading franchise: its terminal ROTCE is argued from asset-light trading lines, and the base and bull terminal returns (20% and 25%) are a trading platform's returns, which come with a trading platform's risk. The stage table's convergence row, below, cuts the other way: a take rate converging toward the large brokers' points to something nearer their risk, which is the case for the Schwab and Interactive Brokers median's 9.7%. That fits the bear case best; at 9.7% there alone, fair value is $36.14. One rate across the scenarios keeps them comparable, at a cost of $0.23.

The proponent's answer is that a business earning 23% on its tangible equity deserves a cost of equity nearer the brokers' than its post-listing beta implies. Our rates go part of the way: the explicit years are measured toward Interactive Brokers (12.9%, against its own 11.0%), and the terminal value goes further (10.5%); Exhibit 5 runs the rest.

What the change costs. The method takes fair value from $81.77 to $28.13 in three steps. The framework, with growth fading to the economy's rate: −$41.98, measured at an 11% cost of equity, at or above the brokers' own CAPM rates (8.5% for Schwab, 11.0% for Interactive Brokers), so that it is not charged with Robinhood's higher beta. The ROTCE fade to its terminal levels: −$1.78. The cost of equity from 11% to 12.9% in every year: −$9.88. The stable-growth rate for the mature business later adds +$7.86, in a row of its own. A dividend-discount model on the same scenarios' capital returns gives $35.96, a cross-check that the arithmetic is consistent.

How long the business grows faster than normal. Every stage length is argued from what drives it, not set by convention.

How growth is funded matters. Robinhood's trading lines are asset-light, but margin lending, the credit card and banking use balance sheet. So the model grows tangible equity with earnings during the growth stages (ROTCE held at its 2027 level) and lets ROTCE fade only in the convergence years, which charges new growth for the capital it uses. Equity also cannot shrink faster than a plausible payout: at most 60% of earnings plus the buyback pace of about $300 million a year. The 60% cap is our assumption; at 80%, fair value is $36.02. Without that floor, holding the bear case's ROTCE as its earnings fall would pay out $4.0 billion in 2027, 271% of that year's earnings. With it, bear ROTCE falls to 10.9% in 2028, the bear case earns less than its cost from 2028 to 2030, and ROTCE recovers to 13.0% by 2036. Part of that shortfall is retained capital the bear case's earnings no longer need, so its terminal prices capital that is not returned, not only a weaker franchise. The floor, a revision to the model's construction, is −$0.53; the base case's five-year asset-gathering stage, argued in the table above, is +$0.46.

The weights are 25 / 50 / 25, the same as July's, on their own argument: the base case runs the median of each dated observable (the circuit split stands, the Fed's median rate path), and each tail moves the legal outcome and the rate path in the same direction; the record gives no reason to rate one Supreme Court outcome likelier than the other while both petitions are undecided, so the tails carry equal weight. Each terminal ROTCE keeps a premium over the 10.5% stable-growth cost of equity, and each premium is argued. Base, 20%: an asset-light broker whose capital is mostly regulatory and cash, drifting down from 23% as banking and the card book add balance sheet. Bull, 25%: the platform stays asset-light while new lines attach to the same customers. Bear, 13%: the thinnest premium, 2.5 points over the stable rate. The take rate converges toward the large brokers', the card book adds credit capital, and with equity floored the bear case's ROTCE recovers only to about 13% by 2036; the terminal is set where it lands.

Return on tangible equity against its cost. Exhibit 4 sets twelve quarters of trailing ROTCE against the cost of equity. Robinhood has earned above 12.9% in every quarter since December 2024. The figures for December 2024 to September 2025 (19.8% to 29.3%) include a $358 million tax benefit from releasing the valuation allowance in the fourth quarter of 2024. Once it drops out, ROTCE runs 23.0%, 23.3%, 25.6% and 23.0%, on a tax rate that reached 23.8% in the third quarter (16.9% trailing). A mechanical reading says "fading," because of that base effect. The business reading is a steady reported return of about 23% (about 21% at a 24% tax rate) on tangible equity that grew 16% in twelve months ($8.0 billion to $9.3 billion), about 8 to 10 points above the explicit-year hurdle.

Relative value, as a cross-check. On normalized trailing earnings of $1.92 a share, Robinhood trades at 49.3 times. Schwab's 15.2 times would give $29.21, the median of five brokers (Schwab, Interactive Brokers, LPL, Virtu, eToro) is the same 15.2 times, and Interactive Brokers' 31.5 times, the richest in the group, would give $60.38. At this note's 10.5% stable-growth cost of equity, those two multiples imply perpetual growth of 4.2% and 7.0%; at each broker's own CAPM cost of equity (8.5% and 11.0%), 3.1% and 8.2%. They are reported, not weighted.

What the price requires. At $94.56, the market values Robinhood at 9.3 times its tangible equity. In the excess-return framework that is a promise of a perpetual ROTCE of about 65% at the 10.5% stable-growth cost of equity (about 87% at the explicit-year 12.9%). Exhibit 5 shows the paths. At our rates (12.9% falling to 10.5%), no length of the base case's 13% growth up to sixty years reaches the price; thirty years of it give $57.65. On the same rates, twenty percent growth from the base case's 2028 would have to last 17 years. On the proponent's own inputs, the bull engines followed by 25% a year for eight years (then a five-year fade to the economy's rate by 2041, with a terminal ROTCE of 30%), the value is $132.17 at our rates and $161.55 with 11% for the explicit years, both above the price. That path takes net income to $21.2 billion in 2036, about 12 times today's normalized level. The growth that matches the price at our rates is about 20% a year. Or the cost of equity can carry it: the base case reaches the price at a single cost of equity of 7.1% in every year (the weighted value at 7.3%), which is the 10-year Treasury plus about half of the 4.5% equity premium.

What else moves the number, and by how much. With 11% for the explicit years fair value is $40.60; with both rates a point lower, $43.95; a point higher, $30.08. Holding 12.9% in every year, with no stable-growth rate, gives $28.06. Terminal growth one point higher: $39.40. Two more years of asset gathering: $39.40. Terminal ROTCE five points higher in every case: $39.12. None of the operating sensitivities moves fair value by more than $3.49. The disagreement sits in two places. The stable-growth rate alone is worth +$7.86. The larger is the frame itself: fading growth to the economy's rate by 2036 is −$41.98, and the proponent's path in Exhibit 5 is where that is contested.

The July note left two triggers armed after the September scorecard. Both are scored here against the 10-Q and the dockets.

July's other tracked claims were scored in September. Two are updated by this quarter. The normalized-multiple claim has played out (§1). The decoupling claim held for a third quarter, with crypto down 55% and revenue up 12%. The convertible correction stands.

Eight tripwires follow, each with its effect on fair value where we can compute one; the reversal conditions come after them.

1. Supreme Court action on 26-299 and 26-338 (by 31 January 2027). A grant changes no number until the decision, because the base case already carries the split. A denial confirms the base case. A ruling for the states: −$3.29 a share across the three scenarios. A ruling for preemption: +$1.33. Silence means relisted or held: no change.
2. The full Ninth Circuit on the Assad case (by 31 March 2027). A denial ends Washington's non-enforcement, already inside the base case. A grant puts the Ninth Circuit states back in play.
3. Fourth-quarter event-contract revenue (print expected 9 February 2027, date not yet confirmed by the company). The base case carries about $170 million. At or above $195 million on volume raises the 2027–28 event line 10%; at or below $145 million cuts it 10%.
4. Take rate (by the second-quarter 2027 report). Two quarters at or above 1.45% with platform assets up at least 15% (not explained by the crypto line alone) raise base asset-gathering growth to 16%: +$1.61. Two at or below 1.30% (not explained by a market rally, or by the crypto line: the take rate excluding crypto at or above 1.36%) cut the stage to one year at 8%: −$2.40.
5. Net deposits in 2027, net of named gifts. At or above 18% of beginning platform assets, asset gathering lasts seven years: +$0.92. At or below 10%: −$2.40. Deposits at half the current pace remove the driver that sets the stage, so the weak branch cuts both its length and its rate.
6. Crypto against the Bitcoin path. Revenue per $1,000 of average Bitcoin at or above 1.60 in both the fourth quarter and the first quarter of 2027, with Robinhood Chain revenue shown or under a fifth of the line, lifts the bear crypto engine: +$0.11. At or below 1.30, no floor.
7. The rate path. Effective fed funds at or below 3.13%, or a 2027 median at or below 3.4% in the projections, moves base net interest to the bear path: −$3.25, before any lift to trading. Effective fed funds at or above 4.38%, or a 2027 median at or above 4.6%, moves it to the bull path: +$1.89.
8. Federal wagering excise. Guidance or a ruling that event contracts are wagers forces a re-estimate. It is not modeled, because no filing gives the notional amount the tax would apply to.

The quarter confirmed what July argued about earnings. The tax rate normalized and reported earnings fell to meet it, while the business beneath grew about 8%. It confirmed the decoupling a third time. It turned the rate risk from a headwind into a tailwind. It made the event-contracts line larger and its legal ground narrower. On July's own method, all of that nets to almost nothing: $81.77 against $84.53.

What changed is how we value it. A multiple ladder of 21 to 34 times assumes Robinhood grows 6.4% to 8.0% a year forever, even at the 10.5% stable-growth cost of equity. Measured instead as returns on its equity above its cost (12.9% for the explicit years, 10.5% once the business is mature), with growth fading to the economy's rate by 2036, Robinhood is worth $35.91 a share (bear $12.20, base $33.47, bull $64.51), against $94.56. The price is fair on the company's strongest terms: earnings compounding about 20% a year for eight years beyond the bull case's 2028, reaching about 9 times today's level. It is still the growth commitment July called it: 2.6 times fair value on this method, about 1.16 times on July's. Exhibit 5 shows what each reader's own hurdle rate implies. The Fund holds no position.