Revenue Held Flat While Market Volume Fell 10%, and the Hack-Day Sell-Off Hands Us the Entry: Upgrading to Outperform
Key Takeaways
- Revenue stopped falling, and the crypto market did not help. Total revenue was flat at $1,215.6M while total market crypto trading volume fell 10%. Prediction markets grew about 60% sequentially, past $200M of annualized revenue, so transaction revenue fell only 1.5%, to $590.2M, against the roughly $460M that July's run-rate implied.
- Costs beat guidance and the operating line turned positive. Adjusted expenses of $975.0M came in below the guide floor, Adjusted EBITDA rose 60% to $332.9M against roughly $285M consensus, and FY26 expense guidance was cut a second time, with the midpoint now below 2025's level. About $58M of the EBITDA gain is a swing in crypto held for operations, so the clean improvement is smaller but still real.
- The $2.51 of GAAP EPS is the crypto mark, not the business. A $715.0M gain on crypto held for investment produced net income of $718.7M. Adjusted net income was $77.7M, about $0.27 per share, and positive for the first time since Q4 2025.
- The reaction session sold the sector, not the print. COIN fell 10.8% to $143.29, after touching a new 52-week intraday low of $136.90, as a record $6.3B theft at Binance surfaced and the Nasdaq-100 fell about 3.5%. The shares now trade below the $150.73 at which we kept our Hold in August, on better fundamentals.
- Rating: Upgrading to Outperform from Hold. Two of our three August upgrade triggers cleared by wide margins, and the third (S&S at $580M) missed by $9.8M while S&S grew; at roughly 15–19x mid-cycle EBITDA, the stock prices a deeper trough than the business shows. Conviction is moderate because the Q4 guide steps expenses up 11%, monthly transacting users are still falling, and the crypto cycle has not turned.
COIN financial model
Income Statement · Dollars in millions, except per share
| Income Statement | ||||||
|---|---|---|---|---|---|---|
| Actual | Estimate | |||||
| 2023 | 2024 | 2025 | 2026E | 2027E | 2028E | |
| Consumer transaction revenue, net | $1,334.0 | $3,430.3 | $3,322.8 | $1,787.5 | $2,359.8 | $2,925.3 |
| Institutional transaction revenue, net | 90.2 | 345.6 | 479.7 | 422.9 | 545.4 | 700.3 |
| Other transaction revenue, net (2023+) | 95.5 | 210.2 | 252.9 | 186.1 | 211.6 | 245.5 |
| Total Transaction Revenue | $1,519.7 | $3,986.1 | $4,055.4 | $2,396.5 | $3,116.9 | $3,871.1 |
| Stablecoin revenue (2023+ presentation) | $694.2 | $910.5 | $1,348.8 | $1,193.6 | $1,296.9 | $1,441.8 |
| Blockchain rewards | 330.9 | 705.8 | 677.4 | 342.1 | 374.8 | 433.8 |
| Custodial fee revenue | 69.5 | 0.0 | 0.0 | 0.0 | ||
| Earn campaign revenue (2021 only; folded 2022) | 0.0 | 0.0 | 0.0 | |||
| Interest income | 0.0 | 0.0 | 0.0 | |||
| Interest and finance fee income (2023+) | 186.7 | 265.8 | 247.0 | 263.5 | 281.9 | 309.6 |
| Other subscription and services revenue | 125.6 | 425.1 | 554.8 | 437.1 | 501.0 | 579.9 |
| Total Subscription & Services Revenue | $1,406.9 | $2,307.1 | $2,828.0 | $2,236.3 | $2,454.7 | $2,765.1 |
| Net Revenue | $2,926.5 | $6,293.2 | $6,883.4 | $4,632.9 | $5,571.6 | $6,636.2 |
| Crypto asset sales revenue (2021-2022) | $0.0 | $0.0 | $0.0 | |||
| Corporate interest and other income | 181.8 | 270.8 | 297.9 | 267.0 | 248.2 | 247.6 |
| Total Other Revenue | $181.8 | $270.8 | $297.9 | $267.0 | $248.2 | $247.6 |
| Total Revenue | $3,108.4 | $6,564.0 | $7,181.3 | $4,899.9 | $5,819.8 | $6,883.8 |
| Less: Transaction expense | ($420.7) | ($897.7) | ($1,020.2) | ($710.1) | ($843.9) | ($998.2) |
| Less: Technology and development | (1,324.5) | (1,468.3) | (1,670.6) | (1,923.4) | (2,002.5) | (2,149.8) |
| Less: Sales and marketing | (332.3) | (654.4) | (1,058.6) | (1,071.1) | (1,283.8) | (1,424.4) |
| Less: General and administrative | (1,074.3) | (1,300.3) | (1,619.6) | (1,432.9) | (1,501.5) | (1,595.9) |
| Less: Losses (gains) on crypto held for operations (2024+) | 0.0 | 71.7 | (20.7) | (66.9) | 0.0 | 0.0 |
| Less: Crypto asset impairment, net (2023 presentation) | 34.7 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |
| Less: Restructuring | (142.6) | 0.0 | 0.0 | (52.4) | 0.0 | 0.0 |
| Less: Other operating expense, net | (10.3) | (7.9) | (356.1) | (24.9) | 0.0 | 0.0 |
| Total Operating Expenses | ($3,270.0) | ($4,256.9) | ($5,745.9) | ($5,281.7) | ($5,631.6) | ($6,168.2) |
| Operating Income (Loss) | ($161.7) | $2,307.2 | $1,435.4 | ($381.8) | $188.2 | $715.6 |
| EBITDA (OpInc + D&A) | ($22.0) | $2,434.7 | $1,623.9 | ($122.5) | $429.3 | $946.1 |
| Interest expense | ($82.8) | ($80.6) | ($85.4) | ($90.1) | ($90.1) | ($90.1) |
| Gains (losses) on crypto held for investment (2024+) | 0.0 | 687.1 | (528.9) | (691.9) | 0.0 | 0.0 |
| Other income (expense), net | 167.6 | 29.1 | 700.9 | 11.7 | 0.0 | 0.0 |
| Income (Loss) Before Income Taxes | ($76.8) | $2,942.6 | $1,522.1 | ($1,152.1) | $98.1 | $625.5 |
| Less: Provision for (Benefit from) Income Taxes | 171.7 | (363.6) | (261.7) | 164.9 | (19.6) | (125.1) |
| Net Income (Loss) | $94.9 | $2,579.1 | $1,260.3 | ($987.2) | $78.5 | $500.4 |
| NI Attributable to Common — Basic (numerator) | $94.8 | $2,577.8 | $1,260.3 | ($987.2) | $78.5 | $500.4 |
| NI Attributable to Common — Diluted (numerator) | 94.8 | 2,591.2 | 1,277.3 | (987.2) | 78.5 | 500.4 |
| Weighted Avg Shares — Basic (M) | 235.8 | 247.4 | 260.1 | 263.8 | 263.4 | 263.4 |
| Weighted Avg Shares — Diluted (M) | 254.4 | 273.4 | 287.2 | 263.8 | 263.4 | 263.4 |
| EPS — Basic | $0.40 | $10.42 | $4.85 | ($3.74) | $0.30 | $1.90 |
| EPS — Diluted | $0.37 | $9.48 | $4.45 | ($3.74) | $0.30 | $1.90 |
| Ratios & Assumptions | ||||||
| Operating Margin | (5.2%) | 35.1% | 20.0% | (7.8%) | 3.2% | 10.4% |
| Net Margin | 3.1% | 39.3% | 17.6% | (20.1%) | 1.3% | 7.3% |
| Effective Tax Rate [DRIVER] | 223.5% | 12.4% | 17.2% | 14.3% | 20.0% | 20.0% |
The full workbook adds 22 historical and 10 projected quarters, plus KPI Drivers · Balance Sheet · Cash Flow Statement — every subtotal a live formula, every projection traced to a driver.
Results vs. Consensus
This is the first quarter since Q3 2025 in which Coinbase beat on the lines that measure the operating business. Total revenue landed in line with to slightly below the most-cited Street figure, but the composition is the point: transaction revenue beat, subscription and services beat, Adjusted EBITDA beat by double digits. The shortfalls were corporate interest income, which fell because Coinbase used cash to retire a convertible note, and stablecoin revenue, which landed about 2% below the one published estimate for that line.
| Metric | Actual | Consensus | Beat/Miss | Magnitude |
|---|---|---|---|---|
| Total revenue | $1,215.6M | ~$1.23B (range $1.17–1.23B) | In line / slight miss | -1.2% to -1.4% vs. most-cited |
| Transaction revenue, net | $590.2M | ~$527M | Beat | +12.0% |
| Subscription & services | $570.2M | ~$562M | Beat | +1.4% |
| Adjusted EBITDA | $332.9M | ~$285M | Beat | +16.8% |
| EPS (GAAP, diluted) | $2.51 | $(0.28) to $(0.10) | Beat | Driven by the crypto mark |
| Adjusted EPS (derived) | ~$0.27 | $(0.06) to $0.06 | Beat | See basis note |
A note on the consensus basis. Published revenue consensus ran from $1.17B to $1.233B, and the two most widely cited figures sat at $1.23B, which is the comparison we use. Subscription and services consensus rests mainly on one provider, with a second figure close to it. Transaction-revenue consensus was published by a single provider, so we put more weight on the bar Coinbase set itself: roughly $130M of transaction revenue through July 26, which at that pace implied about $460M for the quarter. Coinbase does not publish adjusted EPS. Adjusted net income of $77.7M over 288.0M diluted weighted shares gives roughly $0.27, and the two services that published an adjusted figure disagreed on its sign, so we show the range.
The scorecard that matters more: management against its own guide
Last quarter this table had one unchecked line, and it was the one we cared about most. This quarter every line landed on plan or better, with transaction expense at the top of its band.
| Metric | Q3 outlook (given July 2026) | Q3 actual | Result |
|---|---|---|---|
| Subscription & services revenue | $500–580M | $570.2M | Upper half of range |
| Transaction expenses (% of net revenue) | Mid-teens | 16.8% | Top of the band |
| Adjusted expenses | $980–1,080M | $975.0M | Below the low end |
| Stock-based compensation | ~$245M | $233.0M | Below guide |
| Transaction revenue (run-rate disclosure) | ~$130M through July 26 (~$460M at that pace) | $590.2M | +28% vs. the July pace |
The transaction line is the one that changes the story. The July pace was about $5.0M per day. The full quarter averaged $6.4M per day, which means August and September ran at roughly $7.0M per day. July was the weakest stretch Coinbase has disclosed, and our August arithmetic, which annualized it, was too bearish by about $130M of transaction revenue.
Sequential and year-over-year
| $M | Q3'25 | Q4'25 | Q1'26 | Q2'26 | Q3'26 | Q/Q | Y/Y |
|---|---|---|---|---|---|---|---|
| Transaction revenue, net | 1,046.3 | 982.7 | 755.8 | 599.2 | 590.2 | -1.5% | -43.6% |
| Subscription & services | 716.6 | 694.5 | 583.5 | 555.1 | 570.2 | +2.7% | -20.4% |
| Corporate interest & other income | 105.8 | 104.0 | 73.6 | 65.8 | 55.1 | -16.2% | -47.9% |
| Total revenue | 1,868.7 | 1,781.1 | 1,413.0 | 1,220.1 | 1,215.6 | -0.4% | -34.9% |
| Net revenue | 1,762.9 | 1,677.1 | 1,339.3 | 1,154.3 | 1,160.5 | +0.5% | -34.2% |
| Adjusted expenses | 1,088.5 | 1,229.4 | 1,132.9 | 1,034.8 | 975.0 | -5.8% | -10.4% |
| Operating income (loss) | 480.5 | 273.8 | (21.4) | (113.5) | 61.7 | n/a | -87.2% |
| Adjusted EBITDA | 800.7 | 565.9 | 303.3 | 207.8 | 332.9 | +60.2% | -58.4% |
| Net income (loss) | 432.6 | (666.7) | (394.1) | (359.5) | 718.7 | n/a | +66.1% |
| Diluted EPS | 1.50 | (2.49) | (1.49) | (1.36) | 2.51 | n/a | +67.3% |
| Adjusted net income (loss) | 420.7 | 178.0 | (45.6) | (104.9) | 77.7 | n/a | -81.5% |
The year-over-year net income comparison flatters this quarter for the same reason the Q2 comparison punished it: both periods are dominated by marks on the crypto portfolio. The honest year-over-year read is revenue down 34.9%, Adjusted EBITDA down 58.4%, and adjusted expenses down 10.4%. The honest sequential read is better than anything since Q3 2025: revenue flat, costs down 5.8%, and EBITDA up 60%.
- Revenue: Flat on mix, not on market. Total revenue fell $4.5M sequentially: transaction revenue fell $8.9M, subscription and services rose $15.1M, and corporate interest and other income fell $10.6M (the components differ from the total by $0.1M on rounding). Inside transaction revenue, prediction markets grew about 60% and offset most of a 9% decline in consumer crypto spot volume. Inside subscription and services, stablecoin revenue rose $14.8M on a record platform balance and, in management's account, higher interest rates. Nothing here depends on a one-time item.
- Margins and costs: Mostly genuine, with one exception. Adjusted expenses fell $59.8M (5.8%): technology and development fell 5.4% on the full-quarter effect of the May reduction, and G&A fell 10.6% on lower professional services. Sales and marketing rose 1.2% on higher USDC rewards. The exception is that gains and losses on crypto held for operations sit inside Adjusted EBITDA, and they swung from a $31.7M loss in Q2 to a $26.1M gain in Q3. That $57.8M swing is 46% of the $125.1M EBITDA increase. Excluding it, Adjusted EBITDA rose 28%, from $239.5M to $306.8M, which is still a clean beat.
- EPS and below the line: Almost entirely non-operating. The $715.0M gain on crypto held for investment and $66.7M of other income drove pre-tax income to $823.9M. The 12.8% effective tax rate was flattered by a valuation-allowance release and a one-time favorable adjustment from the close of a federal tax audit. Adjusted net income of $77.7M is the operating number, and it is positive for the first time in three quarters.
Revenue
Total revenue of $1,215.6M was essentially unchanged from Q2's $1,220.1M. What makes it notable is the market it was earned in: Coinbase's own market data shows total market crypto spot volume down 6%, total market derivatives volume down 11%, and crypto volatility down 15% to multi-year lows. In Q1 and Q2, revenue fell 20.7% and 13.7% as those inputs weakened. This quarter it did not. The revenue base now has enough non-spot components (prediction markets, derivatives, and a subscription and services line at 49% of net revenue) that a soft spot tape no longer translates one-for-one into a revenue decline. That is the diversification thesis working in the quarter it was most needed.
Margins and operating leverage
Operating income of $61.7M is the first positive operating quarter since Q4 2025 and a $175.2M swing from Q2. It needs to be read carefully, because three items inside operating expenses helped: the $26.1M gain on crypto held for operations, $18.4M of net recoveries related to the 2025 data theft incident, and a $4.6M reversal of the restructuring accrual. Strip all three and operating income is roughly $13M, compared with roughly $(63)M on the same basis in Q2. The improvement is real and large. The cushion is thin. Adjusted expenses consumed 84.0% of net revenue, against 89.6% in Q2, and stock-based compensation of $233.0M fell to 70% of Adjusted EBITDA from 115% in Q2.
Below the line and the crypto mark
The crypto portfolio mark has now dominated GAAP results for five consecutive quarters, and this quarter it ran the other way: a $715.0M gain against losses of $209.5M in Q2, $482.4M in Q1 and $718.2M in Q4 2025. Bitcoin rose roughly 43% during the quarter, and Coinbase has been adding to its position. Crypto assets held for investment stood at $2.22B at quarter-end, and the CEO said on the call that the company had just bought its 18,000th bitcoin. We have treated this mark as a presentation problem rather than an economic one since Q3 2025, and we apply the same view in the flattering direction. The $2.51 of GAAP EPS should not be capitalized, just as the $(1.36) should not have been.
Segment Performance
Transaction revenue
| Line ($M) | Q3'25 | Q2'26 | Q3'26 | Q/Q | Y/Y |
|---|---|---|---|---|---|
| Consumer | 843.5 | 451.7 | 450.1 | -0.3% | -46.6% |
| Institutional | 135.0 | 100.1 | 91.7 | -8.4% | -32.1% |
| Other | 67.7 | 47.4 | 48.5 | +2.2% | -28.4% |
| Total transaction, net | 1,046.3 | 599.2 | 590.2 | -1.5% | -43.6% |
The sequential column is the one to read this quarter. Consumer revenue held flat on falling volume, institutional revenue fell with its market, and the total declined 1.5% after falling 20.7% in Q2. Management said prediction markets and derivatives together made up about 25% of transaction revenue in the quarter.
Consumer
Consumer transaction revenue of $450.1M was flat against Q2's $451.7M, while consumer crypto spot volume fell 9% to $23.4B and total crypto derivatives volume fell 6%. The gap is prediction markets: the product crossed $200M of annualized revenue in the quarter, which on our reading means at least $50M of quarterly revenue. Coinbase's growth disclosures (106% sequential growth in Q2, 82% quarterly compound growth from Q1 to Q3) put Q3 about 60% above Q2. Management again said the new products are incremental to spot rather than cannibalizing it.
"And what's great about that is this is a market where overall spot crypto trading was actually down. We saw overall crypto spot trading volume down 10% quarter-over-quarter. And so the growth in prediction markets offset the decline in overall spot trading. And we delivered flat revenue quarter-over-quarter."
— Alesia Haas, CFO
The deck's own figures put total market crypto spot volume down 6% and total market crypto trading volume, including derivatives, down 10%.
Assessment: The consumer line now has a second engine, and it carried the line through a quarter that would otherwise have produced another double-digit decline. What it did not do is bring users back. Monthly transacting users fell to 7.4M from 7.6M in Q2 and 9.3M a year ago, so the stabilization is revenue per user, not users. That is a better outcome than we modelled and a narrower one than the flat revenue line suggests.
Institutional
Institutional transaction revenue of $91.7M fell 8.4% sequentially, which management described as roughly in line with market declines. Crypto derivatives volume fell 6.2% to $963.0B against an 11% decline in the market, a fourth consecutive quarter of derivatives share gain, and trailing-twelve-month derivatives volume reached $4,341B. Year over year, institutional revenue fell 32.1%. That reverses the +63.9% year-over-year growth we highlighted in Q2, and the reason is the comparison base: Q2 2025 preceded the Deribit acquisition, while Q3 2025 included Deribit from mid-August.
Assessment: We overstated the Q2 institutional datapoint. Much of that year-over-year growth was acquisition, and against a Deribit-inclusive base the institutional line is tracking its market down while gaining share. The INTX/Deribit integration completed in October, and management explicitly warned of near-term volume volatility as participants adjust to the combined platform. Expect noise in Q4 before the unified liquidity shows up in revenue.
Other
Other transaction revenue of $48.5M rose 2.2%, the first sequential increase since Q3 2025, after Q2's decline on lower Base revenue.
Assessment: Too small to move the total. Base's value continues to be built in activity share, borrow and lend balances, and tokenization volume rather than in this line.
Subscription and services
| Line ($M) | Q3'25 | Q2'26 | Q3'26 | Q/Q | Y/Y |
|---|---|---|---|---|---|
| Stablecoin revenue | 324.5 | 292.1 | 306.9 | +5.1% | -5.4% |
| Blockchain rewards | 184.6 | 83.3 | 85.7 | +2.9% | -53.6% |
| Interest & finance fee income | 64.8 | 66.1 | 68.6 | +3.7% | +5.9% |
| Other subscription & services | 142.7 | 113.5 | 109.0 | -4.0% | -23.6% |
| Total subscription & services | 716.6 | 555.1 | 570.2 | +2.7% | -20.4% |
Q3'25 stablecoin revenue reflects Coinbase's reclassification of $30.1M to conform to the current presentation.
Stablecoin revenue
Stablecoin revenue of $306.9M rose 5.1% sequentially, the first sequential increase since Q4 2025, but remains 5.4% below the year-ago quarter. The drivers pulled in opposite directions. Average USDC held in Coinbase products reached a record $22B, reflecting the first full quarter of the Hyperliquid balance that arrived in June, while average USDC market capitalization fell to $73B from $77B. Coinbase's disclosed split shows the effect: revenue on platform balances (excluding corporate balances) rose to roughly $181M from $146M, while revenue on off-platform balances fell to $126M from $146M.
"So we capture about 50% of the overall economics. We hold about 30% of total USDC in circulation as assets on our platform."
— Alesia Haas, CFO
The per-dollar arithmetic explains why the mix shift matters. On the rounded balances Coinbase discloses, a platform-held USDC dollar earned roughly 3.6% annualized in Q3, against roughly 1.0% for an off-platform dollar. Measured against average USDC market capitalization, the blended yield rose to roughly 1.7% from roughly 1.5% in Q2. Management attributes part of the improvement to higher interest rates and lists higher average rates as a Q4 tailwind. That is the opposite of the rate path we assumed in August.
Assessment: Better than we feared and still not a growth line. The improvement came from two sources Coinbase only partly controls: pulling USDC onto its own platform, which it pays for through USDC rewards in sales and marketing, and a more favorable rate backdrop. The USDC float itself shrank 5% sequentially, and the 10-Q shows one counterparty, which we take to be Circle, at 27% of total revenue against 19% a year ago. The pillar stays at risk, but the direction of travel improved for the first time in three quarters.
Blockchain rewards
Blockchain rewards of $85.7M rose 2.9% sequentially on price effects and growth in native units staked, after falling 17.4% in Q2. The line is still down 53.6% year over year.
Assessment: Stabilized with token prices, as a price-levered line should. It will move with the market in both directions, and we continue to treat it as the highest-beta component of the supposedly defensive half of the business.
Interest and finance fee income
Interest and finance fee income of $68.6M rose 3.7% sequentially and 5.9% year over year, on a record average daily loan book that management said was up 74% year over year. Average borrow and lend balances in the app rose 25% sequentially to $1,866M from $1,491M, and loan receivables on the balance sheet reached $1.98B.
Assessment: The steadiest line in the P&L and the clearest evidence that customers are using bitcoin as collateral rather than selling it, a trend the CEO described directly on the call. Still small, at 12% of subscription and services.
Other subscription and services
Other subscription and services revenue of $109.0M fell 4.0%, as a record number of paid Coinbase One subscribers partly offset lower milestone-based partnership revenue. Management flagged this roll-off in Q2. Paid subscribers rose 24% year over year, and Coinbase disclosed that subscribers generate 3.5x the revenue per user and hold 4.2x the balances of non-subscribers, with trading retention more than 50% higher.
Assessment: The relative metrics are new and useful, and they partly answer the unit-economics question we raised in Q2. The absolute subscriber count is still missing for a second quarter, so the growth rate cannot be turned into a revenue contribution.
Key KPIs
| KPI | Q3'26 | Q2'26 | Q3'25 | Trend | Read |
|---|---|---|---|---|---|
| Coinbase crypto trading volume market share | 10.7% | 10.3% | 5.6% | All-time high | Fourth consecutive quarterly gain; spot gain driven by stablecoin demand |
| Assets on platform | $377B | $246B | $516B | +53% Q/Q | Price rebound plus ETF inflows |
| Assets on platform, % of crypto market cap | 12.5% | 11.2% | 12.5% | +130bps Q/Q | Back to peak share after Q2's 80bps decline |
| Average USDC held in Coinbase products | $22B | $20B | $15B | All-time high | About 30% of average USDC market cap |
| Average USDC market cap | $73B | $77B | $68B | -5% Q/Q | Off-platform balances fell to $51B from $57B |
| Crypto derivatives volume (TTM) | $4,341B | $4,221B | $3,241B | +2.8% Q/Q | Quarterly volume -6% vs. market -11% |
| Average borrow/lend balances | $1,866M | $1,491M | $610M | +25% Q/Q | All-time high |
| Prediction markets revenue | $200M+ annualized | $100M+ annualized | n/a | About +60% Q/Q | Daily traders +60% Q/Q |
| Monthly transacting users | 7.4M | 7.6M | 9.3M | -2.6% Q/Q | -20% Y/Y; disclosed in the 10-Q only |
| Consumer crypto spot volume | $23.4B | $25.8B | $57.4B | -9.3% Q/Q | Fell faster than the total market |
| Full-time employees | 4,204 | 4,321 | 4,795 | -2.7% Q/Q | Down 12% Y/Y |
| Ending shares outstanding | 265M | 264M | 269M | +1M Q/Q | Buyback effectively paused in Q3 |
Two KPIs pull against the rest of the table. The first is monthly transacting users. At 7.4M, MTUs are down 2.6% sequentially and 20% year over year, and the only place Coinbase reports them is the 10-Q. Consumer crypto spot volume fell 9.3%, faster than the 6% decline in Coinbase's measure of total market spot volume, and the deck credits the quarter's spot share gain to stablecoin demand rather than to crypto trading. The share statistics are real, but on the disclosed figures they do not show the retail crypto trader coming back.
The second is capital return. Coinbase repurchased 37,995 shares for $5.3M in Q3, all in August at an average of $139.91, against $121.3M in Q2 and $1,188.8M across the nine months (all on the equity-statement basis). The shares ended the quarter at 264.8M, up from 263.8M at the start, because issuance for equity awards exceeded repurchases. Roughly $1.96B of authorization remains. Cash and cash equivalents of $8.6B are down from $11.3B at year-end 2025, mostly because Coinbase repaid a $1.27B convertible note and bought back stock earlier in the year. A pause is defensible on liquidity grounds. It is still notable that the buyback went quiet with the shares near their 52-week low while the company kept adding bitcoin.
Key Topics & Management Commentary
Overall Management Tone: Management was confident and product-led, and for the first time in a year the numbers supported the tone. The live session again carried no financial question, with the results discussion moved to pre-recorded videos, so the posture on the call was strategic rather than accountable. Compared with Q2, management leaned harder on diversification as a proven fact rather than a promise, and was less defensive about the cycle.
1. The live call is now a product forum; the results moved to video
For the second consecutive quarter, the live session on X opened with no prepared remarks. This time the CFO pointed listeners to pre-recorded videos in which she and the CEO discussed the quarter, and the live hour went to four questioners: one institutional research analyst, a crypto data company, an asset manager's digital-assets researcher, and a podcast host.
"If you haven't already, please head over to our IR website to watch the videos we posted of Brian and I discussing our quarterly results. But in this session, right now, we are excited to directly connect with you, our customers, our community, our shareholders and we want to talk about our quarter and answer questions."
— Alesia Haas, CFO
No one asked about the Q4 expense step-up, the paused buyback, the decline in monthly transacting users, the drop in USDC market capitalization, or the stablecoin counterparty concentration. The Binance breach, which began the evening of the print, was not mentioned.
Assessment: An improvement on Q2, because the results were at least discussed by management in some form rather than skipped. The structural issue is unchanged: a format with no follow-up questions on the numbers lets management choose which numbers get discussed. In a strong quarter that costs little. We still count it against the name.
2. Diversification delivered the flat quarter
The CFO's central message was that years of investment in non-spot products produced a flat revenue quarter in a down market. The disclosures support it: prediction markets and derivatives at about 25% of transaction revenue, subscription and services at 49% of net revenue, and 88% of net revenue from sources other than bitcoin spot trading last quarter.
"Predictions plus derivatives, so those 2 combined were about 25% of our total transaction revenue. And if I look at then our total revenue, subscription services was roughly 49% of total net revenues."
— Alesia Haas, CFO
The CEO framed the gap between the operating facts and the headline numbers as not yet understood by the market.
"I don't think that this is widely understood yet. And the facts on the ground are actually looking a lot better than it's showing up maybe in the overall numbers, which are more flattish."
— Brian Armstrong, CEO
Assessment: In Q2 we wrote that diversification across products within one asset class is not diversification across cycles, because the replacement revenue is levered to the same volume, volatility, price and rate variables. This quarter partly answers that objection. Prediction markets grew through slower sports months and multi-year-low crypto volatility, which is the first evidence of a revenue line that is genuinely less correlated to the crypto tape. One quarter does not prove it; it does move the burden of proof.
3. Prediction markets passed the seasonal test
Our main caution on prediction markets in Q2 was the seasonal comparison: the NBA playoffs and the World Cup inflated Q2, so Q3 would be hard. Revenue grew about 60% anyway, to more than $200M annualized, with daily traders up more than 60% sequentially. The deck credits non-sports categories and recurring crypto contracts, such as crypto volatility contracts, for engagement between major events, and cites 82% quarterly compound revenue growth since the February launch.
"So prediction markets is one of our fastest-growing new revenue products. We just launched it in February this year. And during the third quarter, it crossed $200 million of annualized run rate revenue."
— Alesia Haas, CFO
The roadmap is event hubs for the midterm elections, integration with Coinbase One, a prosumer product, and, on economics, a move toward USDC-native, in-house clearing. Transaction expense rose to 16.8% of net revenue, up 32bps, because prediction-market growth carries exchange fees.
Assessment: The best operating line in the release for the second quarter running, and the seasonal objection is now answered. The open question is margin: revenue is disclosed, but the exchange fees that come with it are visible only as a rising transaction-expense ratio, and the plan to bring clearing in-house is an admission that the current economics are not yet the target economics.
4. The second expense cut came from overhead; the Q4 step-up is marketing
FY26 adjusted-expense guidance fell to $4.15–4.30B from $4.20–4.45B, a $100M cut at the midpoint for the second consecutive quarter. The new midpoint of $4,225M is 1.3% below 2025's actual $4,282M. With nine-month adjusted expenses at $3,142.7M, the range implies Q4 adjusted expenses of $1,007–1,157M, which matches the Q4 guide and steps up 11% at the midpoint from Q3's $975.0M. Coinbase attributes the Q4 increase to marketing, driven by USDC rewards and performance marketing, and to professional services.
Assessment: In August we set a downgrade trigger on a second in-year cut that reached the growth investment. The cut arrived, but it came from G&A professional services and the full-quarter run-rate of the May reduction, while sales and marketing rose and the Q4 guide adds more. That is the right shape: cut overhead, spend on growth marketing that management said in Q2 pays back inside a year. The trigger is not hit. The cost of the choice is a lower Q4 EBITDA, which we work through in the guidance section.
5. Circle: cooperation on USDC, competition on infrastructure
Asked where Coinbase's stablecoin efforts overlap with Circle's, the CFO gave the most concrete economics disclosure of the call: Coinbase captures about half of USDC's economics while holding about 30% of USDC in circulation on its platform. She was also direct that the two companies now compete on payment infrastructure.
"Where we compete is we're both building stablecoin payment infrastructure. And so we will compete in some areas but we are both incentivized to grow total USDC market cap and deeply embedded in the overall network of stablecoin usage around the world."
— Alesia Haas, CFO
Assessment: This partly answers the number we said was most missing in Q2, the economics Coinbase retains after sharing. "About 50%" is an average across platform and off-platform dollars, not a marginal figure, but combined with the disclosed revenue split it lets an outside reader rebuild the per-dollar yields. The concentration is the other side of it: one counterparty is now 27% of total revenue, and the relationship is openly competitive at the infrastructure layer.
6. Agentic finance: real growth, small base
The CEO described Coinbase for Agents, the connector that lets third-party AI agents trade and pay through a Coinbase account, as the area he now spends most of his time on. He gave one growth figure and was careful to scale it.
"I mean the agentic trading volume is growing something like 20% week-over-week right now, which is just shocking."
— Brian Armstrong, CEO
He added immediately that the volume is still small relative to the rest of Coinbase's trading, and described an endpoint through which agents send product feedback that engineers turn into production changes within a day.
Assessment: A 20% weekly growth rate is a statement about a tiny base, and management said so. As in Q2, we treat agentic finance as an option with no current cash flows. What is new is the CEO's personal focus on it, which raises the odds that it gets resourced through the cycle.
7. Deribit integration is done; volatility first, unified liquidity later
In Q2, management described U.S. perpetuals and unified Deribit liquidity as coming soon without a date. Both moved this quarter. The INTX/Deribit integration completed in October, Deribit is being made available to U.S. and non-U.S. customers, and Coinbase Pro returns in Q4 as a unified advanced platform covering spot, futures, perpetuals, options and equities.
"So we've got now this unified regulated global exchange that processes everything, perps, options, futures. We're making Deribit now available to both U.S. and ex U.S. So I think we're one of the only companies that can unify the global liquidity."
— Brian Armstrong, CEO
Assessment: A commitment from the Q2 call delivered on schedule, which counts in management's favor. The deck's own warning of near-term volume volatility should be taken literally: the integration lands in the same quarter as an expense step-up, so Q4 institutional revenue could be soft before the combined order book pays off.
8. Tokenized equities: sequencer fees now, liquidity later
Asked whether the Base sequencer fee is the main revenue source for tokenized stocks, the CFO said it is the most direct one today and that the real opportunity is building liquidity around the assets. Coinbase cited about $3B of tokenized equity trading volume for non-U.S. customers since launch, and the deck shows more than $1B of tokenized stock volume on Base.
"So yes, sequencer fees are the most direct monetization point for tokenized equities today. And our main opportunity then is the utility of the assets once they're tokenized."
— Alesia Haas, CFO
Assessment: An honest answer that confirms tokenization is a volume and positioning story for now. The CEO's point that Coinbase issues real securities with dividends and voting rights, rather than synthetic exposure, is a credible differentiator for a regulated custodian. It is not yet a revenue line worth modelling.
9. Security and custody, on the night of the largest exchange theft on record
A question on whether AI could break the elliptic-curve cryptography that secures crypto custody drew a careful technical answer: the CEO said he is not worried about AI on that front but treats quantum computing as a matter of when, not if, and confirmed that Coinbase has begun upgrading its custody to post-quantum standards.
"One is, we have begun the work to upgrade our custody to be post-quantum. So that's important."
— Brian Armstrong, CEO
The timing gave the exchange more weight than it would otherwise have had. The Binance breach, reported at about $6.3B, began shortly after the U.S. close on the day of the print. Binance disclosed it on Thursday and confirmed its size on Friday. Neither the call nor the materials referred to it, and it may not have been public when the call took place.
Assessment: Custody is the franchise. Coinbase stores more crypto than any other platform, custodies a large share of the U.S. spot ETF assets, and holds 80-plus licenses. A record theft at the largest offshore competitor is a near-term negative for crypto prices and sentiment and a medium-term argument for the regulated, audited venue. The 2022 FTX collapse is the obvious precedent, and there too the sector sell-off came first.
10. Balance sheet: buying bitcoin, not stock
Coinbase reported $8.6B of cash and cash equivalents and $11B of available resources, including $2.4B of crypto and marketable investments. During the nine months it repaid a $1.27B convertible note, bought $217.9M of crypto for its investment portfolio, and repurchased $1,188.8M of stock, almost all of it before Q3. The deck reports bitcoin units held up 92%.
"I think we just purchased our 18,000th bitcoin actually on Coinbase balance sheet."
— Brian Armstrong, CEO
Assessment: The balance sheet is not a concern, and the convertible repayment is a sensible use of cash. The capital-allocation signal is mixed. The company kept adding bitcoin while repurchasing $5.3M of its own stock in a quarter when the shares traded as low as $139. If management believes its own diversification message, its own equity was the cheaper asset.
11. Competition, and a CEO turning back to the product
Asked how Coinbase differentiates as more platforms pursue the same multi-asset model, the CEO pointed to crypto infrastructure (Base, USDC, the x402 protocol), to serving consumers, institutions and developers on one platform, and to the scale of the opportunity outside traditional finance. On a separate question, he said his time is shifting away from Washington and back toward customers and product.
"So I'll still be in D.C. periodically but my -- more and more of my focus now is shifting to really just talking to our customers, improving our products. That's going to be my focus much more in 2027."
— Brian Armstrong, CEO
Assessment: The differentiation answer is strongest on infrastructure and weakest on the retail app, which is where competitive pressure from multi-asset brokers shows up and where MTUs are falling. A founder-CEO redirecting attention to product as the regulatory agenda matures is a positive, and it is the right response to the consumer data in this release.
Guidance & Outlook
Coinbase continues to guide subscription and services and the expense lines, and to disclose transaction revenue quarter-to-date. In August that disclosure worked against the stock. This quarter it is modestly constructive.
| Metric | Q4'26 outlook | Q3'26 actual | Implied change |
|---|---|---|---|
| Transaction revenue | ~$165M QTD through October 25 | $590.2M | +2.9% at the QTD pace |
| Subscription & services revenue | $550–630M | $570.2M | +3.5% at midpoint |
| Transaction expenses | Mid-teens % of net revenue | 16.8% | Roughly flat |
| Adjusted expenses | $1,007–1,157M | $975.0M | +11.0% at midpoint |
| Stock-based compensation | ~$240M | $233.0M | +3.0% |
| Amortization of intangibles | $36–46M | $33.8M | Higher |
| FY2026 adjusted expenses | Low | High | Midpoint | Change |
|---|---|---|---|---|
| 2025 actual | n/a | n/a | $4,282M | n/a |
| Initial 2026 outlook | $4,250M | $4,600M | $4,425M | n/a |
| Outlook as of Q2 2026 | $4,200M | $4,450M | $4,325M | -$100M vs. initial |
| Current 2026 outlook | $4,150M | $4,300M | $4,225M | -$100M vs. Q2; -1.3% vs. 2025 |
Implied quarter-over-quarter ramp. Transaction revenue of about $165M over the first 25 days of the quarter is a pace of $6.6M per day, which across 92 days gives roughly $607M, up about 3% from Q3. Two cautions sit on top of the company's own warning against extrapolating. First, the October pace is above the Q3 average of $6.4M per day but below the roughly $7.0M per day that August and September ran at, so activity cooled into the new quarter. Second, bitcoin was about 8% below its September 30 level by the close of the reaction session, and the Deribit integration is expected to unsettle derivatives volume in the near term.
Building Q4 from the company's guidance and its October pace, on the same method that reproduces Q3's reported $332.9M exactly:
- Transaction revenue at the October pace: ~$607M. Subscription and services at the guide midpoint: $590M. Net revenue: ~$1,197M.
- Corporate interest and other income, extending the decline: ~$50M. Total revenue: ~$1,247M.
- Less transaction expense at 16.5% of net revenue (~$198M) and adjusted expenses at the $1,082M midpoint, then add back guided stock-based compensation of ~$240M and roughly $27M of depreciation outside intangible amortization. We assume no gain or loss on crypto held for operations.
- Result: Adjusted EBITDA of roughly $235M at the expense midpoint. At $1,030M of adjusted expenses (in Q3 Coinbase landed about 5% below its midpoint) it is roughly $287M. At the low end it is roughly $310M.
These are our estimates, not company figures. The point is that Q4 EBITDA probably declines from Q3, and the reason is marketing reinvestment and the absence of a crypto-ops gain rather than revenue deterioration. A 16th consecutive positive quarter is not in question on any reasonable path.
Street at: Post-print commentary suggests Street Q4 transaction-revenue estimates were set well above the October pace and are likely to come down, while the subscription and services guide midpoint sits above where pre-print expectations were. The two broadly offset at the net revenue line, but the mix shifts toward the steadier half of the business. The post-print price-target changes we have seen were mostly cuts, and the published range remains very wide, which is consistent with a Street that does not agree on what this franchise earns at mid-cycle.
Guidance style: Conservative on expenses, and reliably so: Q3 adjusted expenses landed below the floor of the range. On subscription and services, the guide recovered its credibility after Q2's miss below the floor, with Q3 landing in the upper half. We would expect Q4 adjusted expenses to land below the midpoint again, which is why our working Q4 EBITDA estimate sits between the midpoint and low-end cases.
Analyst Q&A Highlights
Four questioners took part in the live session, and each asked two or three questions. The exchanges below are the substantive ones. None addressed the financial results directly.
Whether the case for owning bitcoin has changed
The opening question asked whether bitcoin's role has evolved now that it can serve as collateral in DeFi markets, given that it still trades like a risk asset. Management's answer tied the collateral use case to its own lending growth and closed with a cautious read on the cycle.
Q: "Do you feel like the case for Bitcoin ownership broadly has evolved?"
— Andrew Jeffrey, William Blair
A: "So -- and then just broadly, I'd say like in this cycle that we're in, I mean, it's always difficult to call these things exactly but we are seeing volumes rebounding and more institutions coming and onboarding."
— Brian Armstrong, CEO
Assessment: The collateral point is supported by the numbers, with borrow and lend balances up 25% sequentially to a record. The cycle comment is hedged appropriately. "Volumes rebounding" fits the August and September transaction pace better than the October one.
How much of the consumer line is prediction markets
The only question that approached the income statement asked management to separate prediction markets from spot in consumer transaction revenue and to say how to think about it going forward. The answer gave the mix figures and the incrementality claim but no prediction-market revenue figure for the quarter.
Q: "Could you help us parse a little bit how much -- it certainly looks like prediction markets pushing consumer transaction revenue in the quarter. But can you help us sort of unpack that a little bit and tell us maybe how we can think about that going forward?"
— Andrew Jeffrey, William Blair
A: "And what we saw right now in this quarter, a lot of that was around predictions. But I also want to share with you, this isn't cannibalizing our spot trading. What we're seeing is, this is all incremental."
— Alesia Haas, CFO
Assessment: The incrementality claim is now repeated across two quarters, but the evidence offered is behavioral rather than numerical. Falling MTUs and a 9% decline in consumer spot volume are consistent with a smaller, more engaged user base spending more per head, which is a narrower version of "incremental" than the phrase suggests.
Where Coinbase and Circle cooperate and compete
A question framed the Circle relationship as "coopetition" and asked where the two companies' stablecoin efforts overlap. The answer produced the call's most useful number, the share of USDC economics Coinbase retains.
Q: "Can you talk about perhaps where your efforts overlap with Circle's and where your efforts are distinct?"
— Andrew Jeffrey, William Blair
A: "So we capture about 50% of the overall economics. We hold about 30% of total USDC in circulation as assets on our platform."
— Alesia Haas, CFO
Assessment: A direct answer to the question we flagged as unasked in Q2. Read with the disclosed revenue split, it shows why pulling USDC onto the platform is worth paying for through rewards. It also confirms that Coinbase's largest single revenue relationship is with a company it now competes with on infrastructure.
What could go right for agentic trading
A questioner from a crypto data company whose team had integrated Coinbase's agent tooling asked what the upside case looks like for a connector that spans crypto, equities and prediction markets. The CEO gave the call's only agentic growth figure, with an explicit scale caveat.
Q: "And so my question is, what could go right for Coinbase agents because it's the only MCP that includes x402 in prediction markets, equities and crypto. Like what could go right from here?"
— Jonathan Ma, Artemis
A: "Yes, there's still -- the numbers in the grand scheme of things are still relatively small compared to our other trading volume. But to me, it's a clear indication of where things are going."
— Brian Armstrong, CEO
Assessment: Management volunteered the caveat, which is the right posture. The exchange confirms agentic finance as the CEO's priority and leaves its revenue contribution at zero for modelling purposes.
Post-quantum readiness for custody
A question on whether AI-discovered algorithms could break the cryptography behind crypto custody drew a distinction between AI, which the CEO does not consider a threat to that cryptography, and quantum computing, which he does.
Q: "How does that change the security and custody setup? And are you accelerating the post-quantum road map?"
— Lorenzo Valente, ARK Invest
A: "One is, we have begun the work to upgrade our custody to be post-quantum. So that's important. Now, of course, we also need the blockchains themselves to be upgraded and to be quantum resistant."
— Brian Armstrong, CEO
Assessment: A substantive answer with a concrete action. The dependency it names matters: Coinbase can harden its own custody, but the assets it custodies depend on protocol upgrades it does not control. On the night of a record theft at a competitor, the custody franchise is the asset investors should value most.
How tokenized stocks make money
Asked whether Base sequencer fees are the main revenue source for tokenized equities and how fragmentation across issuers resolves, the CFO said sequencer fees are the most direct monetization point today and argued that liquidity, not issuance, decides the winners.
Q: "Is the Base sequencer fee the main revenue driver going forward? Or are you looking at other monetization plans?"
— Lorenzo Valente, ARK Invest
A: "Building depth of liquidity, building network effect, that is the hard part."
— Alesia Haas, CFO
Assessment: The liquidity argument is the same one management used for USDC and Base in Q2, and it has held up there. For tokenized equities it is a positioning claim with no revenue attached yet.
Deribit and the unified derivatives exchange
A question on how the Deribit business has grown and how plans changed with the U.S. regulatory shift drew confirmation that the integration is complete and that Coinbase Pro is returning.
Q: "Have you seen the growth of that product, the perpetual futures and the options? And how has your plans with that shifted with the recent shift in more of a positive regulatory environment in the U.S.?"
— Zion Thomas, Market Bubble
A: "We're -- we said publicly, we're going to actually relaunch our advanced trader platform under the Coinbase Pro brand. So Coinbase Pro is coming back."
— Brian Armstrong, CEO
Assessment: The growth half of the question was not answered with a number, though the deck supplies one: derivatives volume compounding 179% annually over two years. The roadmap commitments from Q2 have been met, and the measure for Q4 is whether the unified exchange holds share through the integration noise.
Whether Coinbase will market harder to retail
A host who described himself as a long-time customer asked whether Coinbase plans more consumer marketing, given the perception that it focuses more on Washington than on traders. The answer defended the regulatory work and signalled a shift in the CEO's attention.
Q: "But I'm curious if you guys have plans to do more marketing along that front to kind of target a different consumer or a different customer segment."
— Zion Thomas, Market Bubble
A: "Now this doesn't mean -- it's not a zero-sum game. Like we've got to -- we can go build products, we can go talk to our customers and help push the regulatory ball forward."
— Brian Armstrong, CEO
Assessment: The marketing question was answered indirectly by the Q4 guide, which steps up performance marketing, rather than on the call. The consumer data argues for exactly that spending, and the CEO's stated shift toward product in 2027 points the same way.
What They're NOT Saying
- Monthly transacting users. MTUs fell to 7.4M from 7.6M in Q2 and 9.3M a year ago. The figure appears only in the 10-Q, not in the deck or on the call, in the quarter management argued that the business is healthier than the headline numbers show.
- Why the buyback stopped. Repurchases fell to $5.3M in Q3 from $121.3M in Q2, with the shares near their 52-week low and $1.96B of authorization remaining. Management did not explain the pause, while the CEO highlighted continued bitcoin purchases.
- The payback on the Q4 marketing step-up. Adjusted expenses step up about $107M at the guide midpoint, led by USDC rewards and performance marketing. Management cited a one-year payback on growth marketing in Q2. It offered no updated figure for the spending it is now adding.
- Prediction-market economics. Revenue run-rate and trader growth are disclosed. The take rate, the exchange fees that pushed transaction expense up 32bps, and the margin are not.
- Why USDC market capitalization fell. Average USDC market cap fell 5% sequentially and off-platform balances fell to $51B from $57B, with off-platform stablecoin revenue down about 14%. The record platform balance got the attention. The shrinking float did not.
- An absolute Coinbase One subscriber count. For the second quarter, the subscriber base is described by growth rate (+24% year over year) and relative metrics, but not by size.
- The running count of products above the revenue threshold. Coinbase disclosed twelve products above $100M of annualized revenue at Q1. The deck now uses a $200M threshold for prediction markets, and neither count has been updated.
- The October slowdown. The quarter-to-date transaction pace is below the August and September pace. The deck says to be cautious about extrapolating, but offers no commentary on what cooled.
- The Binance breach. Neither the call nor the materials addressed it, which may simply reflect its timing. Investors will want management's view on whether a record theft at the largest offshore exchange shifts flows toward regulated venues, and how Coinbase's own exposure to the event, if any, is managed.
Market Reaction
- Pre-print setup: COIN closed at $160.59 on October 29 ahead of the after-market release. The shares entered the print down 29.0% year to date against the S&P 500 up 12.8%, down 53.9% over the trailing twelve months, and down 15.5% over the trailing thirty days. The 52-week closing range was $141.09 to $355.22, so the stock came in about 14% above the bottom of a very wide band.
- After-hours move: Shares rose about 1% in the first hour after the release, to roughly $162.60, then faded to a decline of about 2%, near $157, later in the evening.
- Reaction session (October 30): Opened at $148.23, a 7.7% gap down, traded a range of $136.90 to $148.50, and closed at $143.29, down 10.8% or $17.30. The intraday low undercut the prior 52-week intraday low of $139.11 from July 31. The close held above the $141.09 52-week closing low. Volume of 18.7M shares was 2.2x the 30-day average. The S&P 500 fell about 2%.
- Peers and crypto: Robinhood fell 7.2%, Strategy 7.8%, Circle 9.3% and Galaxy Digital 11.8%, and the Nasdaq-100 fell 3.5%. Bitcoin finished the session down about 1%, near $76,900, after touching a six-week low overnight.
The release itself barely moved the stock. After-hours trading treated it as roughly neutral, up about 1% and then down about 2%. The 7.7% opening gap formed overnight, as news spread that Binance had suffered a theft reported at $6.36B, which reports described as the largest on record, and which began shortly after the U.S. close. Binance said it would cover all losses and worked with stablecoin issuers to freeze roughly $3B of the stolen assets, but crypto equities fell together regardless.
Beta explains most of the size of the move. COIN's 10.8% decline was about three times the Nasdaq-100's 3.5% fall, which is consistent with the stock's beta of more than 3. The crypto-linked peer group fell between 7% and 12%, and COIN sits inside that range. On this evidence the market did not single out Coinbase's quarter. The gap between COIN at -10.8% and bitcoin at roughly -1% is the familiar pattern of crypto equities trading as levered sentiment instruments on days of sector stress.
What the market may be pricing beyond beta is narrower. Post-print commentary pointed to total revenue landing slightly below the most-cited consensus and to the decline in spot volume, and Q4 transaction-revenue estimates look set to come down. Against that, the shares recovered 4.7% from the $136.90 low to close at $143.29, near the session's volume-weighted average price. That shape looks more like a flush on sector news than a re-rating of the business, and we would use it as an entry point.
Street Perspective
Debate: Is the EBITDA rebound an operating turn or a mark?
Bull view: The constructive camp argues that Coinbase has proven it can grow EBITDA without a market recovery: revenue flat, adjusted expenses down 5.8%, EBITDA up 60%, and a second cut to the full-year cost guide. On this view the cost base is now structurally below 2025 while revenue is 35% below last year, so the operating leverage into any recovery is large and under-modelled.
Bear view: The skeptics point out that GAAP profit is almost entirely the crypto portfolio mark, that a $57.8M swing in crypto held for operations sits inside Adjusted EBITDA, and that the Q4 guide steps expenses back up by about 11%. Strip the one-offs and operating income was close to zero, and Q4 EBITDA is likely to fall.
Our take: Both are describing the same quarter accurately, and the bull has the better of the argument on direction. Even excluding the crypto-ops swing, EBITDA rose 28% on flat revenue, which is cost discipline rather than a mark. The bear is right that the Q4 decline is coming, but it is coming from marketing investment the company chose, not from revenue it lost.
Debate: Does a record hack at Binance help or hurt Coinbase?
Bull view: A growing view holds that every major failure at an offshore venue has ended with share moving to regulated, audited platforms, and that Coinbase, as the platform storing the most crypto and a major custodian for U.S. spot ETFs, is the natural beneficiary of a $6.3B theft at its largest competitor.
Bear view: The bear camp contends that a theft of this size damages confidence in the whole asset class, pressures crypto prices that drive every Coinbase revenue line, and invites regulatory attention that does not distinguish between venues. Coinbase is a high-beta proxy for crypto sentiment, and the reaction session showed it.
Our take: The bear has the near term and the bull has the medium term, as in 2022. What is different is that Binance has said it will cover customer losses and about half the stolen assets have been frozen, which, if it holds, limits the contagion that made FTX so damaging. We would treat price weakness tied to the hack as an entry point, not a reason to leave.
Debate: Is prediction-market revenue real diversification?
Bull view: The bull case being made on the Street is that prediction markets have reached $200M-plus of annualized revenue in eight months, grew through slower sports months, and are less correlated to crypto volatility than any other Coinbase line, which makes them the first genuine hedge against the spot cycle.
Bear view: Some desks argue the line is lower quality than it looks: it carries exchange fees that are already pushing up transaction expense, faces regulatory uncertainty and well-funded competitors, and is growing while the user base shrinks, which suggests it is monetizing existing customers rather than adding new ones.
Our take: The bear's points are fair on margin and on users, but they argue about the quality of a revenue line that the bear case said in Q2 would fade after a seasonal peak. It did not. We now give prediction markets credit as a diversifier and will judge it next on whether the in-house clearing plan lifts its margin.
Debate: Is the stock cheap at the lows?
Bull view: The constructive camp notes the shares touched a 52-week intraday low after the best operating quarter in a year, below where many investors last sold, with assets on platform back at peak share and the cost base reset.
Bear view: The skeptics answer that the stock still trades at about 27x trailing Adjusted EBITDA, that trailing EBITDA has more than halved, and that Q4 EBITDA is likely to fall again. On current earnings it is not cheap.
Our take: On trailing earnings the bear is right, and that is the wrong lens for a cyclical at a trough. On mid-cycle EBITDA of $2.0–2.5B, which this cost base can deliver on a partial volume recovery, the shares trade at roughly 15–19x, and the operating trend has now turned in the stock's favor.
Model Update Needed
Our standing model was rolled to 2Q26 actuals in October. Because Coinbase stopped reporting trading volume as a key metric, each transaction line now compounds from its 2Q26 actual, and 3Q26E was anchored to the July pace: $464M of transaction revenue and $540M of subscription and services. Q3 beat both, and the expense lines came in below the model's revenue-ratio drivers.
| Item | Current model | Suggested change | Reason |
|---|---|---|---|
| Transaction revenue | 3Q26E $464M (consumer -25% Q/Q, institutional and other -15%); 4Q26E ~$577M on a rebound | Re-base on the $590.2M actual; 4Q26E ~$600M at the October pace, then low-single-digit sequential growth | Q3 beat the model by 27%; August and September ran about 40% above the July pace |
| Prediction markets | Inside consumer transaction revenue | Separate driver at $200M+ annualized, with growth slowing from roughly 60% per quarter | Now the swing factor in consumer revenue; needs its own line and its own fee drag |
| Stablecoin revenue | +1% in 3Q26E, then +2–3% per quarter | Split into platform balances (~3.6% yield) and off-platform balances (~1.0%) | Actual +5.1%; the mix shift onto the platform, not market-cap growth, drove Q3, and management now calls rates a tailwind |
| Blockchain rewards | -8% in 3Q26E, then +6% to +3% per quarter | Re-base on the $85.7M actual; keep the link to token prices | Actual +2.9% as prices rebounded |
| Interest & finance fee income | -3% in 3Q26E, then +2–3% per quarter | Re-base; drive off the loan book | Actual +3.7% on a record loan book |
| Corporate interest & other | -2% per quarter | ~$50M per quarter | Actual -16.2% Q/Q after the $1.27B convertible repayment reduced cash |
| Transaction expense | 15.0% of net revenue in 3Q26E | 16.5–17% of net revenue | 16.8% in Q3, rising with prediction-market exchange fees |
| T&D + G&A + S&M (GAAP) | Revenue-ratio drivers; FY26E $4,427M | Anchor to the absolute guide: FY26 GAAP $4,290–4,450M (adjusted $4.15–4.30B plus amortization); 4Q26E ~$1,070–1,120M | Q3 actual $1,008.8M, below the guide; ratio drivers overstate cost when revenue beats |
| Gains/losses on crypto held for operations | Zero (marks not forecast) | Keep at zero; flag as a swing item inside Adjusted EBITDA | $(31.7)M in Q2, +$26.1M in Q3; 46% of the Q3 EBITDA increase |
| Stock-based compensation | ~22.9% of revenue in 3Q26E (~$245M guide) | ~$240M per quarter in absolute terms | $233.0M in Q3; Q4 guide ~$240M |
| Buybacks and share count | $300M per quarter in 2H26, $400M per quarter in 2027 | 3Q26 at the $5.3M actual; no buyback until it resumes; ~265M basic, ~288M diluted | Q3 repurchases were 37,995 shares |
| Tax rate | 20% | Keep 20%; treat Q3's 12.8% as one-time | Valuation-allowance release and audit closure |
Valuation impact. At the October 30 close of $143.29 and 264.8M shares, market capitalization is approximately $37.9B. That is roughly 27x trailing-twelve-month Adjusted EBITDA of $1,409.9M and about 28.5x the Q3 run-rate annualized ($1,331.6M). On current earnings the stock is not cheap. Against our mid-cycle anchor of $2.0–2.5B of Adjusted EBITDA, unchanged since August, it trades at roughly 15–19x, and the cost base that would deliver that anchor is now lower than when we set it.
| Scenario (12 months) | Adjusted EBITDA | Multiple | Value per share | vs. $143.29 |
|---|---|---|---|---|
| Bear: Q4 at the guide midpoint (~$235M) becomes the run-rate | ~$0.95B | 25x | ~$90 | -37% |
| Base: partial volume recovery on the current cost base | $2.25B mid-cycle | 20x | ~$170 | +19% |
| Bull: fuller recovery, prediction markets keep scaling | $2.5B | 22x | ~$208 | +45% |
The skew favors the upside, and the base case clears the rating bar. The bear case is not remote, though: it requires only that October's activity level persists for a year. That is why conviction is 5 rather than higher.
Thesis Scorecard Post-Earnings
Scored against the standing thesis as last updated at Q2 2026. Pillar numbering is unchanged so the arc stays comparable. Bear #4 was retired at Q2 and folded into Bear #6.
| Thesis point | Status | Notes |
|---|---|---|
| Bull #1: Platform diversification, S&S as recurring-revenue buffer | Confirmed | AT RISK → ON TRACK. S&S $570.2M (+2.7% Q/Q) landed in the upper half of guidance after Q2's miss below the floor; 49% of net revenue; Q4 guide midpoint $590M. The levers that helped (rates, prices) are the same ones that hurt in Q2 |
| Bull #2: Market-share gain through cycle | Confirmed | ON TRACK. 10.7% total share, a fourth consecutive gain; assets-on-platform share back to its 12.5% peak. Caveat: the spot gain came from stablecoin demand, and consumer crypto spot volume fell faster than the market |
| Bull #3: Derivatives expansion (Deribit plus retail) | Confirmed on volume, flat on revenue | ON TRACK. Derivatives volume -6% vs. market -11%, a fourth consecutive share gain; Deribit integration completed. Institutional revenue -8.4% Q/Q and -32.1% Y/Y; the Q2 year-over-year growth was largely the acquisition |
| Bull #4: Stablecoin / USDC float monetization | Improving | AT RISK (unchanged). Revenue +5.1% Q/Q on a record $22B platform balance and higher rates, but still -5.4% Y/Y; USDC market cap -5% Q/Q; one counterparty now 27% of total revenue |
| Bull #5: Prediction markets as new growth engine | Confirmed + | ON TRACK. $200M+ annualized, about 60% above Q2, through slower sports months; daily traders +60%. The seasonal objection is answered. Margin is the next test |
| Bull #6: Base and onchain payments optionality | Neutral | ON TRACK. Borrow/lend balances at a record $1.87B; tokenized stocks past $1B on Base; agentic volume +20% a week on a small base. Other transaction revenue +2.2%. Optionality, not earnings |
| Bull #7: Operating discipline, positive Adj. EBITDA plus cost framework | Confirmed + | ON TRACK. Adjusted expenses below the guide floor; FY26 guide cut a second time, midpoint now below 2025; Adjusted EBITDA +60% for a 15th consecutive positive quarter; operating income positive. Q4 guide steps up 11% for marketing |
| Bull #8: Revenue diversification across 12-plus products at $100M ARR | Neutral | ON TRACK but undisclosed. 25% of transaction revenue from predictions and derivatives; the product count has not been updated since Q1 |
| Bear #1: Transaction-revenue cyclicality | Materialized, no longer deepening | MATERIALIZING (unchanged). -1.5% Q/Q after -20.7%; October pace +3%. But -43.6% Y/Y, MTUs 7.4M and falling, and volatility at multi-year lows. The sequential damage stopped because of prediction markets, not because spot recovered |
| Bear #2: Equity-portfolio overlay creates GAAP optics | Confirmed, sign reversed | CONTAINED. A $715.0M gain produced $718.7M of net income. The same distortion as Q2 in the flattering direction |
| Bear #3: S&S compression on rate cycle plus staking yields | Receding | MATERIALIZING → EMERGING. Management cites higher rates as a tailwind into Q4; stablecoin and rewards both grew sequentially. Still negative year over year |
| Bear #5: Restructuring as defensive signal | Contained | CONTAINED. $4.6M accrual reversal; headcount down a further 2.7%; no new leadership departures raised |
| Bear #6: Unit economics asserted rather than disclosed | Partly addressed | EMERGING (unchanged). New: about 50% of USDC economics retained, Coinbase One revenue per user 3.5x non-subscribers. Still missing: subscriber count, product count, prediction-market take rate, and MTUs outside the 10-Q |
| Bear #7: Operating loss with no cushion | Improved, cushion thin | EMERGING (unchanged). Operating income $61.7M, but about $13M excluding the crypto-ops gain, data-theft recoveries and the restructuring reversal. SBC fell to 70% of Adjusted EBITDA from 115%. Buyback paused while issuance continued |
Overall: The thesis strengthened for the first time in five quarters. Of the eight bull pillars, four are confirmed or strengthened, two are improving or confirmed on volume, and two are neutral; none is challenged. The S&S buffer returns to ON TRACK after landing in the upper half of its guide, and one bear point recedes as the rate backdrop turned. What did not improve is the consumer user base and the disclosure around it, and the operating cushion is thinner than the headline operating income suggests. The business stopped shrinking before the crypto cycle turned, which is the strongest evidence yet that the diversification strategy changed the shape of the downturn.
Action: Upgrade to Outperform. We would add on weakness tied to the Binance breach rather than chase a rebound. Triggers for the Q4 print: back to Hold on Adjusted EBITDA below $200M, transaction revenue below $550M, or MTUs below 7.0M; higher conviction on a resumed buyback, MTUs stabilizing, and prediction markets past $300M of annualized revenue with a stable transaction-expense ratio. The most informative datapoints before then are crypto volatility and whether the Binance freeze holds.