Everything Exchange Builds, Deribit Lands, USDC Compounds — but Circle Marks and an October Fade Pull Us Back to Hold from Outperform.

Rating Action

We initiated coverage of Coinbase at Hold at Q1 2025 on the view that the operating model had three structural tailwinds — (i) S&S compounding via USDC reserve income and Coinbase One subscriber growth, (ii) institutional take rate maturing as Coinbase Prime and Exchange scaled, and (iii) regulatory clarity converting policy investment into product TAM — offset by the volatility of consumer transaction revenue, which remains the single largest driver of GAAP profitability and is highly correlated to crypto-cycle sentiment. We upgraded to Outperform at Q2 2025 when the structural thesis materially strengthened: GENIUS Act passage and SEC Chair Atkins' Project Crypto framework delivered the regulatory inflection the Hold was waiting on, while S&S held at $656M against -40% transaction volumes — proving the recurring-revenue chassis buffers the cycle. The forward-actionable Q3 S&S guide of $665-745M plus the Deribit + Echo + Everything Exchange roadmap converted the platform-diversification thesis from theoretical to fundable.

Today we are downgrading to Hold from Outperform. The operating-layer read on Q3 is, in isolation, our strongest evidence yet that the platform thesis is real: S&S compounded another 14% QoQ to $747M and is now nearly 40% of revenue, consumer transaction revenue grew 30%, the Everything Exchange went from talk to live products (DEX integrations, U.S. perps, Deribit closed), and CDP institutional logos read like a who's-who of the U.S. financial system. But three new offsets that the Q2 upgrade did not have to underwrite materially narrow the asymmetry. First, the Circle mark. The $381M unrealized loss on CRCL in a single quarter, in a quarter where USDC's market cap actually grew to a $74B ATH, illustrates the negative convexity that comes with concentrating equity exposure in your supply-side partner: when Circle's stock decoupled from USDC's growth, Coinbase ate the spread. The Q2 upgrade thesis assumed CRCL would track USDC; it didn't. Second, the take-rate compression. Consumer spot volume grew 37% but consumer transaction revenue grew only 30% — a 7-point gap explained by a heavier advanced-trading mix and a "white glove" service launched for high-priority traders, both of which trade fee rate for share. The Q2 thesis underwrote stable take rates; the Q3 datapoint is the first material crack. Third, the late-September fade flagged in the Gala question — September was strong, October less so, and management did not contest the framing — combined with a Q4 outlook that moves S&S sideways and steps OpEx up by ~$100M QoQ on the tech/dev and G&A line. The operating thesis is still intact; the risk/reward is no longer asymmetric.

Catalysts to return to Outperform: (i) a clean Q4 print where consumer take rate stabilizes and S&S accelerates rather than holds, (ii) December 17 H2 product event with substantive launches in stablecoin payments (x402, Base app, business payments wait list converting), tokenized stocks, and prediction markets, (iii) USDC supply growth continuing to outpace competitors with no further Circle mark, (iv) Q4 institutional take-rate stability post-Deribit-rebate-rationalization. Catalysts to downgrade further: (i) a Q4 transaction-revenue miss with continued advanced-trading mix shift, (ii) further Circle mark of comparable magnitude, (iii) regulatory reversal in CFTC perps or stablecoin-payments framework, (iv) OpEx growth that runs ahead of revenue into early 2026. Q4 is the gate.

Results vs. Consensus

Q3 was a strong operational quarter on the operating layer with two below-the-line items that distorted the GAAP read. Total revenue of $1.9 billion and adjusted EBITDA of $801 million were both QoQ-up materially, with the underlying drivers split between transaction revenue (consumer +30%, institutional +122% with Deribit contribution) and a 14% QoQ step in S&S to $747M. The market reaction was muted because of the Circle mark and the OpEx step-up in the Q4 outlook.

The headline operating numbers are uniformly strong. The marks on the GAAP P&L are the noise: the $424M crypto-portfolio gain and the $381M Circle loss roughly offset, leaving adjusted net income of $421M only modestly below GAAP net income of $433M. The structural read is that operating-layer profitability is holding even as Coinbase invests heavily in headcount and integrates two acquisitions in the same quarter.

Segment Performance

Transaction Revenue — Consumer (+30% QoQ to $844M)

Consumer spot volume grew 37% QoQ to $59 billion, outgrowing both U.S. spot (+29%) and global spot (+38%) market growth on a comparable basis. The 7-point gap between volume growth (37%) and revenue growth (30%) is the headline tension. Management was explicit that the gap was driven by a heavier mix of advanced trading volume, which carries a lower fee rate, and by a deliberate effort to attract and retain high-priority traders via a new white-glove service. This is rational — whales drive disproportionate volume and stickiness — but it is mathematically a take-rate compression and warrants tracking. The bull read is that price increases on the long-tail of assets partially offset, that the advanced trader cohort is sticky, and that volume share gains in U.S. spot are themselves a leading indicator. The bear read is that the highest-fee-rate cohort (basic retail) is a smaller share of the volume mix than a year ago, and that the structural take rate may be drifting down with each successive cycle.

Transaction Revenue — Institutional (+122% QoQ to $135M)

The headline number is inflated by Deribit's mid-quarter close (Aug 14) contributing $52 million on a partial-quarter basis. Stripping that out, institutional was up roughly 64% QoQ on an organic basis, driven by both Coinbase Prime and Exchange. Combined Coinbase + Deribit derivatives volume of approximately $840 billion in Q3 reflects a fully-integrated derivatives stack that Coinbase did not have a year ago: spot + perps + options under a single risk-and-balance-sheet umbrella. Management called out that scaling back rebates and incentives in derivatives drove "more solid sticky organic open interest growth," meaning the take-rate trajectory in institutional should turn favorable as those incentives roll off in Q4 and into 2026. Deribit on its own already runs at >75% market share in non-U.S. options; the U.S. options TAM is the multi-quarter optionality being underwritten here.

Subscription & Services (+14% QoQ to $747M)

S&S is the cleanest piece of the thesis. The $747M revenue print is now ~40% of total revenue and growing faster, in QoQ terms, than transaction revenue's blend. Management cited three drivers: USDC platform balances ($15B average; $74B market cap ATH supply-side), institutional financing average loan balances, and assets under custody ($516B). All three are native-unit growth lines, not price-led, which means the segment is genuinely compounding on adoption rather than rebasing on rate or fee. The interest-rate sensitivity remains the swing factor — reserve income on USDC is rate-sensitive — but the supply growth in Q3 (USDC compounding more than 2x its largest competitor by management's framing) creates a unit-led offset to any rate compression that may eventually come. Q4 guide of $710M–$790M implies flat-to-up-modestly QoQ; on USDC supply alone the bar should be clearable, and the wider range reflects rate uncertainty more than supply uncertainty.

Other — Crypto Investments and Circle Stake

Coinbase carries a sizable balance-sheet position in crypto investments (long-term crypto investments of $2.6 billion at quarter-end, in addition to $11.9 billion in USD resources) and a strategic equity stake in Circle (CRCL). The fair-value mechanics produced a $424M gain on the crypto portfolio and a $381M unrealized loss on the Circle stake in the same quarter. Net of each other these wash to a $43M net positive, but the magnitude of either side — ~$400M each — will continue to dominate quarterly GAAP optics until Coinbase either trims the Circle position or the market reprices it back up. We treat both lines as below-the-line for thesis purposes, but flag the Circle drag as a real read on the equity-stake mode of capital deployment that deserves a higher hurdle going forward.

Key Topics with Management Commentary

The Everything Exchange — Now Live, Not Just Announced

Q2 introduced the Everything Exchange as a strategic vision; Q3 is the first quarter where it shows up as concrete shipped product. Two pieces in particular — DEX integrations and CFTC-regulated U.S. perps — are non-trivial.

"We turbocharged our trading platform in Q3 by adding decentralized exchange or DEX integrations, which expanded access to tradable assets from about 300 to over 40,000 assets in the U.S. With DEX integrated under the hood, customers get day 1 access to new tokens as they are created, and we capture the upside when one of those takes off."
— Brian Armstrong, Co-Founder & CEO

The strategic significance is that Coinbase is converting itself from a 300-asset spot exchange into a 40,000-asset trading platform without a corresponding inventory or listing-overhead burden. The DEX integration outsources the asset-list problem to the underlying decentralized exchanges while monetizing through the front-end. Combined with U.S.-style perps (where Coinbase claims first-mover regulatory status) and Deribit options, the trading-product portfolio at year-end Q3 is qualitatively wider than it was at year-end Q2.

Stablecoin Payments — The Ramp Is the Story

Armstrong devoted a meaningful portion of his prepared remarks and Q&A response to stablecoin payments. The framing is that USDC's $74B market cap is necessary but not sufficient — the next leg requires usage in commerce, not just trading and treasury. Coinbase's response is a stack: Base layer-2 + USDC + payments APIs + the Base app + Coinbase Business + the x402 protocol. The x402 integration with Cloudflare, Vercel, and Google is the most consequential developer-side hook of Q3 because it embeds stablecoin payments in HTTP itself.

"Cross-border payments are something like $40 trillion in volume annually. B2B is 75% of that, which is an early use case for stablecoins. And we're now seeing about $100 billion in annual stablecoin volume, which is growing rapidly."
— Brian Armstrong

The unit economics are compelling on paper — sub-$0.01 fees vs. 2–3% on cards, sub-second settlement — but adoption is still in the "build the rails" phase. Coinbase Business has 1,000 onboarded with 1,000 on the wait list. Shopify USDC checkout (with 1% consumer cashback) is the first credible large-merchant proof point. The thesis question is whether 2026 monetizes at scale or remains a build year.

USDC Compounding — The S&S Engine

The single cleanest data point of the quarter is the USDC supply trajectory.

"In Q3, Coinbase customers held on average, $15 billion of USDC on platform, making us the largest contributor to USDC's all-time high $74 billion market cap. USDC continues to be the top-performing major stablecoin in the crypto ecosystem, growing more than 2x as much as the largest competitor."
— Brian Armstrong

The strategic value of USDC supply is that it converts the policy-and-regulation tailwind (GENIUS Act, regulatory clarity) into a compounding native-unit balance that drives reserve income directly into S&S. As long as USDC supply growth outpaces rate compression, the segment compounds. The risk is the inverse: a sharp rate-cut cycle without offsetting supply growth would compress S&S in a way that consumer transaction revenue can't easily backfill.

CDP — The Infrastructure Layer Is Becoming a Real Business

"We've been able to close 264 institutions now that are using that product, including large companies like JPMorgan, BlackRock, Citi, PNC, fintechs like Stripe, PayPal, Revolut, Webull. So I think that this is going to increasingly be an important part of our business... it's similar to what Amazon did with AWS."
— Brian Armstrong

The AWS analogy is aspirational but the customer logos are real. CDP monetizes by selling crypto-as-a-service infrastructure (custody, USDC integration, on-chain rails) to institutions that don't want to build it themselves. The Citi partnership announced in mid-October is the marquee logo addition. CDP isn't yet a separately-disclosed reporting line, but management's increasing willingness to cite the institutional client count quarterly suggests it will be soon.

Deribit Integration — Already Material, Not Yet Fully Integrated

"Deribit is already the market leader in options. They had over 75% market share for options. Notably, this is all non-U.S. And so there is path to grow the market for options in the U.S. That is going to be a multi-quarter road map of bringing both the regulatory licenses and product to bear in the U.S., but we think that's a huge opportunity for us."
— Alesia Haas, CFO

Deribit closed Aug 14, contributed $52M to revenue and $30M to OpEx in the partial quarter (including $16M in deal-related amortization), and onboarded 100 employees in September. Cross-margining between spot and derivatives went live for U.S. customers in Q3. The U.S. options ramp is a multi-quarter regulatory path, not a Q4 unlock.

October 10 Liquidation Event — A Resilience Datapoint

James Yaro at Goldman asked about the October 10 crypto-wide liquidation event. The answer was a positive read on Coinbase's risk infrastructure.

"We are really pleased that we did not see significant liquidations on our platform... we actually operated very well without disruption. And we didn't have any downtime or degraded latency around market data or anything like that... Several major exchanges experienced extended outages during that time, and we didn't have any."
— Alesia Haas (with addendum from Brian Armstrong)

This is a quiet but important datapoint. In a market structure where competitor downtime equals share gain, Coinbase's resilience during a stress event is a hidden positive. The corollary — that the AWS outages earlier in the year did briefly affect Coinbase — was acknowledged honestly, and Armstrong's framing of multi-cloud as a cost-benefit trade-off rather than a target was the right answer for a CFO-conscious audience.

Guidance — Q4 2025

The Q4 outlook is best read as a profitable consolidation quarter, not an acceleration quarter. October transaction revenue at ~$385M is below the implied monthly run-rate of Q3's $844M consumer figure (~$281M/month consumer-only), which suggests October is in line on consumer but the full-Q4 print depends on Nov/Dec follow-through. The OpEx step-up of ~$100M QoQ at midpoint — half from full-period Deribit/Echo, half from headcount — is a known quantity, but it does compress incremental margins. Management's explicit signal that 2026 will see "sequential rate of operating expense growth slow as compared to our Q4 rate" is the multi-quarter unlock.

Analyst Q&A — Paraphrased

Craig Siegenthaler (Bank of America) asked how the Echo acquisition expands Coinbase's network in private and public capital formation. Armstrong framed Echo as a vertical-integration play — capital formation moving on-chain, with Coinbase's $516B asset base providing a built-in two-sided marketplace for token-issuance distribution.

Kenneth Worthington (JPMorgan) asked whether the political and regulatory clarity is increasing Coinbase's pace of M&A. Choi confirmed the pace has increased, framed it in the lineage of "best tech companies of all time" using M&A to accelerate adoption, and said the framework is buy/build/partner/invest, evaluated case-by-case, with priorities anchored to trading and payments.

Peter Christiansen (Citi) asked about operating-infrastructure resilience after recent cloud outages. Armstrong acknowledged AWS impact, framed multi-cloud redundancy as a cost-benefit trade-off rather than an absolute, and pointed to October 10 as a resilience datapoint where Coinbase had no downtime while several major exchanges had extended outages. Choi added that 65% of customer support is fully automated and that compliance-investigation LLM agents are being rolled out in 2026.

Benjamin Budish (Barclays) asked about the white-glove service for advanced traders and the broader competitive dynamic in retail crypto. Haas framed white-glove as a high-value-trader concierge offering with personal account management, and emphasized that Coinbase's Q3 trading volume exceeded overall U.S. spot volume — a share-gain read.

Owen Lau (Clear Street) asked about CDP and the Citi partnership. Armstrong used the AWS-for-crypto framing again and disclosed the 264 institutional logos. Haas added that Coinbase Business has onboarded over 1,000 SMBs with another 1,000 on the wait list.

Devin Ryan (Citizens) asked about Deribit integration progress. Haas detailed the August 14 close, $52M revenue and 100-employee onboard, and the spot/derivatives cross-margining go-live in Q3.

Patrick Moley (Piper Sandler) asked about Everything Exchange milestones. Armstrong pointed to live products (DEX integrations, U.S. perps) and the December 17 H2 product event as the next milestone.

James Yaro (Goldman Sachs) asked about the October 10 liquidation event. Haas confirmed Coinbase did not see significant liquidations and the platform performed without disruption, framing transparency from publicly-listed exchanges as a structural risk-reducer over time.

Andrew Jeffrey (William Blair) asked about commercial stablecoin adoption and cross-border timing. Armstrong cited the $40T cross-border market, ~$100B current stablecoin volume, B2B as the early use case, and cited x402 with Cloudflare/Vercel/Google as the developer-side flywheel.

Bo Pei (US Tiger) asked whether the scale-back of rebates in derivatives would expand margins in Q4 and 2026. Haas declined to quantify the take-rate impact but confirmed the rebate scale-back drove "more solid sticky organic open interest growth" in the international derivatives platform.

Alex Markgraff (KeyBanc) asked how management thinks about contribution margin across the expanding product portfolio. Haas framed it as customer-relationship economics rather than per-product margin, with institutional clients now engaged with three-plus products on average.

Dan Dolev (Mizuho) asked about the take-rate trajectory a few quarters out. Haas declined to commit to a forward-take-rate number, citing pricing experimentation on the retail side and the goal of reducing reliance on trading fees as a single monetization vector.

Edward Engel (Compass Point) asked whether Q4 OpEx is the bottom of the reinvestment cycle or whether headcount continues to ramp. Haas confirmed half the QoQ OpEx step-up is from Deribit + Echo full-period and half from continued headcount, with 2026 framed as an absorption year rather than a ramp year.

Zach Gunn (FT Partners) asked about incentive economics for stablecoin payments adoption. Armstrong framed the payments problem as a network-effects challenge but pointed to Coinbase's $500B asset base, the open-network character of crypto, and the B2B cross-border use case as the unmet-need wedges.

Joseph Vafi (Canaccord) asked about Deribit's options-share opportunity and margin structure relative to spot. Haas confirmed Deribit's >75% non-U.S. options share and framed U.S. options as a multi-quarter regulatory road map, declining to provide a clean spot-vs-derivatives margin comparison given product customization.

Gustavo Gala (Monness, Crespi and Hardt) asked about the September-to-October competitive dynamics — specifically why September saw outperformance and October appeared to reverse. Haas did not contest the framing, declined to attribute it to any specific competitor pressure, and reiterated that Coinbase always faces competition. This is the question we noted in the Rating Action above.

What They're NOT Saying

• A specific margin profile for the Base app, the payments stack, or the CDP business. Haas explicitly declined to give per-product contribution margins, framing the company as managing toward total adjusted EBITDA growth rather than per-product margin. This is fine for a multi-product growth phase, but eventually investors will need a way to mark the unit economics of the non-trading businesses to underwrite the platform thesis.
• The Circle position size or any plan to trim it. The $381M unrealized loss on Circle was disclosed and explained, but management gave no signal on whether the position would be held, trimmed, or hedged. In a quarter where USDC's market cap actually grew to a $74B ATH, the gap between Coinbase's USDC unit-growth story and the Circle equity drag is a structural tension that will recur in any quarter the stocks decouple.
• Forward take-rate guidance. Haas declined Dolev's question explicitly. The 7-point gap between consumer volume growth (+37%) and consumer transaction revenue growth (+30%) is the most-asked-about line item of the quarter and the answer was "we're experimenting." Acceptable for now; less acceptable if the gap widens in Q4.
• The October growth trajectory beyond the headline $385M transaction revenue. The Gala question got at this directly — September strong, October less so — and management did not provide additional color. We read this as either (a) October is genuinely softer than September on a per-day basis, or (b) management is being conservative on a single-month QTD print. Either way the silence is itself a signal.
• A specific timeline for tokenized stocks, prediction markets, or U.S. options. Armstrong pointed to December 17 as the next milestone but committed to no specific go-live dates for the named asset classes. The H2 product event will be where these get crystallized; until then the Everything Exchange remains a partial-credit story.
• 2026 OpEx growth in absolute dollars. Haas committed to a slower rate of sequential OpEx growth in 2026, but no absolute dollar guide. With Deribit + Echo full-period in the base, the absorption-year framing matters; investors will quantify it themselves until the company does.

Market Reaction

The print was released after the close on October 30, 2025. The initial market reaction was mixed: the operating-layer numbers (revenue, EBITDA, S&S compounding, USDC growth) read as cleanly positive, but the Circle mark, the take-rate compression in advanced trading, and the Q4 OpEx step-up created two-handed-economist coverage in the after-hours wrap. The shares traded in a wide range overnight as algo readers parsed the GAAP/adjusted bridge. By the morning of October 31, the print was trading as a "good quarter, expensive setup" — a fair description that aligns with our maintain-Hold posture. The single most-debated line on after-hours desks was the consumer transaction take rate; the second was whether the Circle drag will recur. Both are tracker items into Q4.

Street Perspective

The Street has been broadly constructive on COIN through 2025 on the platform-thesis logic: Everything Exchange + Deribit + Base + USDC compounds to a multi-segment franchise that can monetize through the cycle rather than only when retail spot trading is hot. Q3 reinforces that bull case on the operating layer. The bull case being made on the Street is that S&S compounding plus institutional take-rate stability post-Deribit-rebate-rationalization can carry profitability even if consumer transaction revenue is range-bound; that the Everything Exchange will crystallize at the December 17 product event with credible tokenized-stock and prediction-market launches; and that USDC supply growth will continue to outpace competitors regardless of the rate environment. The bear case being made on the Street is that the consumer-transaction take rate is structurally drifting lower with each cycle, that the Circle stake creates negative convexity when CRCL underperforms USDC, and that the Q4 OpEx step-up plus a reinvestment-mode 2026 will cap incremental margin expansion. Our Hold reflects neither bull nor bear fully — we agree with the platform-thesis logic but want one more quarter of clean execution and stable take rate before underwriting the upgrade.

Model Implications

• S&S compounding remains the cleanest line. $747M Q3 + $710M–$790M Q4 guide implies ~$2.95B–$3.0B run-rate exiting 2025, growing on USDC supply and institutional financing. This is now ~40% of revenue and our model adjusts the segment growth contribution upward.
• Consumer take rate adjusts down. The 7-point volume-revenue gap (37% vs. 30%) is meaningful enough to embed in our Q4 and 2026 model assumptions. We move our consumer take-rate expectation modestly lower to reflect ongoing advanced-trading mix shift, with offsetting volume growth.
• Institutional gets a Deribit step-up but a take-rate question. Q4 will be the first full Deribit quarter ($52M was a partial-quarter contribution); we extrapolate to a roughly $130–$150M Deribit-revenue Q4 contribution at constant volumes, plus organic Coinbase Prime and Exchange. The rebate-scale-back is a tailwind into 2026.
• OpEx step-up sized. Tech/Dev + G&A of $925M–$975M Q4 vs. ~$870M Q3 implied, plus S&M $215M–$315M, gives a wide Q4 OpEx range we model toward the midpoint. Half the step-up is acquisition-related and is structural; the other half is headcount and decelerates in 2026.
• Below-the-line items modeled separately. We continue to exclude crypto-portfolio FV gains/losses and the Circle mark from our adjusted-net-income view, but flag that GAAP volatility from these lines will remain a feature, not a bug, of any quarter where crypto prices or CRCL move materially.
• Q4 transaction revenue scenarios. October at $385M annualizes to roughly $1.15B if Nov/Dec hold the run-rate; vs. Q3's $979M total transaction revenue, that would be flat-to-up modestly. We model a Q4 base of $1.05–$1.20B transaction revenue range, depending on November/December crypto-cycle posture.

Thesis Scorecard

Net thesis score: seven confirmed pillars, two compression flags, one building. The platform thesis is intact and is the dominant signal. The two flags — consumer take-rate and the Circle equity-stake mode — are the gating items that pulled us back from Outperform to Hold this quarter. The December 17 H2 product event plus the Q4 print together constitute the gate to return to Outperform; a clean Q4 plus credible product-event launches would crystallize the platform thesis fully. We sit at Hold pending those.