Record Share, Shrinking Pie: Cost Discipline Is Losing the Race to the Revenue Reset

Results vs. Consensus

Coinbase missed on every line that matters and missed for the third consecutive quarter. The composition of the miss is what determines whether this is a cyclical trough or a structural de-rating, and this quarter the composition got modestly worse: the transaction line missed for cyclical reasons we already understood, but the subscription line missed for reasons that are partly structural and partly self-inflicted forecasting.

A note on the EPS basis. The two consensus services we track disagreed sharply on this quarter's EPS: one carried a positive $0.14, the other a loss of $(0.17). That $0.31 spread reflects genuine disagreement about whether the crypto-asset mark and the restructuring charge sit above or below the adjusted line, not a data error. Because they disagree, we present the range rather than manufacture a single number, and we lead on GAAP $(1.36), which is unambiguous. On the company's own adjusted basis, the reported loss was $(104.9)M, roughly $(0.40) per share against the quarter-end share count.

The scorecard that matters more: management against its own guide

Street consensus is a survey. The Q2 outlook Coinbase published in May is a commitment. Management put both on the same page of the deck, and four of the five lines carry a check mark.

Every line management controls came in on or below plan. The one line that depends on markets and rates came in below the floor. That is a clean split, and it is the single most useful thing in the release: it tells you the cost program is executing to specification and it tells you the revenue base is deteriorating faster than management's own forecasting could keep up with, over a horizon of only eleven weeks.

Sequential and year-over-year

The Q2'25 comparatives deserve a caveat before anyone anchors on the year-over-year net income line. That quarter carried $1,506.9M of other income, which is what produced $1,428.9M of net income on an operating loss of $(24.7)M. Comparing this quarter's $(359.5)M to that number measures the absence of a one-time gain, not operating deterioration. The honest year-over-year comparisons are revenue (-18.5%), Adjusted EBITDA (-59.4%), and the operating line, which went from a $(24.7)M loss to a $(113.5)M loss.

Revenue

Total revenue of $1,220.1M was the lowest quarterly print since Q3 2024 and the third consecutive sequential decline. The decomposition is unambiguous: total revenue fell $192.9M sequentially, of which $156.6M was transaction revenue, $28.4M subscription and services, and $7.8M corporate interest and other income (the components differ from the total by $0.1M on rounding). Transaction revenue did outperform its own market, falling 20.7% against a 25% decline in total market crypto spot trading volume, and management is entitled to that point. But outperformance measured against a collapsing benchmark still produces a shrinking business, and the year-over-year comparison (-21.6%) is now nearly identical to the sequential one, which tells you the deterioration is not decelerating.

Margins and operating leverage

The operating loss of $(113.5)M is the number that should concern a shareholder more than the headline net loss, because it contains no marks. It widened more than fivefold from $(21.4)M in Q1 and it is the second consecutive quarterly operating loss. Adjusted expenses at $1,034.8M consumed 89.7% of net revenue, against 84.6% in Q1 and 69.6% in Q2'25. Management is cutting hard and the cuts are landing, but the revenue line is falling roughly twice as fast as the cost line, and that gap is the entire investment question for the next two quarters.

Below the line and the crypto mark

For the fourth consecutive quarter, an unrealized crypto-asset mark dominates the GAAP optics. The $209.5M loss on crypto assets held for investment follows $482.4M in Q1 and $718.2M in Q4'25. We have flagged this since Q3 2025 as a presentation problem rather than an economic one, and that assessment is unchanged: Coinbase holds crypto on the balance sheet as a strategic position and marks it every quarter, which imports the asset class's volatility directly into reported earnings. What has changed is that the mark is no longer the reason for the loss. In Q4'25 and Q1'26 you could strip the mark and find a profitable operating business underneath. This quarter you cannot.

Segment Performance

Transaction revenue

Total row is the income-statement figure. Component rows are the whole-dollar amounts disclosed in the earnings presentation and may not sum to the total exactly.

The year-over-year column is the one to read, and it says something the sequential column hides. Consumer transaction revenue fell 30.5% year over year while institutional transaction revenue rose 63.9%. The entire year-over-year decline in transaction revenue, and more, is consumer. The institutional franchise that Coinbase spent 2025 acquiring and integrating is growing substantially through the worst crypto tape in three years.

Consumer

Consumer transaction revenue of $452M fell 20.3% sequentially against a 24% decline in consumer crypto spot trading volume. Revenue falling less than volume means the consumer take rate improved, and management attributes that to prediction markets, which contribute to transaction revenue but are not captured in the spot-only volume metric. This is the second consecutive quarter in which the newer, higher-take products have cushioned the consumer line against the spot cycle.

Assessment: The mix shift is working exactly as the diversification thesis predicted, and it is worth more than the headline decline suggests. But it is cushioning, not offsetting. A 20% sequential decline in the largest revenue line, in a quarter where the newer products more than doubled, tells you how severe the underlying spot contraction was.

Institutional

Institutional transaction revenue of $100M fell 26.5% sequentially, which management characterized as roughly in line with market declines, but rose 63.9% year over year from $61M. This is the line where the Deribit acquisition and the derivatives build-out were supposed to change the cyclical shape, and on both the volume and the revenue metrics they have. Crypto derivatives trading volume was flat sequentially on a trailing-twelve-month basis at $4.22T while the derivatives market overall fell 12%, delivering a third consecutive quarter of derivatives share gain and an all-time-high global derivatives share.

Assessment: This is the most under-appreciated result in the release and the sequential decline obscures it. A 64% year-over-year increase in institutional transaction revenue, in a year when total crypto market cap fell and spot volumes collapsed, is the derivatives thesis converting from volume statistics into money. It is also the reason to take management's share commentary seriously rather than dismiss it as consolation. Two qualifications keep this from being a rating-changing datapoint. The base is small, at roughly 17% of transaction revenue, so 64% growth off $61M moves the total line by less than $40M. And derivatives are high-notional and low-take by construction, so growing their mix mechanically compresses the blended take rate even when per-product economics are stable. That is a consequence of the strategy working rather than failing, but it needs to be modelled explicitly instead of treated as upside.

Other

Other transaction revenue of $47M fell 11.3% sequentially, which management attributed largely to lower Base revenue. The smallest of the three lines and the shallowest decline.

Assessment: Base's strategic value is being built in volume share and agentic settlement, not in this revenue line, and the deck is explicit about that framing. Investors should not expect Base to become a material revenue contributor on the current architecture.

Subscription and services

Stablecoin revenue

This is the most analytically important line in the quarter. Stablecoin revenue of $292M fell 4.3% sequentially and 5.5% year over year. Over that same year, average USDC held in Coinbase products rose 44% to an all-time-high $20B, average USDC market cap rose from $61B to a record $77B, and more than 30% of USDC in circulation now sits inside Coinbase products. Every volume and balance metric set a record and the revenue line went backwards.

"So we have already met the conditions for the circle contract to renew, it will renew on the same terms. I wanna take away any ambiguity about that, for the market."
— Alesia Haas, CFO

The arithmetic behind the divergence is straightforward. Measured against average USDC market cap, the blended annualized yield Coinbase realized fell from roughly 2.0% in Q2'25 to roughly 1.5% in Q2'26, a compression of about 50 basis points. That is a rate-cycle outcome, and Coinbase controls none of it. What Coinbase does control is the balance base, and it grew that base 26% on market cap and 44% on platform-held balances over the same period.

Assessment: The stablecoin pillar has crossed an important line this quarter. For eight quarters, balance growth outran yield compression and the line grew. It no longer does. Coinbase is now running up an escalator moving down, and unless USDC circulation growth re-accelerates above roughly 30% annually, this line goes sideways to down for as long as the front end keeps falling. We are reclassifying this pillar from compressed-but-durable to at risk. It remains a genuinely valuable, high-margin annuity; it is no longer a growth line, and modeling it as one has been the single largest source of forecast error in our own numbers.

Blockchain rewards

Blockchain rewards of $83M fell 17.8% sequentially and 42.8% year over year, which management attributed to growth in native units staked being more than offset by price effects and lower rewards rates. This line is levered to token prices twice over: the reward accrues in units and is booked in dollars, so a falling market compresses both the rate and the translation.

Assessment: This is the highest-beta line inside what is supposed to be the low-beta half of the business, and at -42.8% year over year it is materially undermining the "subscription and services is a volume-independent buffer" framing. Investors treating the full $555M S&S line as recurring and defensive are over-crediting roughly $83M of it.

Interest and finance fee income

The only line in the entire P&L that grew year over year, up 11.9% to $66M on a new all-time-high average DeFi borrow/lend balance. Average borrow/lend balances reached $1,491M, up from $199M in Q2'25 and $1,330M in Q1'26.

Assessment: Small but structurally interesting. A more than sevenfold year-over-year increase in balances producing only 11.9% revenue growth points to the same rate compression affecting stablecoin. Still, this is the clearest evidence in the release that the onchain product surface is accumulating real, sticky balances rather than transient activity.

Other subscription and services

Other S&S of $114M grew 4.6% sequentially, which management tied to a new all-time high in paid Coinbase One subscribers and the achievement of performance earn-outs. It is also the line management flagged as a headwind into Q3, since those earn-outs roll off.

Assessment: The only sequential grower in S&S, and management has already told us a portion of the growth does not repeat. Notably, no subscriber count was disclosed this quarter after a specific 1M-plus figure at Q1.

Key KPIs

Two KPIs deserve more attention than they will get. The first is assets on platform as a share of total crypto market cap, which fell 80 basis points to 11.2% and broke a long run of stability in the 11.9% to 12.5% band. Management attributed the decline to ETF outflows, given Coinbase's position as primary custodian, noted that trends have stabilized quarter-to-date in Q3, and stated that native units excluding ETFs increased sequentially. That explanation is credible and specific. It is still the first quarter in two years where the custody franchise lost measurable share of the asset pool, and custody share is the upstream input to almost every monetization surface Coinbase operates.

The second is the share count. Coinbase has returned more than $2.0B to shareholders, repurchased more than 10.1M Class A shares, offset more than 85% of stock-based compensation issuance since Q4'24, and has roughly half its authorization remaining. Despite all of that, shares outstanding rose from 256M to 264M year over year, and rose sequentially. At $238.3M of quarterly stock-based compensation against $207.8M of Adjusted EBITDA, the company is issuing more equity value per quarter than it is generating in adjusted cash earnings.

Key Topics & Management Commentary

Overall Management Tone: Management was strategically confident and financially silent, and the gap between the two is the defining feature of this call. There were no prepared remarks and no CFO walk-through of the quarter; the session opened directly into questions and stayed on product, regulation, and competitive positioning for its full length, with the financial deterioration going essentially unaddressed. Tone was materially more forward-leaning than the operating results, and less quantitative than at any point in the four quarters we have covered.

1. The call format itself is now a disclosure issue

Coinbase hosts its earnings call live on X, with a panel that mixes independent creators and podcast hosts with a small number of institutional research analysts. Management opened by framing the audience as customers and community alongside shareholders.

"We are also joined today by a group of independent and institutional research analysts. We are excited to connect directly with you, our customers, our community, our shareholders to talk about our quarter and answer your questions."
— Alesia Haas, CFO

Across the entire call, not one question addressed the $359.5M net loss, the $113.5M operating loss, the $52.4M restructuring charge, the subscription-and-services shortfall against the company's own guide, the 80-basis-point decline in custody share, or the Q3 transaction-revenue run-rate that management had just disclosed. There were no prepared remarks in which the CFO might have addressed them unprompted. The call also did not conclude; it ended when the final questioner's microphone failed and the operator wrapped the session.

Assessment: We have no objection to a public, accessible earnings format, and there is real merit in a CEO taking questions from the customer base. But a format is only additive if it supplements the accountability function rather than replacing it. In a quarter with a third consecutive miss, a widening operating loss, and a guide-down embedded in the outlook, a call that generates zero financial questions is not a communication innovation, it is an absence of scrutiny. We are treating this as a governance-quality mark against the name and it is a contributing reason the rating stays at Hold rather than moving up on the strategic progress.

2. Regulation: CLARITY at the one-yard line

The lead question concerned the CLARITY Act, then out of committee and awaiting a Senate floor vote with the August recess approaching and prediction-market odds of passage cited at roughly 30%. Management was optimistic about a floor vote and, more usefully, explicit about the downside case.

"your question was about what happens if it does not pass, and I think in that world, it is actually kind of just business as usual for Coinbase for a few reasons."
— Brian Armstrong, CEO

The reasoning was that Coinbase already operates to most of what the bill would require, and that the SEC and CFTC have signalled they would issue their own rules regardless. Management's framing was that the marginal loser from failure is the American consumer rather than Coinbase.

Assessment: This is the correct and honest framing, and it de-risks a binary that some of the bull case has been leaning on. It also removes a catalyst. If CLARITY passing is "better on margin" and CLARITY failing is "business as usual," then legislative outcomes are not going to re-rate this stock in either direction, and the investment case rests entirely on the cycle and the cost base.

3. The Circle contract and the multi-stablecoin pivot

Management was asked why Coinbase would join the Onyx USD consortium when some view it as a competing stablecoin platform, and whether the move was leverage in the Circle negotiation. The CFO pre-empted the second half before the first was answered, confirming the Circle contract has already met its auto-renewal conditions and renews on identical terms. The CEO then framed the consortium participation as strategy rather than tactics.

"the short reason is that we are a multi stablecoin platform. We wanna provide stablecoins that all of our customers want to use And where possible, we want to strike good economic arrangements with them."
— Brian Armstrong, CEO

The deck reinforces this, listing PYUSD, USDT, EURC, tGBP and XSGD among supported assets and noting that stablecoins with which Coinbase has a commercial and economic relationship rose to 79% of market stablecoin transaction volume in FY26 year-to-date, from 55% in FY25.

Assessment: The urgency with which the CFO moved to close the Circle question, unprompted and ahead of the actual question asked, tells you how much market anxiety sits on that contract. The clarification is genuinely valuable and the multi-stablecoin framing is strategically sound. But note what it implies: Coinbase is diversifying away from single-issuer dependence at exactly the moment the per-dollar economics on that issuer are compressing. The strategy is right; it is also a hedge against a line item that just turned negative year over year.

4. Sharing USDC economics to buy network effect

The sharpest institutional question of the call concerned the Hyperliquid relationship and whether a large third party holding enough USDC can extract the majority of the network's economics. The CFO did not dispute the mechanism and defended it as deliberate.

"With stablecoins, with underlying protocols, even base, Liquidity network effect are critically important, and so bringing USD deeply into this ecosystem just further drives USDC growth and adoption throughout the globe. So that was our strategy. This is what we think is the right long term strategy for stablecoins, and we are happy to share economics to drive this network effect."
— Alesia Haas, CFO

The CEO added that USDC is already first in stablecoin transaction volume and first among regulated stablecoins, and second to Tether only on market capitalization, arguing that continued economics-sharing is warranted to close that last gap.

Assessment: This is a coherent land-grab strategy and it is also a margin decision that shareholders should price. Coinbase is choosing to share a larger portion of a per-dollar yield that is itself compressing, in exchange for balance growth. That works if circulation compounds faster than yield decays. This quarter it did not: balances hit records and the revenue line fell. Management is buying share of a pool whose economics are deteriorating, and no one asked what the marginal economics of a shared dollar now are.

5. Senior leadership departures

An institutional analyst raised the number of senior departures during the quarter and asked what it signalled about strategy. Both executives pushed back firmly, with the CEO citing bench depth and succession planning.

"there is nothing that is changing about the strategy, but I guess what I would say is 1 of the I am most proud about Coinbase is we have a really deep bench of talent We have a really good succession planning process."
— Brian Armstrong, CEO

The CFO named incoming leaders across people, legal and other functions, adding that each had been groomed by the outgoing leader, and closed with the observation that "these are all individual decisions, so there is nothing from a strategy standpoint to read into these changes."

Assessment: The succession-planning answer is the right answer and the named-successor detail is more than most companies offer. The context still matters: senior turnover clustered in the same quarter as a 14% headcount reduction and a second consecutive operating loss. Individually explicable departures can still aggregate into a signal, and we carry this as an open watch item rather than a resolved one.

6. Cost discipline and the AI-leverage argument

The cost program is the clearest execution story in the release. Headcount fell 13.4% sequentially to 4,321 from 4,988 following the May action. Technology and development fell 10% to $472.8M, G&A fell 5% to $356.9M, and sales and marketing fell 10% to $239.8M. The deck's supporting argument is that AI tooling is decoupling engineering output from headcount, citing pull requests per engineer up 2.2x year over year and integration test coverage across core services up 2.5x over six months, with AI usage growing faster than AI spend.

Full-year adjusted-expense guidance was reduced and narrowed to $4.2–4.45B from an initial $4.25–4.6B, a $100M reduction at the midpoint, against 2025 actual adjusted expenses of $4,282M.

Assessment: This is the pillar that has strengthened most this quarter and it deserves credit. Cutting a cost guide mid-year, in the same release as a revenue miss, is the opposite of the reflexive reinvestment most managements reach for at a cyclical trough. The uncomfortable arithmetic is that the new midpoint of $4,325M is still 1% above 2025's actual spend, in a year where revenue is tracking down roughly 19%. Management is holding costs flat in absolute terms while revenue falls by a fifth. That is discipline, and it is not enough to hold the operating line.

7. Prediction markets and the Everything Exchange

Prediction markets were the standout operating result: contracts and revenue both more than doubled sequentially, up 106%, reaching $100M-plus annualized revenue. The deck attributes the acceleration to new market launches, UI and stability improvements, and seasonally strong sports volume around the NBA playoffs and World Cup soccer, with a new crypto-binaries product driving a 3x increase in daily traders and 4x increase in daily revenue by quarter-end against May's daily average. Management flagged combos and prosumer tooling as the next releases.

The CFO also addressed the cannibalization question directly.

"We have also early signals customers who are engaging with these new products, for example, prediction markets, are also driving incremental spot trading volume. So we are not seeing cannibalization."
— Alesia Haas, CFO

Assessment: Doubling sequentially in a quarter where the core market fell 25% is the single best operating datapoint in the release, and the no-cannibalization evidence, if it holds, makes this genuinely incremental rather than a rotation of the same wallet. Two cautions. The seasonal attribution is management's own, and NBA playoffs plus a World Cup is close to a peak sports calendar, so the Q3 sequential comparison will be hard. And at $100M-plus annualized against $2.4B of trailing transaction revenue, the line still needs to roughly quadruple before it changes the shape of the P&L.

8. Coinbase One and the take-rate question

Management disclosed a new all-time high in paid Coinbase One subscribers, notable for occurring in a down market, and framed subscribers as the most deeply engaged cohort on the platform. Pressed on whether zero-fee trading for members makes the growth accretive or take-rate dilutive, the CFO was unusually candid about the limits of the answer.

"Great question. I am gonna give an unsatisfying answer because when I look at the data on average Coinbase 1 subscribers trade more and have higher unit economics, there is always examples on the edges."
— Alesia Haas, CFO

The substance was that members monetize across staking, card and other surfaces rather than through trading fees, that retention and engagement are better, and that broad adoption would shift where revenue appears in the P&L.

Assessment: Honest, and less reassuring than it sounds. "Revenue will shift through the P&L" is a description of mix migration from a high-visibility line to several lower-visibility ones, and no number was attached to the net effect. Coinbase gave a specific subscriber count at Q1 and gave only a superlative this quarter, which is a step backwards in disclosure on the product it most wants investors to value as an annuity.

9. Agentic finance and whether trust survives machine buyers

The most conceptually interesting exchange challenged the durability of Coinbase's trust-based moat in a world where AI agents transact. The premise was that an agent has no brand loyalty and optimizes purely on cost and latency, which would push Coinbase toward price competition it has historically avoided.

"But I also think AI agents are going to choose infrastructure that is reliable and safe and liquid and compliant and has good uptime. Just like they might choose AWS or some kind of cloud vendor for different types of infrastructure, you could imagine."
— Brian Armstrong, CEO

On sizing, management was explicit that it is early and declined to quantify.

"But it is still quite early, so do not think we have any specific numbers or forecast to share on that at the moment."
— Brian Armstrong, CEO

The deck supports the positioning with shares in the 90%-plus to 99%-plus range across agentic stablecoin settlement volume on Base, x402 protocol usage, and USDC-denominated agentic commerce.

Assessment: The cloud-vendor analogy is the strongest form of the argument and we broadly accept it: agents will optimize a vector, not a scalar, and compliance and uptime are in the vector. But the honest position is that nobody knows, the shares being cited are of a base too small to disclose, and management said so. This is a real option with no current cash flows, and it should be valued as such rather than as a reason to pay up today.

10. The competitive L2 landscape

Asked about Robinhood expanding into crypto and launching its own L2, the CEO reached for a market-structure argument rather than a product one, drawing on what happened when every firm launched a stablecoin.

"So I think what we are what people found out in stablecoins is that there is an actual network effect to stablecoins."
— Brian Armstrong, CEO

The argument extends to blockchains: fragmentation first, consolidation later, with Base positioned as the largest L2 on Ethereum, first in BTC decentralized-exchange spot volume in Q2, and holding what management characterized as roughly a two-year head start. Management also floated the prospect of consolidation among blockchains resembling an M&A process.

Assessment: The stablecoin analogy is well-chosen and the empirical claim underneath it, that incumbent stablecoin share barely moved despite a wave of launches, is checkable and correct. It supports the view that competitor L2s are not an imminent threat to Base's position. It does not address the more immediate competitive question, which is fee and share competition in retail crypto brokerage, where the challenger is not the L2 but the app.

11. Decoupling from Bitcoin

Management was invited to note that Bitcoin-related transactions have fallen from more than half of company revenue to roughly 12%. The deck's own framing is that 88% of net revenue now comes from sources other than BTC spot trading, up from 45% in Q2'20. The CEO's response layered a cyclical claim on top of the structural one.

"And I think Bitcoin will come back in a big way too. By the way, it keeps going through these cycles."
— Brian Armstrong, CEO

Assessment: The diversification is real and measurable, and it is the central achievement of the last three years of strategy. It is worth being precise about what it does and does not buy: revenue has decoupled from Bitcoin's trading fees, not from crypto's cycle. This quarter is the proof. BTC spot is 12% of the business and revenue still fell 13.7% sequentially, because the replacement revenue is levered to the same underlying variables of volume, volatility, token prices and rates. Diversification across products within one asset class is not diversification across cycles.

Guidance & Outlook

Coinbase does not guide to total revenue. It guides to subscription and services revenue, several expense lines, and provides a quarter-to-date transaction-revenue datapoint. That last item is the most consequential disclosure in the release.

Implied quarter-over-quarter ramp. Management disclosed transaction revenue of roughly $130M quarter-to-date through July 26, covering 26 of the quarter's 92 days, and paired it with an explicit caution against extrapolating. Taking the caution seriously and the arithmetic literally: $130M over 26 days is a pace of roughly $5.0M per day, which across 92 days produces approximately $460M of Q3 transaction revenue. That would be a 23% sequential decline, following declines of 20.7% in Q2 and 23.1% in Q1. Three consecutive quarters of low-to-mid-twenties sequential contraction in the largest revenue line.

Street at: Consensus entering the print sat near $1.30B of revenue and roughly $288M of Adjusted EBITDA, both of which were missed by wide margins. Post-print price-target revisions were uniformly downward and the surviving dispersion is extraordinarily wide, spanning a roughly 2.5x range between the lowest and highest published targets. That spread is itself the datapoint: the Street does not agree on what this franchise is worth at a cycle trough, which is a reasonable disagreement to have and a poor foundation for a high-conviction rating in either direction.

Guidance style: Historically conservative on expenses and reliably delivered. This quarter breaks the pattern on the revenue side for the first time in our coverage: the subscription-and-services range set in May proved too high across an eleven-week horizon on the one line management describes as predictable. We are lowering our confidence in the S&S guide as a forecasting instrument accordingly, and we note that the Q3 range of $500–580M is $65M wide against a $555M base, which is a wider relative band than the company set in May.

Analyst Q&A Highlights

With no prepared remarks, the Q&A was the entire call. Six questioners participated, drawn from independent creators and institutional research, and the session ended when the final participant's audio failed. The exchanges below are the substantive artifacts.

Regulatory contingency planning if CLARITY fails

The opening question laid out the legislative state of play precisely, including third-party odds of passage and the timing risk from the approaching recess, and asked what happens to both the company and the consumer in the failure case. It drew the most useful de-risking answer of the call.

Q: "My question is around CLARITY. As CLARITY is at the 1-yard line and now out for the senate for a vote, Prediction markets and Galaxy Research have odds a bit passing around roughly 30%, and August recess is right around the corner."
— Eric Pan, independent analyst

A: "your question was about what happens if it does not pass, and I think in that world, it is actually kind of just business as usual for Coinbase for a few reasons."
— Brian Armstrong, CEO

Assessment: Management answered the question asked, including the uncomfortable half, and did so without hedging. The answer removes tail risk from the bear case and simultaneously removes a catalyst from the bull case. Legislative outcomes are no longer a swing factor for this equity in either direction.

Joining a competing stablecoin consortium and the Circle renewal

A question about participation in a rival stablecoin platform, and whether it functioned as negotiating leverage, produced an unprompted pre-emption: the CFO interrupted the sequence to close the contract question before addressing the platform question at all.

Q: "Could you please talk about the reason Coinbase joining Onyx USD? Some people think, Onyx USD is a major threat. to USDC."
— Owen Lau, Clear Street

A: "So we have already met the conditions for the circle contract to renew, it will renew on the same terms. I wanna take away any ambiguity about that, for the market."
— Alesia Haas, CFO

Assessment: The eagerness to close this specific question, ahead of the one actually posed, is a tell about how much perceived risk sits on that contract. The clarification is valuable and we take it at face value. What went unexamined is that the contract renewing on the same terms is only good news if the underlying economics are stable, and this quarter demonstrated they are not.

Whether shared USDC economics can be defended as the network scales

The most rigorous question of the call observed that a sufficiently large third-party holder can capture the majority of USDC economics, and asked how the network can be grown without giving away the returns to growing it. Management did not contest the premise.

Q: "The relationship with Hyperliquid seems to demonstrate that if a third party has enough USDC, it can leverage the position into commanding the majority of the USDC economics. How do you continue to invest in the USDC network, bring in new participants, and still protect the longer term economics as the network strengthens."
— Kenneth Worthington, JPMorgan

A: "This is what we think is the right long term strategy for stablecoins, and we are happy to share economics to drive this network effect."
— Alesia Haas, CFO

Assessment: Management conceded the mechanism and defended the choice, which is the intellectually honest response and also an acknowledgement that per-dollar economics are a policy variable rather than a fixed rate. No number was offered for what the marginal shared dollar now earns. Given that stablecoin revenue fell year over year on record balances, that number is the one investors most need and it remains undisclosed.

Senior leadership turnover in a single quarter

A direct challenge on the cluster of executive departures. The response leaned on succession-planning process and named the incoming leaders, and both executives were notably firm in rejecting any strategic read.

Q: "There were a number of departures from your senior leadership team this quarter."
— Kenneth Worthington, JPMorgan

A: "But these are all individual decisions, so there is nothing from a strategy standpoint to read into these changes."
— Alesia Haas, CFO

Assessment: Naming successors and describing them as groomed by the outgoing leaders is stronger than the boilerplate most managements offer. The framing that each departure is individual is also, by construction, unfalsifiable. Clustering in the same quarter as a 14% headcount reduction keeps this on the watch list.

Whether the trust moat survives machine buyers

A question that took management's own prior statements about agentic settlement share and turned them into a challenge: if agents optimize purely on cost and latency, a trust-based moat may not transfer, and the business could be forced into price competition it has so far avoided.

Q: "So as agents become a larger share of volume, does that trust mode transfer to them, or does agentic commerce structurally now push Coinbase toward competing on price? In a way that maybe the consumer business never had to?"
— Austin Hankwitz, GRID Capital

A: "But I also think AI agents are going to choose infrastructure that is reliable and safe and liquid and compliant and has good uptime. Just like they might choose AWS or some kind of cloud vendor for different types of infrastructure, you could imagine."
— Brian Armstrong, CEO

Assessment: The best question of the call and a credible answer. Agents optimize an objective function with more than one term, and compliance and reliability plausibly sit inside it. Management paired this with an explicit refusal to forecast the opportunity, which is the right posture and confirms the option has no near-term cash flows to underwrite.

Is subscription growth accretive or take-rate dilutive

A precise question about unit economics: members receive zero-fee trading, so as volume migrates into the subscription tier, does revenue per dollar traded rise or fall. Management labelled its own answer unsatisfying before giving it.

Q: "But as more volume shifts under this Coinbase 1 umbrella, is revenue per dollar traded higher or lower for a Coinbase 1 member than a nonmember? So, you know, should investors read this growth in Coinbase 1 as accretive or as take rate compression?"
— Austin Hankwitz, GRID Capital

A: "Great question. I am gonna give an unsatisfying answer because when I look at the data on average Coinbase 1 subscribers trade more and have higher unit economics, there is always examples on the edges."
— Alesia Haas, CFO

Assessment: Credit for the candour and for not manufacturing a number. But this is the second consecutive quarter in which the subscription product's unit economics have been described qualitatively and never quantified, and this quarter the subscriber count itself was withheld. For a product management wants valued as an annuity, the disclosure is moving in the wrong direction.

Cross-product adoption and marketing efficiency

A question on how product velocity converts into cross-sell, and how marketing dollars are allocated across entry points, drew the only hard efficiency metric offered on the call.

Q: "I am curious to understand how the team is thinking about driving cross product adoption. And specifically bringing new users in through various entry points that are newer products to the platform."
— Alexander Markgraff, KeyBanc Capital Markets

A: "We are typically seeing a 1-year payback on growth marketing efforts but recently we have outperformed this benchmark."
— Alesia Haas, CFO

Assessment: A one-year payback that is currently running better than benchmark is a genuinely good number and the most concrete unit-economics disclosure of the call. It also sits awkwardly beside a 10% sequential cut to sales and marketing spend. If incremental marketing is paying back inside a year and beating plan, the case for cutting it is a liquidity-and-optics case rather than a returns case, and no one asked management to reconcile the two.

What They're NOT Saying

1. Any acknowledgement of the quarter's financial results. No prepared remarks, no CFO walk-through, and not a single question on the $359.5M net loss, the $113.5M operating loss, or the restructuring charge. An entire earnings call passed without the earnings being discussed.
2. Why subscription and services missed the company's own floor. The deck shows the guide and the actual side by side and leaves the one unchecked line uncommented. Management has never missed this guide in our coverage and offered no explanation of what it got wrong eleven weeks ago.
3. A Coinbase One subscriber count. Q1 disclosed a specific figure above one million. Q2 disclosed only "all-time high." Replacing a number with a superlative on the flagship subscription product, in the quarter its economics were directly questioned, is a disclosure regression.
4. An updated count of products above $100M annualized revenue. Management disclosed twelve at Q1 and said prediction markets was on track to be the thirteenth. Prediction markets duly crossed. The running count was not updated, so we cannot tell whether the portfolio grew, held, or lost members as the cycle bit.
5. The marginal economics of a shared USDC dollar. Management defended sharing economics as strategy and declined to quantify what Coinbase now retains per dollar. With stablecoin revenue negative year over year on record balances, this is the single most valuable missing number in the release.
6. Consumer take rate. Still not disclosed, now a full year after we first flagged it. The deck gives consumer revenue and consumer spot volume but the two use different denominators once prediction markets and derivatives are included, which makes the blended rate uncomputable from public disclosure by construction.
7. Any forward revenue or Adjusted EBITDA guide. Coinbase guides expenses precisely and revenue barely. In a stable market that asymmetry is defensible. In the third consecutive quarter of double-digit revenue decline, guiding only the line that is going down on plan while withholding the line that is going down off plan is a choice, and it shifts the forecasting burden entirely onto the buy side.
8. Whether the fifteenth consecutive positive Adjusted EBITDA quarter is at risk. Management restated the all-conditions commitment and published a July run-rate that, on its own guidance midpoints, puts the quarter near $100M. It did not connect those two disclosures, and no one on the call did either.
9. The cause of the custody share decline beyond ETF flows. The ETF explanation is specific and credible, and the statement that ex-ETF native units rose is reassuring. Neither is quantified, so the 80-basis-point decline in share of crypto market cap cannot be decomposed by an outside reader.

Market Reaction

• Pre-print setup: COIN closed at $163.58 on July 30 ahead of the after-market release. The stock entered the print down 27.7% year to date against the S&P 500 up 8.7%, down 56.7% over trailing twelve months, but up 11.9% over the trailing thirty days. The 52-week closing range was $141.09 to $387.27, so the shares came in near the bottom of a very wide band after a modest month-long recovery.
• After-hours move: Shares fell roughly 5% to 6.5% in the after-hours session immediately following the release, trading near $153.
• Reaction session (July 31): Opened at $153.10, a 6.4% gap down, traded a range of $139.11 to $153.68, and closed at $146.26, down 10.6% or $17.32. The intraday low of $139.11 undercut the prior 52-week closing low of $141.09. Volume of 20.9M shares was 2.8x the 30-day average. The S&P 500 was up 0.7% on the same session.
• Follow-through: Closed at $146.50 on August 3 and $150.73 on August 4, recovering $4.47 or 3.1% from the reaction close without retracing the gap. Net move from the pre-print close through August 4 is -7.9%.

The shape of the reaction is more informative than its size. A 6.4% opening gap that extended to a 15.0% intraday drawdown before closing at -10.6% describes a market that sold the headline, kept selling into the release detail, and then found a bid below the prior 52-week low. That pattern is characteristic of positioning capitulation rather than a re-rating: the marginal seller was flushed and the two subsequent sessions recovered a third of the loss on declining volume.

What the market appears to have priced is the third consecutive miss and the Q3 setup, not the underlying franchise. The stock entered the quarter having already de-rated 56.7% over twelve months, so this was not a fresh repricing of the crypto cycle. The specific catalyst was the combination of a revenue line missing by roughly 6% and a July run-rate disclosure that made a fourth consecutive decline arithmetically visible in the release itself. Coinbase is one of very few companies that hands the market a partial quarter's revenue on the call, and in a deteriorating tape that transparency works against the stock.

The subsequent two-session recovery, on volume falling from 20.9M to 7.2M, suggests the sell-off exhausted rather than reversed. We would not read the bounce as a signal about the cycle.

Street Perspective

Debate: Is the cost program sufficient, or is it chasing a falling revenue line?

Bull view: The bull case being made on the Street is that management has done the hard thing early. Cutting headcount 14% and reducing a full-year expense guide mid-cycle, rather than defending headcount into a recovery, means the cost base will be roughly $600M below the 2025 exit run-rate when volumes turn. Because the cuts are structural and AI-assisted rather than discretionary, the argument goes, the operating leverage into any cycle recovery is now unusually steep and largely unmodelled.

Bear view: The bear camp contends that a 1% year-over-year increase in the expense midpoint is not a cost program at all when revenue is tracking down roughly 19%, and that management has cut the growth investment while leaving the fixed base intact. On this reading, cutting sales and marketing 10% in a quarter when management says growth marketing is paying back inside a year is a confession about near-term cash generation, not a demonstration of discipline.

Our take: The bear has the better of the near term and the bull has the better of the medium term, and the resolution is a question of duration rather than of who is right. Both statements are true simultaneously: the cuts are real and structural, and they are insufficient to hold the operating line at current revenue. If the cycle turns within two or three quarters, the bull case pays. If it does not, another cost action is required, and the second one comes out of the growth engine.

Debate: Does record market share matter if revenue keeps falling?

Bull view: A growing consensus view among the constructive camp is that share is the only variable management controls and it is going the right way decisively: an all-time-high 10.3% total trading share, an all-time-high derivatives share for a third consecutive quarter, flat derivatives volume against a market down double digits, and prediction markets doubling. In a cyclical business, share gains compound into the recovery and are worth more than any single quarter's revenue.

Bear view: Some sell-side desks argue that the share gains are concentrated in structurally lower-take-rate products, so each point of share is worth less revenue than the last. On this view the blended take rate is being compressed by the very mix shift that generates the share statistics, and the company is winning volume it cannot monetize at historical rates.

Our take: The bear identifies a real mechanism and draws the wrong conclusion from it. Derivatives are lower-take by construction, and growing them does compress blended economics, but the alternative is not a higher-take mix, it is a smaller business. What we would push back on is the bull framing that share gains are automatically bankable in a recovery. They are bankable only if the products carrying the share have durable unit economics, and management has now declined to quantify unit economics for the subscription tier, prediction markets, and shared USDC in consecutive quarters. Share is being demonstrated; the economics of the share are being asserted.

Debate: Is subscription and services still the defensive floor?

Bull view: The constructive case holds that S&S at 48% of net revenue is doing exactly what it was built to do: it fell 4.9% in a quarter when transaction revenue fell 20.7%, which is a four-to-one dampening ratio and precisely the buffer the diversification strategy promised.

Bear view: The skeptical case is that the buffer is thinner than the aggregate implies. Blockchain rewards, at 15% of the S&S line, fell 42.8% year over year and are levered to token prices twice over. Stablecoin revenue, at 53% of the line, is now shrinking year over year despite record balances. Strip those and what remains is roughly a third of the line.

Our take: The bear is right on the composition and this is the most important thesis revision of the quarter. The dampening ratio is real, but a buffer whose two largest components are levered to token prices and to the front end of the curve is a lower-beta revenue stream, not an uncorrelated one. We are downgrading our own characterization of this pillar accordingly. It still deserves a premium to transaction revenue; it no longer deserves to be modelled as an annuity.

Model Update Needed

Our standing model was built on the Q1 2026 read and carried a recovery-shaped volume path from Q2 2026 forward. That path is now falsified for at least two more quarters, and several S&S driver assumptions were too high.

Valuation impact. At the August 4 close of $150.73 and 264M ending shares, market capitalization is approximately $39.8B. Against trailing-twelve-month Adjusted EBITDA of $1,877.7M that is roughly 21x. Against the Q2 run-rate annualized ($831M) it is roughly 48x. Against a recovery to the Q3'25 run-rate annualized ($3,203M) it would be roughly 12x. Those three numbers, spanning 12x to 48x on the same share price, are the entire argument: this is a cyclical trading at a trough multiple of trough earnings, and the valuation you compute depends almost entirely on which quarter you annualize. We regard normalized mid-cycle Adjusted EBITDA of $2.0–2.5B as the right anchor, which puts the shares at 16x to 20x mid-cycle. That is neither cheap enough to underwrite an Outperform on valuation alone nor expensive enough to justify an Underperform against a franchise gaining share.

Thesis Scorecard Post-Earnings

Scored against the standing thesis carried since Q2 2025 and last updated at Q1 2026. Pillar numbering is unchanged so the arc remains comparable quarter to quarter.

Overall: The thesis is weakened, and for the first time in our coverage the weakening is not purely cyclical. Four of eight bull pillars remain confirmed or strengthened, two are neutral, and two are challenged, with the S&S buffer pillar moving from ON TRACK to AT RISK on a miss against the company's own floor and the stablecoin pillar moving to AT RISK on year-over-year revenue decline against record balances. Two bear points escalated from EMERGING to MATERIALIZING, and two new bear points are added: narrowing disclosure of unit economics, and an operating loss that no longer has a mark to blame. Against that, the operating-discipline pillar strengthened materially and the share-gain pillar is the strongest it has been. The franchise is not the problem. The revenue base is, and the cost program is not closing the gap fast enough.

Action: Hold. Do not add on the sell-off and do not exit on the miss. Specific triggers, in both directions, for the Q3 print: upgrade on transaction revenue exceeding roughly $520M against the $460M July run-rate, S&S landing at or above the $580M guide high, and Adjusted EBITDA above $150M; downgrade to Underperform on any of a negative Adjusted EBITDA quarter, a fourth consecutive S&S guide miss, or a second in-year cut to the expense framework that reaches the growth investment. The single most informative datapoint between now and then is not a company disclosure at all; it is spot volume and the front end of the curve.