
Revisiting This CEX Summer Through Coinbase's Earnings Report
Written by: Fugui
After the US stock market closed on July 30, Coinbase released its report card for the second quarter of 2026. During regular trading hours that day, the stock closed at $163.55, up 2.16%, and market sentiment was originally passable. Once the earnings report came out, the picture changed dramatically, falling all the way to around $153 in after-hours trading, a drop of over 6%. Loss per share was $1.36, while analysts originally expected nearly break-even at -$0.01; the gap was so large it didn't look like a mistake within the "in-line with expectations" range. Total revenue was $1.22 billion, falling short again of Wall Street's expectation range of $1.29 billion to $1.35 billion. This marks the third consecutive quarter where Coinbase's revenue has missed expectations.
The macro background is not complex. Global crypto spot trading volume fell 25% quarter-over-quarter in Q2, and the total crypto market cap shrank by 11%. Volatility for BTC and ETH both dropped to multi-year lows. With the market quiet and uneventful, users were too lazy to move their hands. For an exchange that lives off fees, encountering a quarter without stories, an ugly balance sheet was expected. What truly surprised the market was the magnitude of the loss, exceeding prior psychological expectations.
How Coinbase Makes Money
Trading fee revenue was $599 million, down 21% quarter-over-quarter. Of this, retail spot contributed $452 million, down 20% quarter-over-quarter and down 30.5% year-over-year. Monthly active trading users dropped from 8.2 million in Q1 to 7.6 million. Bitcoin spot ETFs shifted from net inflows to net outflows, with both institutions and retail retreating. Institutional trading revenue actually rose 64.6% year-over-year, thanks to the consolidation of Deribit last year, but quarter-over-quarter it also fell 26%. The pattern in this sector is simple: if coin prices don't move, fees have nowhere to land. Like selling umbrellas, without extreme weather, days are hard.
Interestingly, subscription and services revenue was $555 million, accounting for 48% of total revenue, a historic high. It fell only 5% quarter-over-quarter and 12% year-over-year, a much more decent decline than the trading business, becoming the only reliable stabilizer in this earnings report. However, breaking down the quality of this subscription revenue reveals a worthwhile discrepancy: true subscription products like Coinbase One, membership fees combined account for only a fraction of this revenue; instead, stablecoin-related revenue, specifically USDC reserve interest sharing, accounts for more than half, $292 million, accounting for 53% alone. Staking rewards were $83 million, accounting for 15%; interest and financing income was $66 million, accounting for 12%; the remaining scattered projects like prediction markets, institutional custody, Base ecosystem, and debit card fees made up $114 million, accounting for 20%.
To translate: the subscription revenue Coinbase talks about is actually mostly the Federal Reserve working for it. Users leave USDC on the platform, the platform invests this money into US Treasury bonds, earns the interest spread, and then takes nearly half the proceeds from Circle—this business has nothing to do with the word "subscription" except being written on the same report. The true membership subscription product is actually the most insignificant supporting role in this revenue. This isn't me being sarcastic; the financial statements are slapping themselves in the face. Actually, this strategy isn't anything new; unlisted small and medium-sized exchanges know this trade inside out. Stablecoins users leave untouched can generate interest on anyone's books, it's just no one publicly announces where this money ultimately ends up. Coinbase's uniqueness lies only in being a public company; the SEC requires it to lay out the source of every penny for shareholders to see. Stablecoin interest accounting for over half of services revenue, this business known tacitly within the industry, forced into the 10-Q, suddenly looks like news. Going forward, CEXs won't be competing on who has lower fees, but who has fatter stablecoin reserves and deeper institutional custody. Exchanges are quietly becoming a service business, it's just that many people have been doing this for a long time.
How Coinbase Spends Money
Looking at expenses again. Technology and R&D expenses were $472.8 million, the only item among the three major categories that rose year-over-year instead of falling, up 22% YoY. Although it fell 10% quarter-over-quarter, what fell was headcount, not projects. Base blockchain, prediction markets, derivatives, and the newly launched Coinbase for Agents (opening trading interfaces to AI agents, agents like Claude, ChatGPT can place orders and transfer funds within the scope of user authorization), none stopped. Administrative expenses were $356.9 million, down 5% quarter-over-quarter, basically flat year-over-year. This part consists of fixed expenses for compliance, legal, and regulatory litigation. SEC lawsuits, MiCA implementation costs, are all hard expenses, cannot be cut. What was cut cleanly was Sales and Marketing, $239.8 million, down 10% quarter-over-quarter, the trend is exactly opposite to R&D, one adding positions against the trend, the other contracting with the trend.
Translating this strategy: save everywhere possible, do not save a penny where it cannot be saved. Marketing expenses fell most happily, indicating Coinbase no longer needs to burn money for users—market share has risen to a historic high of 10.3%, scale itself is the cheapest customer acquisition channel. The money saved, along with cash earned from trading business, was all fed into R&D, betting that when the next bull market arrives, derivatives, prediction markets, and the Base ecosystem can immediately realize revenue, rather than starting from zero. Layoffs dropped from 4,988 to 4,321, a 14% decrease. One-time severance pay of $52.4 million was spent in Q2, in exchange for a permanent decrease in fixed costs every quarter going forward.
Two Sets of Books
On the books, Coinbase net loss in Q2 was $359.5 million. Let's put two terms on the table first: GAAP (Generally Accepted Accounting Principles), requires enterprises to revalue self-held crypto assets at market price at the end of each quarter, if it falls, losses must be accrued, even if not a single coin was sold; another term is Adjusted EBITDA, many simplify it to EBIT for convenience, operating profit remaining after excluding interest, taxes, depreciation, amortization, and such non-cash floating losses. Simply put, it is whether the business itself makes money.
According to GAAP standards, total revenue $1.22 billion, minus total operating expenses $1.336 billion (this expenditure already includes $52.4 million one-time restructuring costs), operating level first loses $113.5 million; further down, self-held investment-class crypto assets floating loss $209.5 million, operation-related crypto assets floating loss $31.7 million, plus interest expenses and other financial items included in non-operating losses, several items superimposed, net loss rolls to $359.5 million.
According to Adjusted EBITDA standards, total revenue $1.22 billion minus adjusted operating expenses $1.03 billion, is positive $208 million, positive for the fourteenth consecutive quarter, free cash flow net inflow $197.3 million.
The two sets of books calculate two things: one calculates whether doing business this quarter makes money, the answer is yes; the other calculates if holding Bitcoin and Ethereum in hand were liquidated at today's prices, how much is left on the books, the answer is nothing left.
The market ultimately chose to believe EBITDA. Stock price fell immediately upon earnings release that night, the reason is not complex: revenue missed expectations for three consecutive quarters, plus Q3 guidance is conservative, investors care about whether the trend has bottomed out, not whether this quarter's accounting technical loss counts. As of now, company market cap is about $43.18 billion, P/E ratio slightly over 57 times, considered by some institutions as overvalued—A company losing money on the books, but stock price priced as if making money, this awkwardness, might be the true root of stock price volatility.
Next Door Has Already Closed
Zooming out a bit, Coinbase's trouble is actually a sweet trouble. In the same July, two old-school exchanges announced closure one after another. On July 23, BitMEX, the derivative ancestor that invented 100x leverage perpetual contracts and operated for 11 years, announced it would officially stop operations on September 23, stopping new user registration immediately. Official statement says it's an active decision after strategic assessment, zero hacker theft record in history, the three founders including Arthur Hayes were pardoned by Trump as early as March 2025. Three days later, July 26, BitMart also announced orderly cessation of operations, fully stopping trading on August 26, completely closing station on January 31, 2027.
BitMart closing up shop this time reveals an unexplainable rush. Two months ago, the platform just publicly denied rumors of a withdrawal crisis. Once the announcement came out, on-chain data spoke: some analysis accounts statistics, within 24 hours after announcement, only 58 wallets on the whole platform successfully withdrew about $805,000, in the recent 8-hour tracking window, withdrawal records were zero. Some users received emails saying withdrawal completed, but account page showed on-chain withdrawal frozen; someone tested withdrawing $30, waited half an hour with no follow-up. Platform token BMX fell 81.5% in a week. Four months ago, Polish exchange Zondacrypto's hot wallet Bitcoin balance fell from 55.7 coins to 0.086 coins, evaporating 99.7%, during which the platform always claimed funds were sufficient, about 30,000 users affected, called the largest exchange explosion in Europe after FTX.
These cases confirm, for small and medium CEXs, the theory of three big mountains: regulatory compliance, marketing customer acquisition, network security, one mountain harder to climb than the next.
Regulatory compliance is a visible and tangible one. Fees can be quoted, time can be scheduled, it's just the price is ridiculously high. Coinbase alone, completing the full process of US state licenses, took a full five years, legal, KYC system, compliance team, every item requires real money thrown in. Small and medium platforms cannot afford this time, nor can they afford this budget, so many simply go around it, registering entities in Seychelles, Cayman Islands such regulatory lowlands, if can get license don't get license. BitMEX is typical, headquarters long set in Seychelles, until 2022 only because anti-money laundering system non-compliant pleaded guilty, three founders once faced criminal liability, relying on 2025 pardon to dodge a bullet. Going around this mountain, seems like saved money, bill ultimately must be calculated, just changed time and way to collect debt.
The second mountain is the true core, and harder to climb. Users have already been educated by head platforms, why switch to open an account in a strange place. Small and medium exchanges want to grab people, either rely on high yield inducement, sliding step by step towards Ponzi scheme; or rely on massive ad placement, burning cash through, users exchanged not necessarily stay. Coinbase this quarter happens to stand on the other side of this question: marketing expenses quarter-over-quarter cut by 10%, not lack of money, is no need to spend money, 10.3% market share, years accumulated brand trust, itself is cheapest customer acquisition channel. Same question, answer divides into two directions: one side relies on scale advantage to save money, one side relies on burning money to survive, this question once answered wrong, money on books hits bottom much faster than imagined.
The third mountain is most fatal, because it is completely unpredictable. Ordinary penetration testing can prevent, true trouble is after small and medium CEXs develop to certain scale attracted super DDoS attacks, that kind of out of competitiveness lose money also want to fight, target is let your service interrupt tactic. BitMart actually not considered lightly, found security company Hacken with background from Ukraine Ministry of Defense to do penetration testing, exit equipped with traffic pool prevent DDoS, book homework done not bad. These things behind are all high costs.
What crushed them, never is some specific attack, but when three big mountains press down simultaneously, cash flow first ran out of breath, user confidence followed together broke breath.
The Only Road Left
Ultimately, CEX always carries the shade of shadow banking: users store coins in, exchange turns hand takes to generate interest, lend, market make, books write is assets, doing is maturity mismatch business. Bank run relies on confidence, exchange run relies also on confidence. BitMart two months ago still refuting withdrawal crisis, this time even rumor saved, official announcement closing day, withdrawal itself froze. Truly holding up an exchange, never is proof of reserves on that string of numbers, but is users believe not believe this string of numbers. Coinbase can withstand book losses, relies on credit still there; BitMEX and BitMart cannot withstand, ultimately not lost to which shock, is credit first gone.
Looking forward, this industry probability will go towards two poles. Capable, have cash reserve head platforms, will continue compress marketing and administrative such elastic expenses, put saved money pile into R&D, force self grow subscription, custody, stablecoin yield such decoupled from coin price revenue curve, Coinbase itself given target, is put subscription and services revenue ratio pull to sixty percent above. This road essence, is make exchange into financial infrastructure, use differentiated service swap survival space. And those neither can carry compliance cost, nor pay customer acquisition bill, security budget also long-term tight small and medium platforms, probability will successively walk onto BitMEX and BitMart old road, just decency degree each different, some can choose self close door, some even decency these two characters no time to consider.
Losses in financial statements, ultimately still have next quarter turn around opportunity; closing outside financial statements, even modification room none. This probably is 2026 this summer, CEX industry most plain one dividing line. All financial infrastructure, finally ask are same sentence: account inside that string balance, is your money, or is it a wish promised to you by others.















