Written by: Fortune

After the close of the U.S. stock market on July 30, Coinbase released its second-quarter report for 2026. The stock closed at $163.55 during regular trading hours, up 2.16%, and market sentiment was relatively okay at that time. However, once the earnings report was released, the sentiment changed drastically, and the stock fell to around $153 in after-hours trading, a drop of over 6%. The loss per share was $1.36, while analysts had originally expected a near break-even loss of $0.01, a gap too wide to be considered a simple "miss within expectations." Total revenue was $1.22 billion, falling short of Wall Street's expectation range of $1.29 billion to $1.35 billion, marking Coinbase's third consecutive quarter of disappointing revenue.

The macro background is not complex. In the second quarter, global cryptocurrency spot trading volume decreased by 25% quarter-on-quarter, and the total market capitalization of cryptocurrencies shrank by 11%, with the volatility of BTC and ETH dropping to multi-year lows. With subdued market activity, users were less inclined to trade, which was to be expected for an exchange that relies on transaction fees facing a quarter with no compelling narrative. What truly surprised the market was the extent of the loss, which exceeded prior psychological expectations.

How Coinbase Makes Money

Transaction fee income was $599 million, down 21% quarter-on-quarter. Retail spot trading contributed $452 million, down 20% from the previous quarter and down 30.5% year-on-year. Monthly active trading users dropped from 8.2 million in the first quarter to 7.6 million, and net inflows into Bitcoin spot ETFs turned into net outflows, with institutions and retail investors withdrawing together. Institutional trading income rose by 64.6% year-on-year, thanks to Deribit’s consolidation last year, but also fell by 26% quarter-on-quarter. The rule for this segment is quite simple: When cryptocurrency prices do not move, there are no transaction fees to be earned, similar to selling umbrellas; without extreme weather, business becomes difficult.

Interestingly, subscription and service revenue amounted to $555 million, accounting for 48% of total revenue, a historical high, with a quarter-on-quarter decline of only 5% and a year-on-year decrease of 12%, a much more respectable decline compared to transaction services, making it the only stable point in this earnings report. However, when dissecting this subscription revenue, a noteworthy misalignment emerges: the true subscription product, Coinbase One, contributes only a small fraction to this income; instead, revenue related to stablecoins, specifically the interest share from USDC reserves, constitutes over half at $292 million, accounting for 53%. Staking rewards amounted to $83 million, making up 15%; interest and financing income was $66 million, comprising 12%; the remaining fragmented projects including prediction markets, institutional custodial services, Base ecosystem, and debit card transaction fees brought in $114 million, representing 20%.

To put it differently: Coinbase's claimed subscription revenue mainly comes from the Federal Reserve doing the work for it. Users deposit USDC on the platform, and the platform invests this money in U.S. Treasury bonds, earning the interest spread and then taking nearly half of the profits from Circle — this business has nothing to do with the word "subscription," aside from being reported on the same financial statement. The actual subscription product is, ironically, the least noticeable supporting role in this income. I am not being sarcastic; the financial statements are just revealing their own contradictions. This method isn't new at all; many unlisted small exchanges are well aware of it. Users' stagnant stablecoins can generate interest on anyone's books, but no one publicly discloses whose pockets the money eventually ends up in. Coinbase's uniqueness lies in being a publicly traded company; the SEC requires it to disclose the sources of every penny to shareholders, and stablecoin interest making up over half of service revenue — this business, long tacitly accepted in the industry, being forced into the 10-Q report appears newsworthy. Moving forward, centralized exchanges will compete not on lower fees, but on healthier stablecoin reserves and deeper institutional custody, subtly transforming exchanges into service businesses — a business many are already engaged in.

How Coinbase Spends Money

Now let’s look at expenditures. Technology and research expenses were $472.8 million, the only one of the three major categories that increased year-on-year by 22%; although it decreased by 10% quarter-on-quarter, this was due to staff reduction, not project scaling back. The Base public chain, prediction markets, derivatives, and the newly launched Coinbase for Agents (providing an open trading interface for AI agents like Claude and ChatGPT to place orders within user-authorized scopes) all continued without pause. Administrative and management expenses were $356.9 million, down 5% quarter-on-quarter and roughly flat year-on-year; this portion includes compliance, legal, and regulatory legal costs, with SEC lawsuits and MiCA implementation costs representing hard expenditures that cannot be cut. In contrast, sales and marketing expenses were a sharp $239.8 million, down 10% quarter-on-quarter, showing a reverse trend with R&D: one expanding against the tide, the other contracting.

This strategy translates to: save wherever possible and not a penny less where it can't be saved. A significant reduction in marketing expenses indicates that Coinbase no longer needs to spend heavily to acquire users — its market share has already reached a historical high of 10.3%, and scale itself is the cheapest customer acquisition channel. The saved funds, along with the cash earned from transaction business, were all funneled into R&D, betting that when the next bull market arrives, derivatives, prediction markets, and the Base ecosystem can immediately convert revenues instead of starting from scratch. The workforce was reduced from 4,988 to 4,321, a 14% decrease, and a one-time severance payment of $52.4 million was recorded in the second quarter, leading to a permanent decline in fixed costs in every following quarter.

Two Sets of Accounts

On paper, Coinbase reported a net loss of $359.5 million for the second quarter. Here, we need to explain two terms: GAAP (Generally Accepted Accounting Principles) requires a company to reevaluate its cryptocurrency assets at market value at the end of each quarter. If prices fall, losses must be recognized, even if no coins have been sold; the other term is Adjusted EBITDA, often abbreviated as EBIT, which calculates operating profit by excluding interest, taxes, depreciation, amortization, and similar non-cash losses, simply put, it shows whether the core business is profitable.

According to GAAP, total revenue was $1.22 billion, and when total operating expenses of $1.3336 billion (which includes the one-time restructuring expense of $52.4 million) are subtracted, an operating loss of $113.5 million is recorded. Further down, the floating loss from self-held investment cryptocurrencies amounted to $209.5 million, and the operating-related floating loss was $31.7 million, plus interest and other financial items accounted for in non-operating losses: these combined resulted in a net loss of $359.5 million.

Using adjusted EBITDA, total revenue of $1.22 billion minus adjusted operating expenses of $1.03 billion gives a positive figure of $208 million, marking the fourteenth consecutive quarter of positive adjusted EBITDA, with a net inflow of free cash flow of $197.3 million.

The two sets of accounts represent different aspects: one assesses if doing business in this quarter is profitable, the answer is yes; the other estimates how much would remain if all the Bitcoin and Ethereum were liquidated at today’s prices, the answer is zero.

The market ultimately chose to trust EBITDA. The stock price fell on the night of the earnings report, and the reasons are not complicated: three consecutive quarters of revenue missing expectations, alongside a conservative guidance for the third quarter; investors are concerned with whether the trend has hit the bottom, rather than whether this quarter's accounting technical loss is significant. As of now, the company’s market capitalization is around $43.18 billion, with a price-to-earnings ratio just above 57, considered by some institutions to be overvalued — a company that is financially losing money is priced like one that is earning, this inconsistency may truly be the root cause of stock price volatility.

The Neighbor Has Already Closed Its Doors

Pulling the perspective back a bit, Coinbase's troubles are, in fact, sweet troubles. In the same July, two veteran exchanges announced their closures successively. On July 23, BitMEX, the originator of the 100 times leveraged perpetual contracts and operating for 11 years, announced it would officially stop operations on September 23 and cease new user registrations immediately. The official statement indicated this was an active decision after strategic evaluation, noting a history free of hacks. Founders Arthur Hayes and the others were pardoned by Trump back in March 2025. Three days later, on July 26, BitMart also announced orderly cessation of operations, stopping all trading on August 26 and completely closing down by January 31, 2027.

BitMart’s closure exudes an inexplicable hurry. Two months prior, the platform had publicly denied rumors of a withdrawal crisis. Following the announcement, on-chain data spoke volumes: an analysis of accounts indicated that within 24 hours of the announcement, only 58 wallets successfully withdrew approximately $805,000 across the entire platform, and in the most recent eight-hour tracking window, there were zero withdrawal records. Some users received emails stating their withdrawals were complete, yet their account pages showed on-chain withdrawals were frozen; others tested withdrawing $30 and waited half an hour without any updates. The platform token BMX plummeted by 81.5% within a week. Four months prior, the Polish exchange Zondacrypto saw its hot wallet Bitcoin balance drop from 55.7 coins to 0.086 coins, evaporating by 99.7%, while the platform vehemently claimed funds were sufficient, affecting around thirty thousand users, dubbed the largest exchange failure in Europe post-FTX.

These cases illustrate the three formidable mountains facing smaller centralized exchanges: regulatory compliance, customer acquisition marketing, and cybersecurity — each harder to overcome than the last.

Regulatory compliance is a tangible barrier. Fees can be quoted, and timelines can be scheduled, but the costs are prohibitively high. Coinbase, as a single company, went through the full licensing process across all 50 states, taking a total of five years, needing to invest real money into legal, KYC systems, and compliance teams for each step. Smaller platforms cannot afford this time, let alone this budget, so many choose to circumvent the process altogether, registering the entities in regulatory havens like Seychelles or the Cayman Islands, avoiding obtaining licenses where possible. BitMEX is a prime example, operating its headquarters in Seychelles for years, until 2022, when it pleaded guilty to non-compliance with anti-money laundering regulations, with its three founders facing potential criminal charges, only managing to escape through a pardon in 2025. Skirting around this mountain seemingly saves money, but the bill is still due; it just changes the time and manner of collection.

The second mountain is the real core challenge and is even more difficult to navigate. Users have already been educated by leading platforms; why would they choose to open an account in a strange place? For smaller exchanges looking to capture users, they either rely on high returns to entice users, gradually sliding towards a Ponzi scheme; or they burn cash through massive advertising, acquiring users who might not even stay. Coinbase, in this quarter, stood on the opposite side of this issue: reducing marketing expenses by 10% quarter-on-quarter was not due to a lack of funds, but because it didn’t need to spend money; with a market share of 10.3% and years of accumulated brand trust serving as the cheapest customer acquisition channel. The same problem presents two distinct answers: one side saves money through scale advantages, while the other requires burning cash to survive. If this question is answered incorrectly, the available cash can run out faster than anticipated.

The third mountain poses the greatest peril, as it is entirely unpredictable. Standard penetration testing can be prevented, but the real trouble arises from the super DDoS attacks that hit smaller CEXs once they reach a certain scale, attacks aimed solely at disrupting service, even at a loss for competitive reasons. BitMart did not take this lightly; it hired Hacken, a security company with ties to the Ukrainian Ministry of Defense to perform penetration testing, and set up a traffic pool for DDoS defense; on paper, their preparedness was promising. All of these measures come with substantial costs.

What crushes them is not any single attack but when all three mountains come crashing down simultaneously; cash flow first chokes, and then user confidence follows suit.

The Remaining Path

Ultimately, centralized exchanges have always carried the undertone of shadow banking: users deposit coins, exchanges lend them out, invest them, and engage in market-making, recording assets on the books while conducting a mismatched terms of business. Bank runs rely on confidence, and exchange runs do too. BitMart two months ago was still denying rumors of a withdrawal crisis; this time, even the rumors were unnecessary — on the day of the official announcement to close, withdrawals froze on their own. What upholds an exchange isn’t the string of numbers on reserve proof, but whether users trust that string of numbers. Coinbase can withstand accounting losses because its credibility still remains; BitMEX and BitMart could not hold on, and ultimately, it was not because they lost to any one shock but because their credibility disappeared first.

Looking ahead, this industry will likely gravitate toward polar extremes. Leading platforms with capabilities and cash reserves will continue to cut back on elastic expenses like marketing and administration, piling up the saved money into R&D, pushing themselves to cultivate a revenue curve from subscription, custody, and stablecoin interests that is detached from currency prices. Coinbase's own target is to raise the ratio of subscription and service revenue to over 60%. The essence of this path is to transform exchanges into financial infrastructure, exchanging differentiated services for survival space. In contrast, those smaller platforms that cannot withstand compliance costs, unable to pay customer acquisition bills, and always strapped for security budgets will likely follow the old path of BitMEX and BitMart, albeit with varying degrees of grace; some may choose to close shop themselves, while others won't even have the luxury of maintaining appearances.

Losses on the financial statements ultimately still hold opportunities for redemption next quarter; closures outside the reports leave no room for revision. This could be the simplest division line in the CEX industry in the summer of 2026. All financial infrastructures ultimately ask the same question: that balance in the account — is it your money, or is it just a promise someone made to you?