Two crypto exchanges disappeared this week for reasons that may seem similar, but are not. BitMEX, the venue that pioneered the perpetual swap, was outcompeted by on-chain perpetual platforms that deliver the same product faster and at lower cost. Bitmart, a retail spot venue, simply ran out of retail spot business. The rest of the sector tells the same story. Gemini cut roughly a third of its staff after a $582 million loss, and Coinbase shed 14% of its workforce as consumer transaction revenue fell 45% year over year. Monthly spot volume across centralized venues hit a two-year low near $1.05 trillion, while Korean exchange volumes fell almost 90% with help from local tax policy. 

For all of it, the industry has one explanation: winter. Bitcoin fell from $115,000. Volumes always collapse when prices do, and exchanges always cut at the bottom. Anyone claiming the cycle is dead has history against them. Still, the cycle only accounts for the timing of this contraction. Its shape has a separate cause: centralized exchanges are being squeezed from both ends of their business at once, and the middle that remains was never the most profitable part.

The Business Was the Middle

Strip an exchange to its economic engine, and it was always one thing: retail spot brokerage. 

A price-taking customer crosses the spread and pays anywhere from dozens of basis points to more than a percent on flow that costs almost nothing to serve. Everything else, from institutional volume at single-digit basis points to market-maker rebates and custody, was volume without much margin. 

The engine worked as long as the two highest-value flows in crypto both had to pass through it: blue-chip accumulation on one end, speculative frenzy on the other.

Between 2024 and 2026, each end found a better route.

Front One: the ETFs Took the Top

Spot ETFs raised demand for BTC and ETH while also rerouting it. The marginal blue-chip buyer became an Authorized Participant (AP) or an OTC desk, transacting at wholesale prices that a retail taker never sees. Much of that flow never reaches an exchange at all. Desks internalize what they can against institutional sellers and push only residual risk onto public order books at institutional pricing. In-kind creations and redemptions, approved in late 2025, shrank even that residual leg: APs can now settle creations by delivering coin instead of buying it, and large holders move existing BTC into the wrapper through participating dealers. 

The more permanent half of this front is that retail followed the product. The customer who once paid up to 1% to buy bitcoin on an exchange now buys an ETF at zero commission in the brokerage account where their index funds live (often inside an IRA). Migrations into cheaper, more convenient, tax-advantaged wrappers are one-way doors. That customer is unlikely to return when sentiment improves.

Front Two: the Chains Took the Bottom

At the other end, the speculative flow that once defined exchange P&Ls moved on-chain. Memecoin-era trading lived natively on DEXs and launchpads; centralized venues captured only the richest slice (graduated listings, leveraged perpetuals, and a long tail of tokens at full retail take rates and wide spreads). That slice was extraordinarily profitable per dollar of volume, a multiple of what bitcoin flow earns, and for six quarters it flattered aggregate revenue enough to keep multiples high and questions unasked. Nothing was hidden: the institutional mix shift and grinding take rates sat in every quarterly disclosure. Nothing needed hiding when the top line was growing. It was simply ignored.

Then the froth ended, as froth does. The perp-DEX generation began taking derivatives flow too, which is how a pioneer like BitMEX can die in the same week as a spot venue: two fronts, one squeeze.

Reading the Cross-Section

Is this just winter wearing a thesis? A single downturn cannot settle that, but the cross-section can. The mix shift toward low-margin flow was visible while prices were still rising, which cycles do not explain. The carnage is deepest precisely where the squeeze predicts: pure retail spot venues, retail-heavy geographies, and exchanges with no derivatives or services franchise. 

The resilient names are the ones that exited the middle early: the large exchange that captured the ETF complex through custody, built stablecoin and subscription income, and bought its way into derivatives; the diversified brokerage whose stock fell a fraction of its pure-crypto peers. The survivors are not counterevidence. They came out on top because they repriced their business model before the market repriced it for them.

The scope of this claim is limited to spot retail brokerage economics rather than crypto trading as a whole. Perpetual futures remain the industry's largest fee pool, and ETFs do not touch them. But that pool is now contested by on-chain venues rather than protected by incumbency. That is a competition problem layered on top of a cannibalization problem.

Floor and Ceiling

The precise claim is simple: exchange revenue used to be a leveraged bet on retail spot activity, and that leverage has been permanently reduced. Cycles will continue, but their amplitude, measured in exchange fees, will not return to the old line. The floor of this downturn and the ceiling of the next recovery have both shifted lower.

That yields a testable prediction: in the next bull market, watch revenue per dollar of volume rather than revenue. Prices and volumes will recover; spot transaction revenue at centralized venues will lag severely (or poorly) because the returning flow will route through ETFs, OTC internalization, in-kind plumbing, and on-chain venues rather than retail order books. If blended take rates recover with the cycle, this thesis is wrong.

And there is a second act. The exchange businesses that endure are already becoming infrastructure: custodians to the ETF complex, stablecoin and payments platforms, Onchain Earn programs, prediction markets, derivatives operators, and listing venues for tokenized equities and RWAs. In short: businesses that may eventually out-earn the retail spot engine they replace. 

The firms writing shutdown notices this month are not casualties of a bear market. They were toll booths in the middle of a road that users learned to avoid.