The tollbooth survives. The operator does not.

Solana DEX volume is down 82 percent from the January 2025 peak. Pump.fun’s fees are down 46 percent over the same window.

Photon runs a version of the same business. It is down 99.4 percent.

I pulled the daily fee data for every major venue on Solana and aggregated it by month. The spread between those three numbers is the whole story of this layer.

Photon was the king of the last cycle. It charged 1 percent on every buy and every sell. No subscription, no volume discount, no token. In January 2025 that model collected $84.6 million in a month. In the 24 hours to July 25, 2026, it collected $14,311.

Trojan is down 97 percent. BullX went from $54.6 million in a single month to nothing.

None of them died of the bear market. They died of each other. Every one charged the same 1 percent, executed fast, and owned nothing underneath. Switching cost on a trading terminal is the time it takes to import a private key.

Now take Pump.fun’s monthly fees and divide by total Solana DEX volume. That is what the platform pulls out of every dollar moving on the chain.

January 2025, at the top: 4.8 basis points. July 2026, at the bottom: 14.6.

Volume collapsed by 82 percent and the take rate tripled. It did not raise the headline price. It added floors to the building. PumpSwap, its own AMM, is now $50.6 million of the $79.1 million it earned last month. The launchpad that made the company famous is the smaller half. Then it shipped Terminal, its own front end, aimed at the terminals above. Terminal did $2.96 million in 30 days. Photon and Trojan together did $1.6 million.

Fees in this business come from turnover, not from price. Panic selling at the lows pays the same rake as chasing at the highs. Run the correlation across 26 months and Pump.fun scores 0.23 against the SOL price and 0.77 against volume. Photon scores 0.65 against price, which is what a beta wrapper looks like from the inside.

Then there is Fomo, a phone app rather than a terminal, charging half what the terminals charge, selling a feed instead of execution speed. It scores minus 0.68 against price. Its best month ever is this one, with SOL at $77.

So the layer does survive the cycle. Near zero marginal cost, revenue tied to churn instead of price, and a supply of things to gamble on that regenerates for free.

But zero marginal cost and zero switching cost are the same sentence read from opposite ends. Nothing cheap to run is expensive to replace. The crown moved three times in 24 months.

I have traded on all four of these. Nothing in the product told me which one would still be here.

The trench always pays rent. The landlord keeps changing.