Perps didn’t just “rise.” They erased the old fee hierarchy and rebuilt it around trading behaviour instead of capital. Hyperliquid clearing $110M in 30-day fees on $266B volume and Jupiter Perps doing $80.33M on a totally different architecture are pointing to the same structural shift: the systems tied to flow now print more than the systems tied to liquidity. Look at the receipts: • @HyperliquidX: $105M fees / $95.73M revenue • @JupiterExchange: $75M fees / $18.6M revenue • @edgeX_exchange: $68.7M / $49.6M • @Lighter_xyz: $27.1M fees / $27.1M revenue • @GMX_IO — $5.1M fees / $1.8M revenue Now compare to the “old guard”: • @Uniswap: $93.4M fees • @aave v3: $85.9M fees / $10.8M revenue • @CurveFinance: $13.4M fees / $6.4M revenue The fees are huge, but none of them match perp engines on earnings efficiency. The reason is structural, not narrative: Lending = deposit behaviour AMMs = liquidity behaviour Perps = trading behaviour Trading behaviour is the only one that scales with volatility, leverage, liquidations, hedging, and momentum; the things that actually move money on-chain. That’s why perps monetize reflexive cycles while lending and swaps flatten out when activity cools. Three design truths flipped the fee stack: 1. Behaviour > deposits: Every trade is revenue. TVL becomes optional, not foundational. 2. Volatility > optimism: Perps print in both directions. AMMs and lenders do not. 3. Execution > liquidity: Latency, oracles, unified margin, and liquidation engines compound revenue more than TVL does. @HyperliquidX didn’t need incentives. @JupiterExchange didn’t need a farm. When the engine is correct, flow pays for itself. Perps aren’t an appendix to DeFi anymore. They are the fee economy. Everything else sits downstream of execution. This isn’t a phase. It’s the new structure.