We build a model to show that decentralized exchanges (DEX) require less computing power on average than traditional exchanges to accommodate the demand for high-speed trading services. Centralized exchanges acquire excess processing capacity to accommodate activity surges: The idle capacity’s opportunity cost is an externality of low-latency trading. On DEXs, HFTs bid on gas fees in real-time to acquire time priority from a network of miners. The price of speed surges as HFTs compete during activity bursts. HFT-driven demand for speed peaks higher on DEX, but spans a shorter time interval. On aggregate, DEX infrastructure is more cost-efficient.