<p>We analyze the effect of block propagation latency on the performance and design of Nakamoto-style blockchains. Miners strategically choose block capacity, balancing the risk of invalidation from forking with transaction fee income. The model identifies a unique and symmetric Nash equilibrium block capacity, which increases with the ratio of block production time to transmission delay and decreases with the ratio of coinbase reward to transaction fee rate. We endogenize blockchain growth and derive the Master Equation for pending mempool data. The results reveal a trade-off between efficiency (low transaction load) and security (high miner participation). Reducing the coinbase reward while raising transaction fees improves efficiency but may weaken security. We also discuss testable implications and extend the model to include uncle block rewards and discrete latency.</p>