I find most of the bitcoin returns are realized in three-minute windows before blocks are added to the blockchain. These minting events are natural experiments and unpredictable due to the proof-of-work mechanism. Using high-frequency event studies, I show bitcoin returns are: (1) significantly positive in the minutes leading up to the mint, and (2) zero during post-mint periods. These patterns exist in other cryptocurrencies and are robust to a variety of tests, including controls for other calendar effects, data snooping, and outliers. I consider several alternative explanations, however, none of them can fully reconcile the new facts about bitcoin prices around mining activity.