Current cryptocurrencies provide a heavily limited transaction throughput that is clearly insufficient to cater their growing adop- tion. Payment-channel networks (PCNs) have emerged as an inter- esting solution to the scalability issue and are currently deployed by popular cryptocurrencies such as Bitcoin and Ethereum. While PCNs do increase the transaction throughput by processing pay- ments off-chain and using the blockchain only as a dispute arbitra- tor, they unfortunately require high collateral (i.e., they lock coins for a non-constant time along the payment path) and are restricted to payments in a path from sender to receiver. These issues have severe consequences in practice. The high collateral enables denial- of-service attacks that hamper the throughput and utility of the PCN. Moreover, the limited functionality hinders the applicability of current PCNs in many important application scenarios. Unfortu- nately, current proposals do not solve either of these issues, or they require Turing-complete language support, which severely limit their applicability. In this work, we present AMCU, the first protocol for atomic multi-channel updates and reduced collateral that is compatible with Bitcoin (and other cryptocurrencies with reduced scripting ca- pabilities). We provide a formal model in the Universal Composabil- ity framework and show that AMCU realizes it, thus demonstrating that AMCU achieves atomicity and value privacy. Moreover, the reduced collateral mitigates the consequences of griefing attacks in PCNs while the (multi-payment) atomicity achieved by AMCU opens the door to new applications such as credit rebalancing and crowdfunding that are not possible otherwise. Moreover, our eval- uation results demonstrate that AMCU has a performance in line with that of the Lightning Network (the most widely deployed PCN) and thus is ready to be deployed in practice.