Cloud providers are much better at managing infrastructure, services and data stores than organisations. We want to leverage managed services as much as possible, especially for storing state. We can provision quickly, significantly reduce costs and reliably store data with effectively infinite scale. For example, CosmosDB has an SLA of 99.999% uptime (5-nines), which is the equivalent of about 5 minutes downtime per year. Data is replicated across zones and/or regions. This is much more reliable (and much cheaper) than we could ever hope to achieve if we were managing our own databases.