We serve data in milliseconds to more than 1.2 billion users across 40 geographic regions. To reach that scale, we worked closely with @AzureCosmosDB on our storage architecture, achieving a 12x improvement in global median read latency.